Update to igb and em:

em revision 7.0.0:
	- Using driver devclass, seperate legacy (pre-pcie) code
	  into a seperate source file. This will at least help
	  protect against regression issues. It compiles along
	  with em, and is transparent to end use, devices in each
	  appear to be 'emX'. When using em in a modular form this
	  also allows the legacy stuff to be defined out.
	- Add tx and rx rings as in igb, in the 82574 this becomes
	  actual multiqueue for the first time (2 queues) while in
	  other PCIE adapters its just make code cleaner.
	- Add RX mbuf handling logic that matches igb, this will
	  eliminate packet drops due to temporary mbuf shortage.

igb revision 1.9.3:
	- Following the ixgbe code, use a new approach in what
	  was called 'get_buf', the routine now has been made
	  independent of rxeof, it now does the update to the
	  engine TDT register, this design allows temporary
	  mbuf resources to become non-critical, not requiring
	  a packet to be discarded, instead it just returns and
	  does not increment the tail pointer.
	- With the above change it was also unnecessary to keep
	  'spare' maps around, since we do not have the discard
	  issue.
	- Performance tweaks and improvements to the code also.

MFC in a week

1 files changed, 100 insertions, 131 deletions

diff --git a/sys/dev/e1000/if_em.h b/sys/dev/e1000/if_em.h
index fe5a99b..d3a1019 100644
--- a/sys/dev/e1000/if_em.h
+++ b/sys/dev/e1000/if_em.h
@@ -52,7 +52,6 @@
  *      (num_tx_desc * sizeof(struct e1000_tx_desc)) % 128 == 0
  */
 #define EM_MIN_TXD		80
-#define EM_MAX_TXD_82543	256
 #define EM_MAX_TXD		4096
 #define EM_DEFAULT_TXD		1024
 
@@ -70,7 +69,6 @@
  *      (num_tx_desc * sizeof(struct e1000_tx_desc)) % 128 == 0
  */
 #define EM_MIN_RXD		80
-#define EM_MAX_RXD_82543	256
 #define EM_MAX_RXD		4096
 #define EM_DEFAULT_RXD		1024
 
@@ -144,7 +142,6 @@
  * transmit descriptors.
  */
 #define EM_TX_CLEANUP_THRESHOLD	(adapter->num_tx_desc / 8)
-#define EM_TX_OP_THRESHOLD	(adapter->num_tx_desc / 32)
 
 /*
  * This parameter controls whether or not autonegotation is enabled.
@@ -182,7 +179,7 @@
 #define EM_DEFAULT_PBA                  0x00000030
 #define EM_SMARTSPEED_DOWNSHIFT         3
 #define EM_SMARTSPEED_MAX               15
-#define EM_MAX_INTR			10
+#define EM_MAX_LOOP			10
 
 #define MAX_NUM_MULTICAST_ADDRESSES     128
 #define PCI_ANY_ID                      (~0U)
@@ -191,11 +188,6 @@
 #define EM_EEPROM_APME			0x400;
 #define EM_82544_APME			0x0004;
 
-/* Code compatilbility between 6 and 7 */
-#ifndef ETHER_BPF_MTAP
-#define ETHER_BPF_MTAP			BPF_MTAP
-#endif
-
 /*
  * TDBA/RDBA should be aligned on 16 byte boundary. But TDLEN/RDLEN should be
  * multiple of 128 bytes. So we align TDBA/RDBA on 128 byte boundary. This will
@@ -209,7 +201,6 @@
 #define EM_BAR_TYPE(v)		((v) & EM_BAR_TYPE_MASK)
 #define EM_BAR_TYPE_MASK	0x00000001
 #define EM_BAR_TYPE_MMEM	0x00000000
-#define EM_BAR_TYPE_IO		0x00000001
 #define EM_BAR_TYPE_FLASH	0x0014 
 #define EM_BAR_MEM_TYPE(v)	((v) & EM_BAR_MEM_TYPE_MASK)
 #define EM_BAR_MEM_TYPE_MASK	0x00000006
@@ -237,6 +228,7 @@
 #define EM_TSO_SIZE		(65535 + sizeof(struct ether_vlan_header))
 #define EM_TSO_SEG_SIZE		4096	/* Max dma segment size */
 #define EM_MSIX_MASK		0x01F00000 /* For 82574 use */
+#define EM_MSIX_LINK		0x01000000 /* For 82574 use */
 #define ETH_ZLEN		60
 #define ETH_ADDR_LEN		6
 #define CSUM_OFFLOAD		7	/* Offload bits in mbuf flag */
@@ -249,18 +241,6 @@
  */
 #define EM_EIAC 0x000DC
 
-/* Used in for 82547 10Mb Half workaround */
-#define EM_PBA_BYTES_SHIFT	0xA
-#define EM_TX_HEAD_ADDR_SHIFT	7
-#define EM_PBA_TX_MASK		0xFFFF0000
-#define EM_FIFO_HDR		0x10
-#define EM_82547_PKT_THRESH	0x3e0
-
-/* Precision Time Sync (IEEE 1588) defines */
-#define ETHERTYPE_IEEE1588	0x88F7
-#define PICOSECS_PER_TICK	20833
-#define TSYNC_PORT		319 /* UDP port for the protocol */
-
 /*
  * Bus dma allocation structure used by
  * e1000_dma_malloc and e1000_dma_free.
@@ -282,12 +262,76 @@ struct em_int_delay_info {
 	int value;			/* Current value in usecs */
 };
 
+/*
+ * The transmit ring, one per tx queue
+ */
+struct tx_ring {
+        struct adapter          *adapter;
+        struct mtx              tx_mtx;
+        char                    mtx_name[16];
+        u32                     me;
+        u32                     msix;
+	u32			ims;
+        bool                    watchdog_check;
+        int                     watchdog_time;
+	struct em_dma_alloc	txdma;
+	struct e1000_tx_desc	*tx_base;
+        struct task             tx_task;
+        struct taskqueue        *tq;
+        u32                     next_avail_desc;
+        u32                     next_to_clean;
+        struct em_buffer	*tx_buffers;
+        volatile u16            tx_avail;
+	u32			tx_tso;		/* last tx was tso */
+        u16			last_hw_offload;
+#if __FreeBSD_version >= 800000
+	struct buf_ring         *br;
+#endif
+	/* Interrupt resources */
+        bus_dma_tag_t           txtag;
+	void                    *tag;
+	struct resource         *res;
+        u64                     tx_irq;
+        u64                     no_desc_avail;
+};
+
+/*
+ * The Receive ring, one per rx queue
+ */
+struct rx_ring {
+        struct adapter          *adapter;
+        u32                     me;
+        u32                     msix;
+	u32			ims;
+        struct mtx              rx_mtx;
+        char                    mtx_name[16];
+        u32                     payload;
+        struct task             rx_task;
+        struct taskqueue        *tq;
+        struct e1000_rx_desc	*rx_base;
+        struct em_dma_alloc	rxdma;
+        unsigned int            next_to_refresh;
+        unsigned int            next_to_check;
+        struct em_buffer	*rx_buffers;
+	struct mbuf		*fmp;
+	struct mbuf		*lmp;
+
+        /* Interrupt resources */
+        void                    *tag;
+        struct resource         *res;
+        bus_dma_tag_t           rxtag;
+        bus_dmamap_t            rx_sparemap;
+
+        /* Soft stats */
+        u64                     rx_irq;
+        u64                     rx_packets;
+        u64                     rx_bytes;
+};
+
+
 /* Our adapter structure */
 struct adapter {
 	struct ifnet	*ifp;
-#if __FreeBSD_version >= 800000
-	struct buf_ring	*br;
-#endif
 	struct e1000_hw	hw;
 
 	/* FreeBSD operating-system-specific structures. */
@@ -296,42 +340,50 @@ struct adapter {
 
 	struct resource *memory;
 	struct resource *flash;
-	struct resource *msix;
+	struct resource *msix_mem;
 
-	struct resource	*ioport;
-	int		io_rid;
-
-	/* 82574 may use 3 int vectors */
-	struct resource	*res[3];
-	void		*tag[3];
-	int		rid[3];
+	struct resource	*res;
+	void		*tag;
+	u32		linkvec;
+	u32		ivars;
 
 	struct ifmedia	media;
 	struct callout	timer;
-	struct callout	tx_fifo_timer;
-	bool		watchdog_check;
-	int		watchdog_time;
-	int		msi;
+	int		msix;
 	int		if_flags;
 	int		max_frame_size;
 	int		min_frame_size;
 	struct mtx	core_mtx;
-	struct mtx	tx_mtx;
-	struct mtx	rx_mtx;
 	int		em_insert_vlan_header;
+	u32		ims;
+	bool		in_detach;
 
 	/* Task for FAST handling */
 	struct task     link_task;
-	struct task     rxtx_task;
-	struct task     rx_task;
-	struct task     tx_task;
+	struct task     que_task;
 	struct taskqueue *tq;           /* private task queue */
 
-#if __FreeBSD_version >= 700029
 	eventhandler_tag vlan_attach;
 	eventhandler_tag vlan_detach;
-	u32	num_vlans;
-#endif
+
+	u16	num_vlans;
+	u16	num_queues;
+
+        /*
+         * Transmit rings:
+         *      Allocated at run time, an array of rings.
+         */
+        struct tx_ring  *tx_rings;
+        int             num_tx_desc;
+        u32		txd_cmd;
+
+        /*
+         * Receive rings:
+         *      Allocated at run time, an array of rings.
+         */
+        struct rx_ring  *rx_rings;
+        int             num_rx_desc;
+        u32             rx_process_limit;
 
 	/* Management and WOL features */
 	u32		wol;
@@ -348,96 +400,26 @@ struct adapter {
 	struct em_int_delay_info rx_int_delay;
 	struct em_int_delay_info rx_abs_int_delay;
 
-	/*
-	 * Transmit definitions
-	 *
-	 * We have an array of num_tx_desc descriptors (handled
-	 * by the controller) paired with an array of tx_buffers
-	 * (at tx_buffer_area).
-	 * The index of the next available descriptor is next_avail_tx_desc.
-	 * The number of remaining tx_desc is num_tx_desc_avail.
-	 */
-	struct em_dma_alloc	txdma;		/* bus_dma glue for tx desc */
-	struct e1000_tx_desc	*tx_desc_base;
-	uint32_t		next_avail_tx_desc;
-	uint32_t		next_tx_to_clean;
-	volatile uint16_t	num_tx_desc_avail;
-        uint16_t		num_tx_desc;
-        uint16_t		last_hw_offload;
-        uint32_t		txd_cmd;
-	struct em_buffer	*tx_buffer_area;
-	bus_dma_tag_t		txtag;		/* dma tag for tx */
-	uint32_t	   	tx_tso;		/* last tx was tso */
-
-	/* 
-	 * Receive definitions
-	 *
-	 * we have an array of num_rx_desc rx_desc (handled by the
-	 * controller), and paired with an array of rx_buffers
-	 * (at rx_buffer_area).
-	 * The next pair to check on receive is at offset next_rx_desc_to_check
-	 */
-	struct em_dma_alloc	rxdma;		/* bus_dma glue for rx desc */
-	struct e1000_rx_desc	*rx_desc_base;
-	uint32_t		next_rx_desc_to_check;
-	uint32_t		rx_buffer_len;
-	uint16_t		num_rx_desc;
-	int			rx_process_limit;
-	struct em_buffer	*rx_buffer_area;
-	bus_dma_tag_t		rxtag;
-	bus_dmamap_t		rx_sparemap;
-
-	/*
-	 * First/last mbuf pointers, for
-	 * collecting multisegment RX packets.
-	 */
-	struct mbuf	       *fmp;
-	struct mbuf	       *lmp;
-
 	/* Misc stats maintained by the driver */
 	unsigned long	dropped_pkts;
 	unsigned long	mbuf_alloc_failed;
 	unsigned long	mbuf_cluster_failed;
-	unsigned long	no_tx_desc_avail1;
-	unsigned long	no_tx_desc_avail2;
 	unsigned long	no_tx_map_avail;
         unsigned long	no_tx_dma_setup;
-	unsigned long	watchdog_events;
 	unsigned long	rx_overruns;
-	unsigned long	rx_irq;
-	unsigned long	tx_irq;
+	unsigned long	watchdog_events;
 	unsigned long	link_irq;
 
-	/* 82547 workaround */
-	uint32_t	tx_fifo_size;
-	uint32_t	tx_fifo_head;
-	uint32_t	tx_fifo_head_addr;
-	uint64_t	tx_fifo_reset_cnt;
-	uint64_t	tx_fifo_wrk_cnt;
-	uint32_t	tx_head_addr;
-
-        /* For 82544 PCIX Workaround */
-	boolean_t       pcix_82544;
-	boolean_t       in_detach;
-
-#ifdef EM_IEEE1588
-	/* IEEE 1588 precision time support */
-	struct cyclecounter     cycles;
-	struct nettimer         clock;
-	struct nettime_compare  compare;
-	struct hwtstamp_ctrl    hwtstamp;
-#endif
-
 	struct e1000_hw_stats stats;
 };
 
-/* ******************************************************************************
+/********************************************************************************
  * vendor_info_array
  *
  * This array contains the list of Subvendor/Subdevice IDs on which the driver
  * should load.
  *
- * ******************************************************************************/
+ ********************************************************************************/
 typedef struct _em_vendor_info_t {
 	unsigned int vendor_id;
 	unsigned int device_id;
@@ -452,19 +434,6 @@ struct em_buffer {
         bus_dmamap_t    map;         /* bus_dma map for packet */
 };
 
-/* For 82544 PCIX  Workaround */
-typedef struct _ADDRESS_LENGTH_PAIR
-{
-	uint64_t   address;
-	uint32_t   length;
-} ADDRESS_LENGTH_PAIR, *PADDRESS_LENGTH_PAIR;
-
-typedef struct _DESCRIPTOR_PAIR
-{
-	ADDRESS_LENGTH_PAIR descriptor[4];
-	uint32_t   elements;
-} DESC_ARRAY, *PDESC_ARRAY;
-
 #define	EM_CORE_LOCK_INIT(_sc, _name) \
 	mtx_init(&(_sc)->core_mtx, _name, "EM Core Lock", MTX_DEF)
 #define	EM_TX_LOCK_INIT(_sc, _name) \