Skip to contentSkip to conversation

updated cxgb4 submission

PATCH In review Dimitris Michailidis 4 participants 18 messages Mar 30, 2010 – Apr 1, 2010

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover updated cxgb4 submission

This is an updated patch series for the cxgb4 driver. Since the previous posting there have been some fixes, simplifications, and updates here and there but nothing extensive in any particular area.

The patches are against net-next as of 36c0d8c.

 drivers/net/Kconfig            |   25 +
 drivers/net/Makefile           |    1 +
 drivers/net/cxgb4/Makefile     |    7 +
 drivers/net/cxgb4/cxgb4.h      |  743 +++++++
 drivers/net/cxgb4/cxgb4_main.c | 4240 ++++++++++++++++++++++++++++++++++++++++
 drivers/net/cxgb4/cxgb4_uld.h  |  235 +++
 drivers/net/cxgb4/l2t.c        |  626 ++++++
 drivers/net/cxgb4/l2t.h        |  110 ++
 drivers/net/cxgb4/sge.c        | 2431 +++++++++++++++++++++++
 drivers/net/cxgb4/t4_hw.c      | 3131 +++++++++++++++++++++++++++++
 drivers/net/cxgb4/t4_hw.h      |  100 +
 drivers/net/cxgb4/t4_msg.h     |  664 +++++++
 drivers/net/cxgb4/t4_regs.h    |  878 +++++++++
 drivers/net/cxgb4/t4fw_api.h   | 1584 +++++++++++++++
 14 files changed, 14775 insertions(+), 0 deletions(-)
 create mode 100644 drivers/net/cxgb4/Makefile
 create mode 100644 drivers/net/cxgb4/cxgb4.h
 create mode 100644 drivers/net/cxgb4/cxgb4_main.c
 create mode 100644 drivers/net/cxgb4/cxgb4_uld.h
 create mode 100644 drivers/net/cxgb4/l2t.c
 create mode 100644 drivers/net/cxgb4/l2t.h
 create mode 100644 drivers/net/cxgb4/sge.c
 create mode 100644 drivers/net/cxgb4/t4_hw.c
 create mode 100644 drivers/net/cxgb4/t4_hw.h
 create mode 100644 drivers/net/cxgb4/t4_msg.h
 create mode 100644 drivers/net/cxgb4/t4_regs.h
 create mode 100644 drivers/net/cxgb4/t4fw_api.h
#1 lore raw

PATCH 1/6 cxgb4: Add register, message, and FW definitions

---
 drivers/net/cxgb4/t4_msg.h   |  664 ++++++++++++++++++
 drivers/net/cxgb4/t4_regs.h  |  878 +++++++++++++++++++++++
 drivers/net/cxgb4/t4fw_api.h | 1584 ++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 3126 insertions(+), 0 deletions(-)
 create mode 100644 drivers/net/cxgb4/t4_msg.h
 create mode 100644 drivers/net/cxgb4/t4_regs.h
 create mode 100644 drivers/net/cxgb4/t4fw_api.h
drivers/net/cxgb4/t4_msg.h new file +664 -0
new file mode 100644
index 0000000..6df2b92
@@ -0,0 +1,664 @@
+/*
+ * This file is part of the Chelsio T4 Ethernet driver for Linux.
+ *
+ * Copyright (c) 2003-2010 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.
+ */
+
+#ifndef __T4_MSG_H
+#define __T4_MSG_H
+
+#include <linux/types.h>
+
+enum {
+	CPL_PASS_OPEN_REQ     = 0x1,
+	CPL_PASS_ACCEPT_RPL   = 0x2,
+	CPL_ACT_OPEN_REQ      = 0x3,
+	CPL_SET_TCB_FIELD     = 0x5,
+	CPL_GET_TCB           = 0x6,
+	CPL_CLOSE_CON_REQ     = 0x8,
+	CPL_CLOSE_LISTSRV_REQ = 0x9,
+	CPL_ABORT_REQ         = 0xA,
+	CPL_ABORT_RPL         = 0xB,
+	CPL_RX_DATA_ACK       = 0xD,
+	CPL_TX_PKT            = 0xE,
+	CPL_L2T_WRITE_REQ     = 0x12,
+	CPL_TID_RELEASE       = 0x1A,
+
+	CPL_CLOSE_LISTSRV_RPL = 0x20,
+	CPL_L2T_WRITE_RPL     = 0x23,
+	CPL_PASS_OPEN_RPL     = 0x24,
+	CPL_ACT_OPEN_RPL      = 0x25,
+	CPL_PEER_CLOSE        = 0x26,
+	CPL_ABORT_REQ_RSS     = 0x2B,
+	CPL_ABORT_RPL_RSS     = 0x2D,
+
+	CPL_CLOSE_CON_RPL     = 0x32,
+	CPL_ISCSI_HDR         = 0x33,
+	CPL_RDMA_CQE          = 0x35,
+	CPL_RDMA_CQE_READ_RSP = 0x36,
+	CPL_RDMA_CQE_ERR      = 0x37,
+	CPL_RX_DATA           = 0x39,
+	CPL_SET_TCB_RPL       = 0x3A,
+	CPL_RX_PKT            = 0x3B,
+	CPL_RX_DDP_COMPLETE   = 0x3F,
+
+	CPL_ACT_ESTABLISH     = 0x40,
+	CPL_PASS_ESTABLISH    = 0x41,
+	CPL_RX_DATA_DDP       = 0x42,
+	CPL_PASS_ACCEPT_REQ   = 0x44,
+
+	CPL_RDMA_READ_REQ     = 0x60,
+
+	CPL_PASS_OPEN_REQ6    = 0x81,
+	CPL_ACT_OPEN_REQ6     = 0x83,
+
+	CPL_RDMA_TERMINATE    = 0xA2,
+	CPL_RDMA_WRITE        = 0xA4,
+	CPL_SGE_EGR_UPDATE    = 0xA5,
+
+	CPL_TRACE_PKT         = 0xB0,
+
+	CPL_FW4_MSG           = 0xC0,
+	CPL_FW4_PLD           = 0xC1,
+	CPL_FW4_ACK           = 0xC3,
+
+	CPL_FW6_MSG           = 0xE0,
+	CPL_FW6_PLD           = 0xE1,
+	CPL_TX_PKT_LSO        = 0xED,
+	CPL_TX_PKT_XT         = 0xEE,
+
+	NUM_CPL_CMDS
+};
+
+enum CPL_error {
+	CPL_ERR_NONE               = 0,
+	CPL_ERR_TCAM_FULL          = 3,
+	CPL_ERR_BAD_LENGTH         = 15,
+	CPL_ERR_BAD_ROUTE          = 18,
+	CPL_ERR_CONN_RESET         = 20,
+	CPL_ERR_CONN_EXIST_SYNRECV = 21,
+	CPL_ERR_CONN_EXIST         = 22,
+	CPL_ERR_ARP_MISS           = 23,
+	CPL_ERR_BAD_SYN            = 24,
+	CPL_ERR_CONN_TIMEDOUT      = 30,
+	CPL_ERR_XMIT_TIMEDOUT      = 31,
+	CPL_ERR_PERSIST_TIMEDOUT   = 32,
+	CPL_ERR_FINWAIT2_TIMEDOUT  = 33,
+	CPL_ERR_KEEPALIVE_TIMEDOUT = 34,
+	CPL_ERR_RTX_NEG_ADVICE     = 35,
+	CPL_ERR_PERSIST_NEG_ADVICE = 36,
+	CPL_ERR_ABORT_FAILED       = 42,
+	CPL_ERR_IWARP_FLM          = 50,
+};
+
+enum {
+	ULP_MODE_NONE          = 0,
+	ULP_MODE_ISCSI         = 2,
+	ULP_MODE_RDMA          = 4,
+	ULP_MODE_FCOE          = 6,
+};
+
+enum {
+	ULP_CRC_HEADER = 1 << 0,
+	ULP_CRC_DATA   = 1 << 1
+};
+
+enum {
+	CPL_ABORT_SEND_RST = 0,
+	CPL_ABORT_NO_RST,
+};
+
+enum {                     /* TX_PKT_XT checksum types */
+	TX_CSUM_TCP    = 0,
+	TX_CSUM_UDP    = 1,
+	TX_CSUM_CRC16  = 4,
+	TX_CSUM_CRC32  = 5,
+	TX_CSUM_CRC32C = 6,
+	TX_CSUM_FCOE   = 7,
+	TX_CSUM_TCPIP  = 8,
+	TX_CSUM_UDPIP  = 9,
+	TX_CSUM_TCPIP6 = 10,
+	TX_CSUM_UDPIP6 = 11,
+	TX_CSUM_IP     = 12,
+};
+
+union opcode_tid {
+	__be32 opcode_tid;
+	__u8 opcode;
+};
+
+#define CPL_OPCODE(x) ((x) << 24)
+#define MK_OPCODE_TID(opcode, tid) (CPL_OPCODE(opcode) | (tid))
+#define OPCODE_TID(cmd) ((cmd)->ot.opcode_tid)
+#define GET_TID(cmd) (ntohl(OPCODE_TID(cmd)) & 0xFFFFFF)
+
+/* partitioning of TID fields that also carry a queue id */
+#define GET_TID_TID(x) ((x) & 0x3fff)
+#define GET_TID_QID(x) (((x) >> 14) & 0x3ff)
+#define TID_QID(x)     ((x) << 14)
+
+struct rss_header {
+	__u8 opcode;
+#if defined(__LITTLE_ENDIAN_BITFIELD)
+	__u8 channel:2;
+	__u8 filter_hit:1;
+	__u8 filter_tid:1;
+	__u8 hash_type:2;
+	__u8 ipv6:1;
+	__u8 send2fw:1;
+#else
+	__u8 send2fw:1;
+	__u8 ipv6:1;
+	__u8 hash_type:2;
+	__u8 filter_tid:1;
+	__u8 filter_hit:1;
+	__u8 channel:2;
+#endif
+	__be16 qid;
+	__be32 hash_val;
+};
+
+struct work_request_hdr {
+	__be32 wr_hi;
+	__be32 wr_mid;
+	__be64 wr_lo;
+};
+
+#define WR_HDR struct work_request_hdr wr
+
+struct cpl_pass_open_req {
+	WR_HDR;
+	union opcode_tid ot;
+	__be16 local_port;
+	__be16 peer_port;
+	__be32 local_ip;
+	__be32 peer_ip;
+	__be64 opt0;
+#define TX_CHAN(x)    ((x) << 2)
+#define DELACK(x)     ((x) << 5)
+#define ULP_MODE(x)   ((x) << 8)
+#define RCV_BUFSIZ(x) ((x) << 12)
+#define DSCP(x)       ((x) << 22)
+#define SMAC_SEL(x)   ((__u64)(x) << 28)
+#define L2T_IDX(x)    ((__u64)(x) << 36)
+#define NAGLE(x)      ((__u64)(x) << 49)
+#define WND_SCALE(x)  ((__u64)(x) << 50)
+#define KEEP_ALIVE(x) ((__u64)(x) << 54)
+#define MSS_IDX(x)    ((__u64)(x) << 60)
+	__be64 opt1;
+#define SYN_RSS_ENABLE   (1 << 0)
+#define SYN_RSS_QUEUE(x) ((x) << 2)
+#define CONN_POLICY_ASK  (1 << 22)
+};
+
+struct cpl_pass_open_req6 {
+	WR_HDR;
+	union opcode_tid ot;
+	__be16 local_port;
+	__be16 peer_port;
+	__be64 local_ip_hi;
+	__be64 local_ip_lo;
+	__be64 peer_ip_hi;
+	__be64 peer_ip_lo;
+	__be64 opt0;
+	__be64 opt1;
+};
+
+struct cpl_pass_open_rpl {
+	union opcode_tid ot;
+	__u8 rsvd[3];
+	__u8 status;
+};
+
+struct cpl_pass_accept_rpl {
+	WR_HDR;
+	union opcode_tid ot;
+	__be32 opt2;
+#define RSS_QUEUE(x)         ((x) << 0)
+#define RSS_QUEUE_VALID      (1 << 10)
+#define RX_COALESCE_VALID(x) ((x) << 11)
+#define RX_COALESCE(x)       ((x) << 12)
+#define TX_QUEUE(x)          ((x) << 23)
+#define RX_CHANNEL(x)        ((x) << 26)
+#define WND_SCALE_EN(x)      ((x) << 28)
+#define TSTAMPS_EN(x)        ((x) << 29)
+#define SACK_EN(x)           ((x) << 30)
+	__be64 opt0;
+};
+
+struct cpl_act_open_req {
+	WR_HDR;
+	union opcode_tid ot;
+	__be16 local_port;
+	__be16 peer_port;
+	__be32 local_ip;
+	__be32 peer_ip;
+	__be64 opt0;
+	__be32 params;
+	__be32 opt2;
+};
+
+struct cpl_act_open_req6 {
+	WR_HDR;
+	union opcode_tid ot;
+	__be16 local_port;
+	__be16 peer_port;
+	__be64 local_ip_hi;
+	__be64 local_ip_lo;
+	__be64 peer_ip_hi;
+	__be64 peer_ip_lo;
+	__be64 opt0;
+	__be32 params;
+	__be32 opt2;
+};
+
+struct cpl_act_open_rpl {
+	union opcode_tid ot;
+	__be32 atid_status;
+#define GET_AOPEN_STATUS(x) ((x) & 0xff)
+#define GET_AOPEN_ATID(x)   (((x) >> 8) & 0xffffff)
+};
+
+struct cpl_pass_establish {
+	union opcode_tid ot;
+	__be32 rsvd;
+	__be32 tos_stid;
+#define GET_POPEN_TID(x) ((x) & 0xffffff)
+#define GET_POPEN_TOS(x) (((x) >> 24) & 0xff)
+	__be16 mac_idx;
+	__be16 tcp_opt;
+#define GET_TCPOPT_WSCALE_OK(x)  (((x) >> 5) & 1)
+#define GET_TCPOPT_SACK(x)       (((x) >> 6) & 1)
+#define GET_TCPOPT_TSTAMP(x)     (((x) >> 7) & 1)
+#define GET_TCPOPT_SND_WSCALE(x) (((x) >> 8) & 0xf)
+#define GET_TCPOPT_MSS(x)        (((x) >> 12) & 0xf)
+	__be32 snd_isn;
+	__be32 rcv_isn;
+};
+
+struct cpl_act_establish {
+	union opcode_tid ot;
+	__be32 rsvd;
+	__be32 tos_atid;
+	__be16 mac_idx;
+	__be16 tcp_opt;
+	__be32 snd_isn;
+	__be32 rcv_isn;
+};
+
+struct cpl_get_tcb {
+	WR_HDR;
+	union opcode_tid ot;
+	__be16 reply_ctrl;
+#define QUEUENO(x)    ((x) << 0)
+#define REPLY_CHAN(x) ((x) << 14)
+#define NO_REPLY(x)   ((x) << 15)
+	__be16 cookie;
+};
+
+struct cpl_set_tcb_field {
+	WR_HDR;
+	union opcode_tid ot;
+	__be16 reply_ctrl;
+	__be16 word_cookie;
+#define TCB_WORD(x)   ((x) << 0)
+#define TCB_COOKIE(x) ((x) << 5)
+	__be64 mask;
+	__be64 val;
+};
+
+struct cpl_set_tcb_rpl {
+	union opcode_tid ot;
+	__be16 rsvd;
+	__u8   cookie;
+	__u8   status;
+	__be64 oldval;
+};
+
+struct cpl_close_con_req {
+	WR_HDR;
+	union opcode_tid ot;
+	__be32 rsvd;
+};
+
+struct cpl_close_con_rpl {
+	union opcode_tid ot;
+	__u8  rsvd[3];
+	__u8  status;
+	__be32 snd_nxt;
+	__be32 rcv_nxt;
+};
+
+struct cpl_close_listsvr_req {
+	WR_HDR;
+	union opcode_tid ot;
+	__be16 reply_ctrl;
+#define LISTSVR_IPV6 (1 << 14)
+	__be16 rsvd;
+};
+
+struct cpl_close_listsvr_rpl {
+	union opcode_tid ot;
+	__u8 rsvd[3];
+	__u8 status;
+};
+
+struct cpl_abort_req_rss {
+	union opcode_tid ot;
+	__u8  rsvd[3];
+	__u8  status;
+};
+
+struct cpl_abort_req {
+	WR_HDR;
+	union opcode_tid ot;
+	__be32 rsvd0;
+	__u8  rsvd1;
+	__u8  cmd;
+	__u8  rsvd2[6];
+};
+
+struct cpl_abort_rpl_rss {
+	union opcode_tid ot;
+	__u8  rsvd[3];
+	__u8  status;
+};
+
+struct cpl_abort_rpl {
+	WR_HDR;
+	union opcode_tid ot;
+	__be32 rsvd0;
+	__u8  rsvd1;
+	__u8  cmd;
+	__u8  rsvd2[6];
+};
+
+struct cpl_peer_close {
+	union opcode_tid ot;
+	__be32 rcv_nxt;
+};
+
+struct cpl_tid_release {
+	WR_HDR;
+	union opcode_tid ot;
+	__be32 rsvd;
+};
+
+struct cpl_tx_pkt_core {
+	__be32 ctrl0;
+#define TXPKT_VF(x)        ((x) << 0)
+#define TXPKT_PF(x)        ((x) << 8)
+#define TXPKT_VF_VLD       (1 << 11)
+#define TXPKT_OVLAN_IDX(x) ((x) << 12)
+#define TXPKT_INTF(x)      ((x) << 16)
+#define TXPKT_INS_OVLAN    (1 << 21)
+#define TXPKT_OPCODE(x)    ((x) << 24)
+	__be16 pack;
+	__be16 len;
+	__be64 ctrl1;
+#define TXPKT_CSUM_END(x)   ((x) << 12)
+#define TXPKT_CSUM_START(x) ((x) << 20)
+#define TXPKT_IPHDR_LEN(x)  ((__u64)(x) << 20)
+#define TXPKT_CSUM_LOC(x)   ((__u64)(x) << 30)
+#define TXPKT_ETHHDR_LEN(x) ((__u64)(x) << 34)
+#define TXPKT_CSUM_TYPE(x)  ((__u64)(x) << 40)
+#define TXPKT_VLAN(x)       ((__u64)(x) << 44)
+#define TXPKT_VLAN_VLD      (1ULL << 60)
+#define TXPKT_IPCSUM_DIS    (1ULL << 62)
+#define TXPKT_L4CSUM_DIS    (1ULL << 63)
+};
+
+struct cpl_tx_pkt {
+	WR_HDR;
+	struct cpl_tx_pkt_core c;
+};
+
+#define cpl_tx_pkt_xt cpl_tx_pkt
+
+struct cpl_tx_pkt_lso {
+	WR_HDR;
+	__be32 lso_ctrl;
+#define LSO_TCPHDR_LEN(x) ((x) << 0)
+#define LSO_IPHDR_LEN(x)  ((x) << 4)
+#define LSO_ETHHDR_LEN(x) ((x) << 16)
+#define LSO_IPV6(x)       ((x) << 20)
+#define LSO_LAST_SLICE    (1 << 22)
+#define LSO_FIRST_SLICE   (1 << 23)
+#define LSO_OPCODE(x)     ((x) << 24)
+	__be16 ipid_ofst;
+	__be16 mss;
+	__be32 seqno_offset;
+	__be32 len;
+	/* encapsulated CPL (TX_PKT, TX_PKT_XT or TX_DATA) follows here */
+};
+
+struct cpl_iscsi_hdr {
+	union opcode_tid ot;
+	__be16 pdu_len_ddp;
+#define ISCSI_PDU_LEN(x) ((x) & 0x7FFF)
+#define ISCSI_DDP        (1 << 15)
+	__be16 len;
+	__be32 seq;
+	__be16 urg;
+	__u8 rsvd;
+	__u8 status;
+};
+
+struct cpl_rx_data {
+	union opcode_tid ot;
+	__be16 rsvd;
+	__be16 len;
+	__be32 seq;
+	__be16 urg;
+#if defined(__LITTLE_ENDIAN_BITFIELD)
+	__u8 dack_mode:2;
+	__u8 psh:1;
+	__u8 heartbeat:1;
+	__u8 ddp_off:1;
+	__u8 :3;
+#else
+	__u8 :3;
+	__u8 ddp_off:1;
+	__u8 heartbeat:1;
+	__u8 psh:1;
+	__u8 dack_mode:2;
+#endif
+	__u8 status;
+};
+
+struct cpl_rx_data_ack {
+	WR_HDR;
+	union opcode_tid ot;
+	__be32 credit_dack;
+#define RX_CREDITS(x)   ((x) << 0)
+#define RX_FORCE_ACK(x) ((x) << 28)
+};
+
+struct cpl_rx_pkt {
+	__u8 opcode;
+#if defined(__LITTLE_ENDIAN_BITFIELD)
+	__u8 iff:4;
+	__u8 csum_calc:1;
+	__u8 ipmi_pkt:1;
+	__u8 vlan_ex:1;
+	__u8 ip_frag:1;
+#else
+	__u8 ip_frag:1;
+	__u8 vlan_ex:1;
+	__u8 ipmi_pkt:1;
+	__u8 csum_calc:1;
+	__u8 iff:4;
+#endif
+	__be16 csum;
+	__be16 vlan;
+	__be16 len;
+	__be32 l2info;
+#define RXF_UDP (1 << 22)
+#define RXF_TCP (1 << 23)
+	__be16 hdr_len;
+	__be16 err_vec;
+};
+
+struct cpl_trace_pkt {
+	__u8 opcode;
+	__u8 intf;
+#if defined(__LITTLE_ENDIAN_BITFIELD)
+	__u8 runt:4;
+	__u8 filter_hit:4;
+	__u8 :6;
+	__u8 err:1;
+	__u8 trunc:1;
+#else
+	__u8 filter_hit:4;
+	__u8 runt:4;
+	__u8 trunc:1;
+	__u8 err:1;
+	__u8 :6;
+#endif
+	__be16 rsvd;
+	__be16 len;
+	__be64 tstamp;
+};
+
+struct cpl_l2t_write_req {
+	WR_HDR;
+	union opcode_tid ot;
+	__be16 params;
+#define L2T_W_INFO(x)    ((x) << 2)
+#define L2T_W_PORT(x)    ((x) << 8)
+#define L2T_W_NOREPLY(x) ((x) << 15)
+	__be16 l2t_idx;
+	__be16 vlan;
+	__u8   dst_mac[6];
+};
+
+struct cpl_l2t_write_rpl {
+	union opcode_tid ot;
+	__u8 status;
+	__u8 rsvd[3];
+};
+
+struct cpl_rdma_terminate {
+	union opcode_tid ot;
+	__be16 rsvd;
+	__be16 len;
+};
+
+struct cpl_sge_egr_update {
+	__be32 opcode_qid;
+#define EGR_QID(x) ((x) & 0x1FFFF)
+	__be16 cidx;
+	__be16 pidx;
+};
+
+struct cpl_fw4_pld {
+	u8 opcode;
+	u8 rsvd0[3];
+	u8 type;
+	u8 rsvd1;
+	__be16 len;
+	__be64 data;
+	__be64 rsvd2;
+};
+
+struct cpl_fw6_pld {
+	u8 opcode;
+	u8 rsvd[5];
+	__be16 len;
+	__be64 data[4];
+};
+
+struct cpl_fw4_msg {
+	u8 opcode;
+	u8 type;
+	__be16 rsvd0;
+	__be32 rsvd1;
+	__be64 data[2];
+};
+
+struct cpl_fw4_ack {
+	union opcode_tid ot;
+	u8 credits;
+	u8 rsvd0[2];
+	u8 seq_vld;
+	__be32 snd_nxt;
+	__be32 snd_una;
+	__be64 rsvd1;
+};
+
+struct cpl_fw6_msg {
+	u8 opcode;
+	u8 type;
+	__be16 rsvd0;
+	__be32 rsvd1;
+	__be64 data[4];
+};
+
+enum {
+	ULP_TX_MEM_READ = 2,
+	ULP_TX_MEM_WRITE = 3,
+	ULP_TX_PKT = 4
+};
+
+enum {
+	ULP_TX_SC_NOOP = 0x80,
+	ULP_TX_SC_IMM  = 0x81,
+	ULP_TX_SC_DSGL = 0x82,
+	ULP_TX_SC_ISGL = 0x83
+};
+
+struct ulptx_sge_pair {
+	__be32 len[2];
+	__be64 addr[2];
+};
+
+struct ulptx_sgl {
+	__be32 cmd_nsge;
+#define ULPTX_CMD(x) ((x) << 24)
+#define ULPTX_NSGE(x) ((x) << 0)
+	__be32 len0;
+	__be64 addr0;
+	struct ulptx_sge_pair sge[0];
+};
+
+struct ulp_mem_io {
+	WR_HDR;
+	__be32 cmd;
+#define ULP_MEMIO_ORDER(x) ((x) << 23)
+	__be32 len16;             /* command length */
+	__be32 dlen;              /* data length in 32-byte units */
+#define ULP_MEMIO_DATA_LEN(x) ((x) << 0)
+	__be32 lock_addr;
+#define ULP_MEMIO_ADDR(x) ((x) << 0)
+#define ULP_MEMIO_LOCK(x) ((x) << 31)
+};
+
+#endif  /* __T4_MSG_H */
drivers/net/cxgb4/t4_regs.h new file +878 -0

878 lines not on this page · raw message

drivers/net/cxgb4/t4fw_api.h new file +1584 -0

1,458 lines not on this page · raw message

The rest of this diff is not shown. View the raw message.

Signature · 1 line
1.5.4

Part of the diff is not shown here. View the raw message.

#2 lore raw

PATCH 2/6 cxgb4: Add HW and FW support code

---
 drivers/net/cxgb4/t4_hw.c | 3131 +++++++++++++++++++++++++++++++++++++++++++++
 drivers/net/cxgb4/t4_hw.h |  100 ++
 2 files changed, 3231 insertions(+), 0 deletions(-)
 create mode 100644 drivers/net/cxgb4/t4_hw.c
 create mode 100644 drivers/net/cxgb4/t4_hw.h
drivers/net/cxgb4/t4_hw.c new file +3131 -0
new file mode 100644
index 0000000..a814a3a
@@ -0,0 +1,3131 @@
+/*
+ * This file is part of the Chelsio T4 Ethernet driver for Linux.
+ *
+ * Copyright (c) 2003-2010 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/init.h>
+#include <linux/delay.h>
+#include "cxgb4.h"
+#include "t4_regs.h"
+#include "t4fw_api.h"
+
+/**
+ *	t4_wait_op_done_val - wait until an operation is completed
+ *	@adapter: the adapter performing the operation
+ *	@reg: the register to check for completion
+ *	@mask: a single-bit field within @reg that indicates completion
+ *	@polarity: the value of the field when the operation is completed
+ *	@attempts: number of check iterations
+ *	@delay: delay in usecs between iterations
+ *	@valp: where to store the value of the register at completion time
+ *
+ *	Wait until an operation is completed by checking a bit in a register
+ *	up to @attempts times.  If @valp is not NULL the value of the register
+ *	at the time it indicated completion is stored there.  Returns 0 if the
+ *	operation completes and	-EAGAIN	otherwise.
+ */
+int t4_wait_op_done_val(struct adapter *adapter, int reg, u32 mask,
+			int polarity, int attempts, int delay, u32 *valp)
+{
+	while (1) {
+		u32 val = t4_read_reg(adapter, reg);
+
+		if (!!(val & mask) == polarity) {
+			if (valp)
+				*valp = val;
+			return 0;
+		}
+		if (--attempts == 0)
+			return -EAGAIN;
+		if (delay)
+			udelay(delay);
+	}
+}
+
+static inline int t4_wait_op_done(struct adapter *adapter, int reg, u32 mask,
+				  int polarity, int attempts, int delay)
+{
+	return t4_wait_op_done_val(adapter, reg, mask, polarity, attempts,
+				   delay, NULL);
+}
+
+/**
+ *	t4_set_reg_field - set a register field to a value
+ *	@adapter: the adapter to program
+ *	@addr: the register address
+ *	@mask: specifies the portion of the register to modify
+ *	@val: the new value for the register field
+ *
+ *	Sets a register field specified by the supplied mask to the
+ *	given value.
+ */
+void t4_set_reg_field(struct adapter *adapter, unsigned int addr, u32 mask,
+		      u32 val)
+{
+	u32 v = t4_read_reg(adapter, addr) & ~mask;
+
+	t4_write_reg(adapter, addr, v | val);
+	(void) t4_read_reg(adapter, addr);      /* flush */
+}
+
+/**
+ *	t4_read_indirect - read indirectly addressed registers
+ *	@adap: the adapter
+ *	@addr_reg: register holding the indirect address
+ *	@data_reg: register holding the value of the indirect register
+ *	@vals: where the read register values are stored
+ *	@nregs: how many indirect registers to read
+ *	@start_idx: index of first indirect register to read
+ *
+ *	Reads registers that are accessed indirectly through an address/data
+ *	register pair.
+ */
+void t4_read_indirect(struct adapter *adap, unsigned int addr_reg,
+		      unsigned int data_reg, u32 *vals, unsigned int nregs,
+		      unsigned int start_idx)
+{
+	while (nregs--) {
+		t4_write_reg(adap, addr_reg, start_idx);
+		*vals++ = t4_read_reg(adap, data_reg);
+		start_idx++;
+	}
+}
+
+/**
+ *	t4_write_indirect - write indirectly addressed registers
+ *	@adap: the adapter
+ *	@addr_reg: register holding the indirect addresses
+ *	@data_reg: register holding the value for the indirect registers
+ *	@vals: values to write
+ *	@nregs: how many indirect registers to write
+ *	@start_idx: address of first indirect register to write
+ *
+ *	Writes a sequential block of registers that are accessed indirectly
+ *	through an address/data register pair.
+ */
+void t4_write_indirect(struct adapter *adap, unsigned int addr_reg,
+		       unsigned int data_reg, const u32 *vals,
+		       unsigned int nregs, unsigned int start_idx)
+{
+	while (nregs--) {
+		t4_write_reg(adap, addr_reg, start_idx++);
+		t4_write_reg(adap, data_reg, *vals++);
+	}
+}
+
+/*
+ * Get the reply to a mailbox command and store it in @rpl in big-endian order.
+ */
+static void get_mbox_rpl(struct adapter *adap, __be64 *rpl, int nflit,
+			 u32 mbox_addr)
+{
+	for ( ; nflit; nflit--, mbox_addr += 8)
+		*rpl++ = cpu_to_be64(t4_read_reg64(adap, mbox_addr));
+}
+
+/*
+ * Handle a FW assertion reported in a mailbox.
+ */
+static void fw_asrt(struct adapter *adap, u32 mbox_addr)
+{
+	struct fw_debug_cmd asrt;
+
+	get_mbox_rpl(adap, (__be64 *)&asrt, sizeof(asrt) / 8, mbox_addr);
+	dev_alert(adap->pdev_dev,
+		  "FW assertion at %.16s:%u, val0 %#x, val1 %#x\n",
+		  asrt.u.assert.filename_0_7, ntohl(asrt.u.assert.line),
+		  ntohl(asrt.u.assert.x), ntohl(asrt.u.assert.y));
+}
+
+static void dump_mbox(struct adapter *adap, int mbox, u32 data_reg)
+{
+	dev_err(adap->pdev_dev,
+		"mbox %d: %llx %llx %llx %llx %llx %llx %llx %llx\n", mbox,
+		(unsigned long long)t4_read_reg64(adap, data_reg),
+		(unsigned long long)t4_read_reg64(adap, data_reg + 8),
+		(unsigned long long)t4_read_reg64(adap, data_reg + 16),
+		(unsigned long long)t4_read_reg64(adap, data_reg + 24),
+		(unsigned long long)t4_read_reg64(adap, data_reg + 32),
+		(unsigned long long)t4_read_reg64(adap, data_reg + 40),
+		(unsigned long long)t4_read_reg64(adap, data_reg + 48),
+		(unsigned long long)t4_read_reg64(adap, data_reg + 56));
+}
+
+/**
+ *	t4_wr_mbox_meat - send a command to FW through the given mailbox
+ *	@adap: the adapter
+ *	@mbox: index of the mailbox to use
+ *	@cmd: the command to write
+ *	@size: command length in bytes
+ *	@rpl: where to optionally store the reply
+ *	@sleep_ok: if true we may sleep while awaiting command completion
+ *
+ *	Sends the given command to FW through the selected mailbox and waits
+ *	for the FW to execute the command.  If @rpl is not %NULL it is used to
+ *	store the FW's reply to the command.  The command and its optional
+ *	reply are of the same length.  FW can take up to %FW_CMD_MAX_TIMEOUT ms
+ *	to respond.  @sleep_ok determines whether we may sleep while awaiting
+ *	the response.  If sleeping is allowed we use progressive backoff
+ *	otherwise we spin.
+ *
+ *	The return value is 0 on success or a negative errno on failure.  A
+ *	failure can happen either because we are not able to execute the
+ *	command or FW executes it but signals an error.  In the latter case
+ *	the return value is the error code indicated by FW (negated).
+ */
+int t4_wr_mbox_meat(struct adapter *adap, int mbox, const void *cmd, int size,
+		    void *rpl, bool sleep_ok)
+{
+	static int delay[] = {
+		1, 1, 3, 5, 10, 10, 20, 50, 100, 200
+	};
+
+	u32 v;
+	u64 res;
+	int i, ms, delay_idx;
+	const __be64 *p = cmd;
+	u32 data_reg = PF_REG(mbox, CIM_PF_MAILBOX_DATA);
+	u32 ctl_reg = PF_REG(mbox, CIM_PF_MAILBOX_CTRL);
+
+	if ((size & 15) || size > MBOX_LEN)
+		return -EINVAL;
+
+	v = MBOWNER_GET(t4_read_reg(adap, ctl_reg));
+	for (i = 0; v == MBOX_OWNER_NONE && i < 3; i++)
+		v = MBOWNER_GET(t4_read_reg(adap, ctl_reg));
+
+	if (v != MBOX_OWNER_DRV)
+		return v ? -EBUSY : -ETIMEDOUT;
+
+	for (i = 0; i < size; i += 8)
+		t4_write_reg64(adap, data_reg + i, be64_to_cpu(*p++));
+
+	t4_write_reg(adap, ctl_reg, MBMSGVALID | MBOWNER(MBOX_OWNER_FW));
+	t4_read_reg(adap, ctl_reg);          /* flush write */
+
+	delay_idx = 0;
+	ms = delay[0];
+
+	for (i = 0; i < FW_CMD_MAX_TIMEOUT; i += ms) {
+		if (sleep_ok) {
+			ms = delay[delay_idx];  /* last element may repeat */
+			if (delay_idx < ARRAY_SIZE(delay) - 1)
+				delay_idx++;
+			msleep(ms);
+		} else
+			mdelay(ms);
+
+		v = t4_read_reg(adap, ctl_reg);
+		if (MBOWNER_GET(v) == MBOX_OWNER_DRV) {
+			if (!(v & MBMSGVALID)) {
+				t4_write_reg(adap, ctl_reg, 0);
+				continue;
+			}
+
+			res = t4_read_reg64(adap, data_reg);
+			if (FW_CMD_OP_GET(res >> 32) == FW_DEBUG_CMD) {
+				fw_asrt(adap, data_reg);
+				res = FW_CMD_RETVAL(EIO);
+			} else if (rpl)
+				get_mbox_rpl(adap, rpl, size / 8, data_reg);
+
+			if (FW_CMD_RETVAL_GET((int)res))
+				dump_mbox(adap, mbox, data_reg);
+			t4_write_reg(adap, ctl_reg, 0);
+			return -FW_CMD_RETVAL_GET((int)res);
+		}
+	}
+
+	dump_mbox(adap, mbox, data_reg);
+	dev_err(adap->pdev_dev, "command %#x in mailbox %d timed out\n",
+		*(const u8 *)cmd, mbox);
+	return -ETIMEDOUT;
+}
+
+/**
+ *	t4_mc_read - read from MC through backdoor accesses
+ *	@adap: the adapter
+ *	@addr: address of first byte requested
+ *	@data: 64 bytes of data containing the requested address
+ *	@ecc: where to store the corresponding 64-bit ECC word
+ *
+ *	Read 64 bytes of data from MC starting at a 64-byte-aligned address
+ *	that covers the requested address @addr.  If @parity is not %NULL it
+ *	is assigned the 64-bit ECC word for the read data.
+ */
+int t4_mc_read(struct adapter *adap, u32 addr, __be32 *data, u64 *ecc)
+{
+	int i;
+
+	if (t4_read_reg(adap, MC_BIST_CMD) & START_BIST)
+		return -EBUSY;
+	t4_write_reg(adap, MC_BIST_CMD_ADDR, addr & ~0x3fU);
+	t4_write_reg(adap, MC_BIST_CMD_LEN, 64);
+	t4_write_reg(adap, MC_BIST_DATA_PATTERN, 0xc);
+	t4_write_reg(adap, MC_BIST_CMD, BIST_OPCODE(1) | START_BIST |
+		     BIST_CMD_GAP(1));
+	i = t4_wait_op_done(adap, MC_BIST_CMD, START_BIST, 0, 10, 1);
+	if (i)
+		return i;
+
+#define MC_DATA(i) MC_BIST_STATUS_REG(MC_BIST_STATUS_RDATA, i)
+
+	for (i = 15; i >= 0; i--)
+		*data++ = htonl(t4_read_reg(adap, MC_DATA(i)));
+	if (ecc)
+		*ecc = t4_read_reg64(adap, MC_DATA(16));
+#undef MC_DATA
+	return 0;
+}
+
+/**
+ *	t4_edc_read - read from EDC through backdoor accesses
+ *	@adap: the adapter
+ *	@idx: which EDC to access
+ *	@addr: address of first byte requested
+ *	@data: 64 bytes of data containing the requested address
+ *	@ecc: where to store the corresponding 64-bit ECC word
+ *
+ *	Read 64 bytes of data from EDC starting at a 64-byte-aligned address
+ *	that covers the requested address @addr.  If @parity is not %NULL it
+ *	is assigned the 64-bit ECC word for the read data.
+ */
+int t4_edc_read(struct adapter *adap, int idx, u32 addr, __be32 *data, u64 *ecc)
+{
+	int i;
+
+	idx *= EDC_STRIDE;
+	if (t4_read_reg(adap, EDC_BIST_CMD + idx) & START_BIST)
+		return -EBUSY;
+	t4_write_reg(adap, EDC_BIST_CMD_ADDR + idx, addr & ~0x3fU);
+	t4_write_reg(adap, EDC_BIST_CMD_LEN + idx, 64);
+	t4_write_reg(adap, EDC_BIST_DATA_PATTERN + idx, 0xc);
+	t4_write_reg(adap, EDC_BIST_CMD + idx,
+		     BIST_OPCODE(1) | BIST_CMD_GAP(1) | START_BIST);
+	i = t4_wait_op_done(adap, EDC_BIST_CMD + idx, START_BIST, 0, 10, 1);
+	if (i)
+		return i;
+
+#define EDC_DATA(i) (EDC_BIST_STATUS_REG(EDC_BIST_STATUS_RDATA, i) + idx)
+
+	for (i = 15; i >= 0; i--)
+		*data++ = htonl(t4_read_reg(adap, EDC_DATA(i)));
+	if (ecc)
+		*ecc = t4_read_reg64(adap, EDC_DATA(16));
+#undef EDC_DATA
+	return 0;
+}
+
+#define VPD_ENTRY(name, len) \
+	u8 name##_kword[2]; u8 name##_len; u8 name##_data[len]
+
+/*
+ * Partial EEPROM Vital Product Data structure.  Includes only the ID and
+ * VPD-R sections.
+ */
+struct t4_vpd {
+	u8  id_tag;
+	u8  id_len[2];
+	u8  id_data[ID_LEN];
+	u8  vpdr_tag;
+	u8  vpdr_len[2];
+	VPD_ENTRY(pn, 16);                     /* part number */
+	VPD_ENTRY(ec, EC_LEN);                 /* EC level */
+	VPD_ENTRY(sn, SERNUM_LEN);             /* serial number */
+	VPD_ENTRY(na, 12);                     /* MAC address base */
+	VPD_ENTRY(port_type, 8);               /* port types */
+	VPD_ENTRY(gpio, 14);                   /* GPIO usage */
+	VPD_ENTRY(cclk, 6);                    /* core clock */
+	VPD_ENTRY(port_addr, 8);               /* port MDIO addresses */
+	VPD_ENTRY(rv, 1);                      /* csum */
+	u32 pad;                  /* for multiple-of-4 sizing and alignment */
+};
+
+#define EEPROM_STAT_ADDR   0x7bfc
+#define VPD_BASE           0
+
+/**
+ *	t4_seeprom_wp - enable/disable EEPROM write protection
+ *	@adapter: the adapter
+ *	@enable: whether to enable or disable write protection
+ *
+ *	Enables or disables write protection on the serial EEPROM.
+ */
+int t4_seeprom_wp(struct adapter *adapter, bool enable)
+{
+	unsigned int v = enable ? 0xc : 0;
+	int ret = pci_write_vpd(adapter->pdev, EEPROM_STAT_ADDR, 4, &v);
+	return ret < 0 ? ret : 0;
+}
+
+/**
+ *	get_vpd_params - read VPD parameters from VPD EEPROM
+ *	@adapter: adapter to read
+ *	@p: where to store the parameters
+ *
+ *	Reads card parameters stored in VPD EEPROM.
+ */
+static int get_vpd_params(struct adapter *adapter, struct vpd_params *p)
+{
+	int ret;
+	struct t4_vpd vpd;
+	u8 *q = (u8 *)&vpd, csum;
+
+	ret = pci_read_vpd(adapter->pdev, VPD_BASE, sizeof(vpd), &vpd);
+	if (ret < 0)
+		return ret;
+
+	for (csum = 0; q <= vpd.rv_data; q++)
+		csum += *q;
+
+	if (csum) {
+		dev_err(adapter->pdev_dev,
+			"corrupted VPD EEPROM, actual csum %u\n", csum);
+		return -EINVAL;
+	}
+
+	p->cclk = simple_strtoul(vpd.cclk_data, NULL, 10);
+	memcpy(p->id, vpd.id_data, sizeof(vpd.id_data));
+	strim(p->id);
+	memcpy(p->ec, vpd.ec_data, sizeof(vpd.ec_data));
+	strim(p->ec);
+	memcpy(p->sn, vpd.sn_data, sizeof(vpd.sn_data));
+	strim(p->sn);
+	return 0;
+}
+
+/* serial flash and firmware constants */
+enum {
+	SF_ATTEMPTS = 10,             /* max retries for SF operations */
+
+	/* flash command opcodes */
+	SF_PROG_PAGE    = 2,          /* program page */
+	SF_WR_DISABLE   = 4,          /* disable writes */
+	SF_RD_STATUS    = 5,          /* read status register */
+	SF_WR_ENABLE    = 6,          /* enable writes */
+	SF_RD_DATA_FAST = 0xb,        /* read flash */
+	SF_ERASE_SECTOR = 0xd8,       /* erase sector */
+
+	FW_START_SEC = 8,             /* first flash sector for FW */
+	FW_END_SEC = 15,              /* last flash sector for FW */
+	FW_IMG_START = FW_START_SEC * SF_SEC_SIZE,
+	FW_MAX_SIZE = (FW_END_SEC - FW_START_SEC + 1) * SF_SEC_SIZE,
+};
+
+/**
+ *	sf1_read - read data from the serial flash
+ *	@adapter: the adapter
+ *	@byte_cnt: number of bytes to read
+ *	@cont: whether another operation will be chained
+ *	@lock: whether to lock SF for PL access only
+ *	@valp: where to store the read data
+ *
+ *	Reads up to 4 bytes of data from the serial flash.  The location of
+ *	the read needs to be specified prior to calling this by issuing the
+ *	appropriate commands to the serial flash.
+ */
+static int sf1_read(struct adapter *adapter, unsigned int byte_cnt, int cont,
+		    int lock, u32 *valp)
+{
+	int ret;
+
+	if (!byte_cnt || byte_cnt > 4)
+		return -EINVAL;
+	if (t4_read_reg(adapter, SF_OP) & BUSY)
+		return -EBUSY;
+	cont = cont ? SF_CONT : 0;
+	lock = lock ? SF_LOCK : 0;
+	t4_write_reg(adapter, SF_OP, lock | cont | BYTECNT(byte_cnt - 1));
+	ret = t4_wait_op_done(adapter, SF_OP, BUSY, 0, SF_ATTEMPTS, 5);
+	if (!ret)
+		*valp = t4_read_reg(adapter, SF_DATA);
+	return ret;
+}
+
+/**
+ *	sf1_write - write data to the serial flash
+ *	@adapter: the adapter
+ *	@byte_cnt: number of bytes to write
+ *	@cont: whether another operation will be chained
+ *	@lock: whether to lock SF for PL access only
+ *	@val: value to write
+ *
+ *	Writes up to 4 bytes of data to the serial flash.  The location of
+ *	the write needs to be specified prior to calling this by issuing the
+ *	appropriate commands to the serial flash.
+ */
+static int sf1_write(struct adapter *adapter, unsigned int byte_cnt, int cont,
+		     int lock, u32 val)
+{
+	if (!byte_cnt || byte_cnt > 4)
+		return -EINVAL;
+	if (t4_read_reg(adapter, SF_OP) & BUSY)
+		return -EBUSY;
+	cont = cont ? SF_CONT : 0;
+	lock = lock ? SF_LOCK : 0;
+	t4_write_reg(adapter, SF_DATA, val);
+	t4_write_reg(adapter, SF_OP, lock |
+		     cont | BYTECNT(byte_cnt - 1) | OP_WR);
+	return t4_wait_op_done(adapter, SF_OP, BUSY, 0, SF_ATTEMPTS, 5);
+}
+
+/**
+ *	flash_wait_op - wait for a flash operation to complete
+ *	@adapter: the adapter
+ *	@attempts: max number of polls of the status register
+ *	@delay: delay between polls in ms
+ *
+ *	Wait for a flash operation to complete by polling the status register.
+ */
+static int flash_wait_op(struct adapter *adapter, int attempts, int delay)
+{
+	int ret;
+	u32 status;
+
+	while (1) {
+		if ((ret = sf1_write(adapter, 1, 1, 1, SF_RD_STATUS)) != 0 ||
+		    (ret = sf1_read(adapter, 1, 0, 1, &status)) != 0)
+			return ret;
+		if (!(status & 1))
+			return 0;
+		if (--attempts == 0)
+			return -EAGAIN;
+		if (delay)
+			msleep(delay);
+	}
+}
+
+/**
+ *	t4_read_flash - read words from serial flash
+ *	@adapter: the adapter
+ *	@addr: the start address for the read
+ *	@nwords: how many 32-bit words to read
+ *	@data: where to store the read data
+ *	@byte_oriented: whether to store data as bytes or as words
+ *
+ *	Read the specified number of 32-bit words from the serial flash.
+ *	If @byte_oriented is set the read data is stored as a byte array
+ *	(i.e., big-endian), otherwise as 32-bit words in the platform's
+ *	natural endianess.
+ */
+int t4_read_flash(struct adapter *adapter, unsigned int addr,
+		  unsigned int nwords, u32 *data, int byte_oriented)
+{
+	int ret;
+
+	if (addr + nwords * sizeof(u32) > SF_SIZE || (addr & 3))
+		return -EINVAL;
+
+	addr = swab32(addr) | SF_RD_DATA_FAST;
+
+	if ((ret = sf1_write(adapter, 4, 1, 0, addr)) != 0 ||
+	    (ret = sf1_read(adapter, 1, 1, 0, data)) != 0)
+		return ret;
+
+	for ( ; nwords; nwords--, data++) {
+		ret = sf1_read(adapter, 4, nwords > 1, nwords == 1, data);
+		if (nwords == 1)
+			t4_write_reg(adapter, SF_OP, 0);    /* unlock SF */
+		if (ret)
+			return ret;
+		if (byte_oriented)
+			*data = htonl(*data);
+	}
+	return 0;
+}
+
+/**
+ *	t4_write_flash - write up to a page of data to the serial flash
+ *	@adapter: the adapter
+ *	@addr: the start address to write
+ *	@n: length of data to write in bytes
+ *	@data: the data to write
+ *
+ *	Writes up to a page of data (256 bytes) to the serial flash starting
+ *	at the given address.  All the data must be written to the same page.
+ */
+static int t4_write_flash(struct adapter *adapter, unsigned int addr,
+			  unsigned int n, const u8 *data)
+{
+	int ret;
+	u32 buf[64];
+	unsigned int i, c, left, val, offset = addr & 0xff;
+
+	if (addr >= SF_SIZE || offset + n > SF_PAGE_SIZE)
+		return -EINVAL;
+
+	val = swab32(addr) | SF_PROG_PAGE;
+
+	if ((ret = sf1_write(adapter, 1, 0, 1, SF_WR_ENABLE)) != 0 ||
+	    (ret = sf1_write(adapter, 4, 1, 1, val)) != 0)
+		goto unlock;
+
+	for (left = n; left; left -= c) {
+		c = min(left, 4U);
+		for (val = 0, i = 0; i < c; ++i)
+			val = (val << 8) + *data++;
+
+		ret = sf1_write(adapter, c, c != left, 1, val);
+		if (ret)
+			goto unlock;
+	}
+	ret = flash_wait_op(adapter, 5, 1);
+	if (ret)
+		goto unlock;
+
+	t4_write_reg(adapter, SF_OP, 0);    /* unlock SF */
+
+	/* Read the page to verify the write succeeded */
+	ret = t4_read_flash(adapter, addr & ~0xff, ARRAY_SIZE(buf), buf, 1);
+	if (ret)
+		return ret;
+
+	if (memcmp(data - n, (u8 *)buf + offset, n)) {
+		dev_err(adapter->pdev_dev,
+			"failed to correctly write the flash page at %#x\n",
+			addr);
+		return -EIO;
+	}
+	return 0;
+
+unlock:
+	t4_write_reg(adapter, SF_OP, 0);    /* unlock SF */
+	return ret;
+}
+
+/**
+ *	get_fw_version - read the firmware version
+ *	@adapter: the adapter
+ *	@vers: where to place the version
+ *
+ *	Reads the FW version from flash.
+ */
+static int get_fw_version(struct adapter *adapter, u32 *vers)
+{
+	return t4_read_flash(adapter,
+			     FW_IMG_START + offsetof(struct fw_hdr, fw_ver), 1,
+			     vers, 0);
+}
+
+/**
+ *	get_tp_version - read the TP microcode version
+ *	@adapter: the adapter
+ *	@vers: where to place the version
+ *
+ *	Reads the TP microcode version from flash.
+ */
+static int get_tp_version(struct adapter *adapter, u32 *vers)
+{
+	return t4_read_flash(adapter, FW_IMG_START + offsetof(struct fw_hdr,
+							      tp_microcode_ver),
+			     1, vers, 0);
+}
+
+/**
+ *	t4_check_fw_version - check if the FW is compatible with this driver
+ *	@adapter: the adapter
+ *
+ *	Checks if an adapter's FW is compatible with the driver.  Returns 0
+ *	if there's exact match, a negative error if the version could not be
+ *	read or there's a major version mismatch, and a positive value if the
+ *	expected major version is found but there's a minor version mismatch.
+ */
+int t4_check_fw_version(struct adapter *adapter)
+{
+	u32 api_vers[2];
+	int ret, major, minor, micro;
+
+	ret = get_fw_version(adapter, &adapter->params.fw_vers);
+	if (!ret)
+		ret = get_tp_version(adapter, &adapter->params.tp_vers);
+	if (!ret)
+		ret = t4_read_flash(adapter,
+			FW_IMG_START + offsetof(struct fw_hdr, intfver_nic),
+			2, api_vers, 1);
+	if (ret)
+		return ret;
+
+	major = FW_HDR_FW_VER_MAJOR_GET(adapter->params.fw_vers);
+	minor = FW_HDR_FW_VER_MINOR_GET(adapter->params.fw_vers);
+	micro = FW_HDR_FW_VER_MICRO_GET(adapter->params.fw_vers);
+	memcpy(adapter->params.api_vers, api_vers,
+	       sizeof(adapter->params.api_vers));
+
+	if (major != FW_VERSION_MAJOR) {            /* major mismatch - fail */
+		dev_err(adapter->pdev_dev,
+			"card FW has major version %u, driver wants %u\n",
+			major, FW_VERSION_MAJOR);
+		return -EINVAL;
+	}
+
+	if (minor == FW_VERSION_MINOR && micro == FW_VERSION_MICRO)
+		return 0;                                   /* perfect match */
+
+	/* Minor/micro version mismatch.  Report it but often it's OK. */
+	return 1;
+}
+
+/**
+ *	t4_flash_erase_sectors - erase a range of flash sectors
+ *	@adapter: the adapter
+ *	@start: the first sector to erase
+ *	@end: the last sector to erase
+ *
+ *	Erases the sectors in the given inclusive range.
+ */
+static int t4_flash_erase_sectors(struct adapter *adapter, int start, int end)
+{
+	int ret = 0;
+
+	while (start <= end) {
+		if ((ret = sf1_write(adapter, 1, 0, 1, SF_WR_ENABLE)) != 0 ||
+		    (ret = sf1_write(adapter, 4, 0, 1,
+				     SF_ERASE_SECTOR | (start << 8))) != 0 ||
+		    (ret = flash_wait_op(adapter, 5, 500)) != 0) {
+			dev_err(adapter->pdev_dev,
+				"erase of flash sector %d failed, error %d\n",
+				start, ret);
+			break;
+		}
+		start++;
+	}
+	t4_write_reg(adapter, SF_OP, 0);    /* unlock SF */
+	return ret;
+}
+
+/**
+ *	t4_load_fw - download firmware
+ *	@adap: the adapter
+ *	@fw_data: the firmware image to write
+ *	@size: image size
+ *
+ *	Write the supplied firmware image to the card's serial flash.
+ */
+int t4_load_fw(struct adapter *adap, const u8 *fw_data, unsigned int size)
+{
+	u32 csum;
+	int ret, addr;
+	unsigned int i;
+	u8 first_page[SF_PAGE_SIZE];
+	const u32 *p = (const u32 *)fw_data;
+	const struct fw_hdr *hdr = (const struct fw_hdr *)fw_data;
+
+	if (!size) {
+		dev_err(adap->pdev_dev, "FW image has no data\n");
+		return -EINVAL;
+	}
+	if (size & 511) {
+		dev_err(adap->pdev_dev,
+			"FW image size not multiple of 512 bytes\n");
+		return -EINVAL;
+	}
+	if (ntohs(hdr->len512) * 512 != size) {
+		dev_err(adap->pdev_dev,
+			"FW image size differs from size in FW header\n");
+		return -EINVAL;
+	}
+	if (size > FW_MAX_SIZE) {
+		dev_err(adap->pdev_dev, "FW image too large, max is %u bytes\n",
+			FW_MAX_SIZE);
+		return -EFBIG;
+	}
+
+	for (csum = 0, i = 0; i < size / sizeof(csum); i++)
+		csum += ntohl(p[i]);
+
+	if (csum != 0xffffffff) {
+		dev_err(adap->pdev_dev,
+			"corrupted firmware image, checksum %#x\n", csum);
+		return -EINVAL;
+	}
+
+	i = DIV_ROUND_UP(size, SF_SEC_SIZE);        /* # of sectors spanned */
+	ret = t4_flash_erase_sectors(adap, FW_START_SEC, FW_START_SEC + i - 1);
+	if (ret)
+		goto out;
+
+	/*
+	 * We write the correct version at the end so the driver can see a bad
+	 * version if the FW write fails.  Start by writing a copy of the
+	 * first page with a bad version.
+	 */
+	memcpy(first_page, fw_data, SF_PAGE_SIZE);
+	((struct fw_hdr *)first_page)->fw_ver = htonl(0xffffffff);
+	ret = t4_write_flash(adap, FW_IMG_START, SF_PAGE_SIZE, first_page);
+	if (ret)
+		goto out;
+
+	addr = FW_IMG_START;
+	for (size -= SF_PAGE_SIZE; size; size -= SF_PAGE_SIZE) {
+		addr += SF_PAGE_SIZE;
+		fw_data += SF_PAGE_SIZE;
+		ret = t4_write_flash(adap, addr, SF_PAGE_SIZE, fw_data);
+		if (ret)
+			goto out;
+	}
+
+	ret = t4_write_flash(adap,
+			     FW_IMG_START + offsetof(struct fw_hdr, fw_ver),
+			     sizeof(hdr->fw_ver), (const u8 *)&hdr->fw_ver);
+out:
+	if (ret)
+		dev_err(adap->pdev_dev, "firmware download failed, error %d\n",
+			ret);
+	return ret;
+}
+
+#define ADVERT_MASK (FW_PORT_CAP_SPEED_100M | FW_PORT_CAP_SPEED_1G |\
+		     FW_PORT_CAP_SPEED_10G | FW_PORT_CAP_ANEG)
+
+/**
+ *	t4_link_start - apply link configuration to MAC/PHY
+ *	@phy: the PHY to setup
+ *	@mac: the MAC to setup
+ *	@lc: the requested link configuration
+ *
+ *	Set up a port's MAC and PHY according to a desired link configuration.
+ *	- If the PHY can auto-negotiate first decide what to advertise, then
+ *	  enable/disable auto-negotiation as desired, and reset.
+ *	- If the PHY does not auto-negotiate just reset it.
+ *	- If auto-negotiation is off set the MAC to the proper speed/duplex/FC,
+ *	  otherwise do it later based on the outcome of auto-negotiation.
+ */
+int t4_link_start(struct adapter *adap, unsigned int mbox, unsigned int port,
+		  struct link_config *lc)
+{
+	struct fw_port_cmd c;
+	unsigned int fc = 0, mdi = FW_PORT_MDI(FW_PORT_MDI_AUTO);
+
+	lc->link_ok = 0;
+	if (lc->requested_fc & PAUSE_RX)
+		fc |= FW_PORT_CAP_FC_RX;
+	if (lc->requested_fc & PAUSE_TX)
+		fc |= FW_PORT_CAP_FC_TX;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_portid = htonl(FW_CMD_OP(FW_PORT_CMD) | FW_CMD_REQUEST |
+			       FW_CMD_EXEC | FW_PORT_CMD_PORTID(port));
+	c.action_to_len16 = htonl(FW_PORT_CMD_ACTION(FW_PORT_ACTION_L1_CFG) |
+				  FW_LEN16(c));
+
+	if (!(lc->supported & FW_PORT_CAP_ANEG)) {
+		c.u.l1cfg.rcap = htonl((lc->supported & ADVERT_MASK) | fc);
+		lc->fc = lc->requested_fc & (PAUSE_RX | PAUSE_TX);
+	} else if (lc->autoneg == AUTONEG_DISABLE) {
+		c.u.l1cfg.rcap = htonl(lc->requested_speed | fc | mdi);
+		lc->fc = lc->requested_fc & (PAUSE_RX | PAUSE_TX);
+	} else
+		c.u.l1cfg.rcap = htonl(lc->advertising | fc | mdi);
+
+	return t4_wr_mbox(adap, mbox, &c, sizeof(c), NULL);
+}
+
+/**
+ *	t4_restart_aneg - restart autonegotiation
+ *	@adap: the adapter
+ *	@mbox: mbox to use for the FW command
+ *	@port: the port id
+ *
+ *	Restarts autonegotiation for the selected port.
+ */
+int t4_restart_aneg(struct adapter *adap, unsigned int mbox, unsigned int port)
+{
+	struct fw_port_cmd c;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_portid = htonl(FW_CMD_OP(FW_PORT_CMD) | FW_CMD_REQUEST |
+			       FW_CMD_EXEC | FW_PORT_CMD_PORTID(port));
+	c.action_to_len16 = htonl(FW_PORT_CMD_ACTION(FW_PORT_ACTION_L1_CFG) |
+				  FW_LEN16(c));
+	c.u.l1cfg.rcap = htonl(FW_PORT_CAP_ANEG);
+	return t4_wr_mbox(adap, mbox, &c, sizeof(c), NULL);
+}
+
+/**
+ *	t4_set_vlan_accel - configure HW VLAN extraction
+ *	@adap: the adapter
+ *	@ports: bitmap of adapter ports to operate on
+ *	@on: enable (1) or disable (0) HW VLAN extraction
+ *
+ *	Enables or disables HW extraction of VLAN tags for the ports specified
+ *	by @ports.  @ports is a bitmap with the ith bit designating the port
+ *	associated with the ith adapter channel.
+ */
+void t4_set_vlan_accel(struct adapter *adap, unsigned int ports, int on)
+{
+	ports <<= VLANEXTENABLE_SHIFT;
+	t4_set_reg_field(adap, TP_OUT_CONFIG, ports, on ? ports : 0);
+}
+
+struct intr_info {
+	unsigned int mask;       /* bits to check in interrupt status */
+	const char *msg;         /* message to print or NULL */
+	short stat_idx;          /* stat counter to increment or -1 */
+	unsigned short fatal;    /* whether the condition reported is fatal */
+};
+
+/**
+ *	t4_handle_intr_status - table driven interrupt handler
+ *	@adapter: the adapter that generated the interrupt
+ *	@reg: the interrupt status register to process
+ *	@acts: table of interrupt actions
+ *
+ *	A table driven interrupt handler that applies a set of masks to an
+ *	interrupt status word and performs the corresponding actions if the
+ *	interrupts described by the mask have occured.  The actions include
+ *	optionally emitting a warning or alert message.  The table is terminated
+ *	by an entry specifying mask 0.  Returns the number of fatal interrupt
+ *	conditions.
+ */
+static int t4_handle_intr_status(struct adapter *adapter, unsigned int reg,
+				 const struct intr_info *acts)
+{
+	int fatal = 0;
+	unsigned int mask = 0;
+	unsigned int status = t4_read_reg(adapter, reg);
+
+	for ( ; acts->mask; ++acts) {
+		if (!(status & acts->mask))
+			continue;
+		if (acts->fatal) {
+			fatal++;
+			dev_alert(adapter->pdev_dev, "%s (0x%x)\n", acts->msg,
+				  status & acts->mask);
+		} else if (acts->msg && printk_ratelimit())
+			dev_warn(adapter->pdev_dev, "%s (0x%x)\n", acts->msg,
+				 status & acts->mask);
+		mask |= acts->mask;
+	}
+	status &= mask;
+	if (status)                           /* clear processed interrupts */
+		t4_write_reg(adapter, reg, status);
+	return fatal;
+}
+
+/*
+ * Interrupt handler for the PCIE module.
+ */
+static void pcie_intr_handler(struct adapter *adapter)
+{
+	static struct intr_info sysbus_intr_info[] = {
+		{ RNPP, "RXNP array parity error", -1, 1 },
+		{ RPCP, "RXPC array parity error", -1, 1 },
+		{ RCIP, "RXCIF array parity error", -1, 1 },
+		{ RCCP, "Rx completions control array parity error", -1, 1 },
+		{ RFTP, "RXFT array parity error", -1, 1 },
+		{ 0 }
+	};
+	static struct intr_info pcie_port_intr_info[] = {
+		{ TPCP, "TXPC array parity error", -1, 1 },
+		{ TNPP, "TXNP array parity error", -1, 1 },
+		{ TFTP, "TXFT array parity error", -1, 1 },
+		{ TCAP, "TXCA array parity error", -1, 1 },
+		{ TCIP, "TXCIF array parity error", -1, 1 },
+		{ RCAP, "RXCA array parity error", -1, 1 },
+		{ OTDD, "outbound request TLP discarded", -1, 1 },
+		{ RDPE, "Rx data parity error", -1, 1 },
+		{ TDUE, "Tx uncorrectable data error", -1, 1 },
+		{ 0 }
+	};
+	static struct intr_info pcie_intr_info[] = {
+		{ MSIADDRLPERR, "MSI AddrL parity error", -1, 1 },
+		{ MSIADDRHPERR, "MSI AddrH parity error", -1, 1 },
+		{ MSIDATAPERR, "MSI data parity error", -1, 1 },
+		{ MSIXADDRLPERR, "MSI-X AddrL parity error", -1, 1 },
+		{ MSIXADDRHPERR, "MSI-X AddrH parity error", -1, 1 },
+		{ MSIXDATAPERR, "MSI-X data parity error", -1, 1 },
+		{ MSIXDIPERR, "MSI-X DI parity error", -1, 1 },
+		{ PIOCPLPERR, "PCI PIO completion FIFO parity error", -1, 1 },
+		{ PIOREQPERR, "PCI PIO request FIFO parity error", -1, 1 },
+		{ TARTAGPERR, "PCI PCI target tag FIFO parity error", -1, 1 },
+		{ CCNTPERR, "PCI CMD channel count parity error", -1, 1 },
+		{ CREQPERR, "PCI CMD channel request parity error", -1, 1 },
+		{ CRSPPERR, "PCI CMD channel response parity error", -1, 1 },
+		{ DCNTPERR, "PCI DMA channel count parity error", -1, 1 },
+		{ DREQPERR, "PCI DMA channel request parity error", -1, 1 },
+		{ DRSPPERR, "PCI DMA channel response parity error", -1, 1 },
+		{ HCNTPERR, "PCI HMA channel count parity error", -1, 1 },
+		{ HREQPERR, "PCI HMA channel request parity error", -1, 1 },
+		{ HRSPPERR, "PCI HMA channel response parity error", -1, 1 },
+		{ CFGSNPPERR, "PCI config snoop FIFO parity error", -1, 1 },
+		{ FIDPERR, "PCI FID parity error", -1, 1 },
+		{ INTXCLRPERR, "PCI INTx clear parity error", -1, 1 },
+		{ MATAGPERR, "PCI MA tag parity error", -1, 1 },
+		{ PIOTAGPERR, "PCI PIO tag parity error", -1, 1 },
+		{ RXCPLPERR, "PCI Rx completion parity error", -1, 1 },
+		{ RXWRPERR, "PCI Rx write parity error", -1, 1 },
+		{ RPLPERR, "PCI replay buffer parity error", -1, 1 },
+		{ PCIESINT, "PCI core secondary fault", -1, 1 },
+		{ PCIEPINT, "PCI core primary fault", -1, 1 },
+		{ UNXSPLCPLERR, "PCI unexpected split completion error", -1, 0 },
+		{ 0 }
+	};
+
+	int fat;
+
+	fat = t4_handle_intr_status(adapter,
+				    PCIE_CORE_UTL_SYSTEM_BUS_AGENT_STATUS,
+				    sysbus_intr_info) +
+	      t4_handle_intr_status(adapter,
+				    PCIE_CORE_UTL_PCI_EXPRESS_PORT_STATUS,
+				    pcie_port_intr_info) +
+	      t4_handle_intr_status(adapter, PCIE_INT_CAUSE, pcie_intr_info);
+	if (fat)
+		t4_fatal_err(adapter);
+}
+
+/*
+ * TP interrupt handler.
+ */
+static void tp_intr_handler(struct adapter *adapter)
+{
+	static struct intr_info tp_intr_info[] = {
+		{ 0x3fffffff, "TP parity error", -1, 1 },
+		{ FLMTXFLSTEMPTY, "TP out of Tx pages", -1, 1 },
+		{ 0 }
+	};
+
+	if (t4_handle_intr_status(adapter, TP_INT_CAUSE, tp_intr_info))
+		t4_fatal_err(adapter);
+}
+
+/*
+ * SGE interrupt handler.
+ */
+static void sge_intr_handler(struct adapter *adapter)
+{
+	u64 v;
+
+	static struct intr_info sge_intr_info[] = {
+		{ ERR_CPL_EXCEED_IQE_SIZE,
+		  "SGE received CPL exceeding IQE size", -1, 1 },
+		{ ERR_INVALID_CIDX_INC,
+		  "SGE GTS CIDX increment too large", -1, 0 },
+		{ ERR_CPL_OPCODE_0, "SGE received 0-length CPL", -1, 0 },
+		{ ERR_DROPPED_DB, "SGE doorbell dropped", -1, 0 },
+		{ ERR_DATA_CPL_ON_HIGH_QID1 | ERR_DATA_CPL_ON_HIGH_QID0,
+		  "SGE IQID > 1023 received CPL for FL", -1, 0 },
+		{ ERR_BAD_DB_PIDX3, "SGE DBP 3 pidx increment too large", -1,
+		  0 },
+		{ ERR_BAD_DB_PIDX2, "SGE DBP 2 pidx increment too large", -1,
+		  0 },
+		{ ERR_BAD_DB_PIDX1, "SGE DBP 1 pidx increment too large", -1,
+		  0 },
+		{ ERR_BAD_DB_PIDX0, "SGE DBP 0 pidx increment too large", -1,
+		  0 },
+		{ ERR_ING_CTXT_PRIO,
+		  "SGE too many priority ingress contexts", -1, 0 },
+		{ ERR_EGR_CTXT_PRIO,
+		  "SGE too many priority egress contexts", -1, 0 },
+		{ INGRESS_SIZE_ERR, "SGE illegal ingress QID", -1, 0 },
+		{ EGRESS_SIZE_ERR, "SGE illegal egress QID", -1, 0 },
+		{ 0 }
+	};
+
+	v = (u64)t4_read_reg(adapter, SGE_INT_CAUSE1) |
+	    ((u64)t4_read_reg(adapter, SGE_INT_CAUSE2) << 32);
+	if (v) {
+		dev_alert(adapter->pdev_dev, "SGE parity error (%#llx)\n",
+			 (unsigned long long)v);
+		t4_write_reg(adapter, SGE_INT_CAUSE1, v);
+		t4_write_reg(adapter, SGE_INT_CAUSE2, v >> 32);
+	}
+
+	if (t4_handle_intr_status(adapter, SGE_INT_CAUSE3, sge_intr_info) ||
+	    v != 0)
+		t4_fatal_err(adapter);
+}
+
+/*
+ * CIM interrupt handler.
+ */
+static void cim_intr_handler(struct adapter *adapter)
+{
+	static struct intr_info cim_intr_info[] = {
+		{ PREFDROPINT, "CIM control register prefetch drop", -1, 1 },
+		{ OBQPARERR, "CIM OBQ parity error", -1, 1 },
+		{ IBQPARERR, "CIM IBQ parity error", -1, 1 },
+		{ MBUPPARERR, "CIM mailbox uP parity error", -1, 1 },
+		{ MBHOSTPARERR, "CIM mailbox host parity error", -1, 1 },
+		{ TIEQINPARERRINT, "CIM TIEQ outgoing parity error", -1, 1 },
+		{ TIEQOUTPARERRINT, "CIM TIEQ incoming parity error", -1, 1 },
+		{ 0 }
+	};
+	static struct intr_info cim_upintr_info[] = {
+		{ RSVDSPACEINT, "CIM reserved space access", -1, 1 },
+		{ ILLTRANSINT, "CIM illegal transaction", -1, 1 },
+		{ ILLWRINT, "CIM illegal write", -1, 1 },
+		{ ILLRDINT, "CIM illegal read", -1, 1 },
+		{ ILLRDBEINT, "CIM illegal read BE", -1, 1 },
+		{ ILLWRBEINT, "CIM illegal write BE", -1, 1 },
+		{ SGLRDBOOTINT, "CIM single read from boot space", -1, 1 },
+		{ SGLWRBOOTINT, "CIM single write to boot space", -1, 1 },
+		{ BLKWRBOOTINT, "CIM block write to boot space", -1, 1 },
+		{ SGLRDFLASHINT, "CIM single read from flash space", -1, 1 },
+		{ SGLWRFLASHINT, "CIM single write to flash space", -1, 1 },
+		{ BLKWRFLASHINT, "CIM block write to flash space", -1, 1 },
+		{ SGLRDEEPROMINT, "CIM single EEPROM read", -1, 1 },
+		{ SGLWREEPROMINT, "CIM single EEPROM write", -1, 1 },
+		{ BLKRDEEPROMINT, "CIM block EEPROM read", -1, 1 },
+		{ BLKWREEPROMINT, "CIM block EEPROM write", -1, 1 },
+		{ SGLRDCTLINT , "CIM single read from CTL space", -1, 1 },
+		{ SGLWRCTLINT , "CIM single write to CTL space", -1, 1 },
+		{ BLKRDCTLINT , "CIM block read from CTL space", -1, 1 },
+		{ BLKWRCTLINT , "CIM block write to CTL space", -1, 1 },
+		{ SGLRDPLINT , "CIM single read from PL space", -1, 1 },
+		{ SGLWRPLINT , "CIM single write to PL space", -1, 1 },
+		{ BLKRDPLINT , "CIM block read from PL space", -1, 1 },
+		{ BLKWRPLINT , "CIM block write to PL space", -1, 1 },
+		{ REQOVRLOOKUPINT , "CIM request FIFO overwrite", -1, 1 },
+		{ RSPOVRLOOKUPINT , "CIM response FIFO overwrite", -1, 1 },
+		{ TIMEOUTINT , "CIM PIF timeout", -1, 1 },
+		{ TIMEOUTMAINT , "CIM PIF MA timeout", -1, 1 },
+		{ 0 }
+	};
+
+	int fat;
+
+	fat = t4_handle_intr_status(adapter, CIM_HOST_INT_CAUSE,
+				    cim_intr_info) +
+	      t4_handle_intr_status(adapter, CIM_HOST_UPACC_INT_CAUSE,
+				    cim_upintr_info);
+	if (fat)
+		t4_fatal_err(adapter);
+}
+
+/*
+ * ULP RX interrupt handler.
+ */
+static void ulprx_intr_handler(struct adapter *adapter)
+{
+	static struct intr_info ulprx_intr_info[] = {
+		{ 0x7fffff, "ULPRX parity error", -1, 1 },
+		{ 0 }
+	};
+
+	if (t4_handle_intr_status(adapter, ULP_RX_INT_CAUSE, ulprx_intr_info))
+		t4_fatal_err(adapter);
+}
+
+/*
+ * ULP TX interrupt handler.
+ */
+static void ulptx_intr_handler(struct adapter *adapter)
+{
+	static struct intr_info ulptx_intr_info[] = {
+		{ PBL_BOUND_ERR_CH3, "ULPTX channel 3 PBL out of bounds", -1,
+		  0 },
+		{ PBL_BOUND_ERR_CH2, "ULPTX channel 2 PBL out of bounds", -1,
+		  0 },
+		{ PBL_BOUND_ERR_CH1, "ULPTX channel 1 PBL out of bounds", -1,
+		  0 },
+		{ PBL_BOUND_ERR_CH0, "ULPTX channel 0 PBL out of bounds", -1,
+		  0 },
+		{ 0xfffffff, "ULPTX parity error", -1, 1 },
+		{ 0 }
+	};
+
+	if (t4_handle_intr_status(adapter, ULP_TX_INT_CAUSE, ulptx_intr_info))
+		t4_fatal_err(adapter);
+}
+
+/*
+ * PM TX interrupt handler.
+ */
+static void pmtx_intr_handler(struct adapter *adapter)
+{
+	static struct intr_info pmtx_intr_info[] = {
+		{ PCMD_LEN_OVFL0, "PMTX channel 0 pcmd too large", -1, 1 },
+		{ PCMD_LEN_OVFL1, "PMTX channel 1 pcmd too large", -1, 1 },
+		{ PCMD_LEN_OVFL2, "PMTX channel 2 pcmd too large", -1, 1 },
+		{ ZERO_C_CMD_ERROR, "PMTX 0-length pcmd", -1, 1 },
+		{ PMTX_FRAMING_ERROR, "PMTX framing error", -1, 1 },
+		{ OESPI_PAR_ERROR, "PMTX oespi parity error", -1, 1 },
+		{ DB_OPTIONS_PAR_ERROR, "PMTX db_options parity error", -1, 1 },
+		{ ICSPI_PAR_ERROR, "PMTX icspi parity error", -1, 1 },
+		{ C_PCMD_PAR_ERROR, "PMTX c_pcmd parity error", -1, 1},
+		{ 0 }
+	};
+
+	if (t4_handle_intr_status(adapter, PM_TX_INT_CAUSE, pmtx_intr_info))
+		t4_fatal_err(adapter);
+}
+
+/*
+ * PM RX interrupt handler.
+ */
+static void pmrx_intr_handler(struct adapter *adapter)
+{
+	static struct intr_info pmrx_intr_info[] = {
+		{ ZERO_E_CMD_ERROR, "PMRX 0-length pcmd", -1, 1 },
+		{ PMRX_FRAMING_ERROR, "PMRX framing error", -1, 1 },
+		{ OCSPI_PAR_ERROR, "PMRX ocspi parity error", -1, 1 },
+		{ DB_OPTIONS_PAR_ERROR, "PMRX db_options parity error", -1, 1 },
+		{ IESPI_PAR_ERROR, "PMRX iespi parity error", -1, 1 },
+		{ E_PCMD_PAR_ERROR, "PMRX e_pcmd parity error", -1, 1},
+		{ 0 }
+	};
+
+	if (t4_handle_intr_status(adapter, PM_RX_INT_CAUSE, pmrx_intr_info))
+		t4_fatal_err(adapter);
+}
+
+/*
+ * CPL switch interrupt handler.
+ */
+static void cplsw_intr_handler(struct adapter *adapter)
+{
+	static struct intr_info cplsw_intr_info[] = {
+		{ CIM_OP_MAP_PERR, "CPLSW CIM op_map parity error", -1, 1 },
+		{ CIM_OVFL_ERROR, "CPLSW CIM overflow", -1, 1 },
+		{ TP_FRAMING_ERROR, "CPLSW TP framing error", -1, 1 },
+		{ SGE_FRAMING_ERROR, "CPLSW SGE framing error", -1, 1 },
+		{ CIM_FRAMING_ERROR, "CPLSW CIM framing error", -1, 1 },
+		{ ZERO_SWITCH_ERROR, "CPLSW no-switch error", -1, 1 },
+		{ 0 }
+	};
+
+	if (t4_handle_intr_status(adapter, CPL_INTR_CAUSE, cplsw_intr_info))
+		t4_fatal_err(adapter);
+}
+
+/*
+ * LE interrupt handler.
+ */
+static void le_intr_handler(struct adapter *adap)
+{
+	static struct intr_info le_intr_info[] = {
+		{ LIPMISS, "LE LIP miss", -1, 0 },
+		{ LIP0, "LE 0 LIP error", -1, 0 },
+		{ PARITYERR, "LE parity error", -1, 1 },
+		{ UNKNOWNCMD, "LE unknown command", -1, 1 },
+		{ REQQPARERR, "LE request queue parity error", -1, 1 },
+		{ 0 }
+	};
+
+	if (t4_handle_intr_status(adap, LE_DB_INT_CAUSE, le_intr_info))
+		t4_fatal_err(adap);
+}
+
+/*
+ * MPS interrupt handler.
+ */
+static void mps_intr_handler(struct adapter *adapter)
+{
+	static struct intr_info mps_rx_intr_info[] = {
+		{ 0xffffff, "MPS Rx parity error", -1, 1 },
+		{ 0 }
+	};
+	static struct intr_info mps_tx_intr_info[] = {
+		{ TPFIFO, "MPS Tx TP FIFO parity error", -1, 1 },
+		{ NCSIFIFO, "MPS Tx NC-SI FIFO parity error", -1, 1 },
+		{ TXDATAFIFO, "MPS Tx data FIFO parity error", -1, 1 },
+		{ TXDESCFIFO, "MPS Tx desc FIFO parity error", -1, 1 },
+		{ BUBBLE, "MPS Tx underflow", -1, 1 },
+		{ SECNTERR, "MPS Tx SOP/EOP error", -1, 1 },
+		{ FRMERR, "MPS Tx framing error", -1, 1 },
+		{ 0 }
+	};
+	static struct intr_info mps_trc_intr_info[] = {
+		{ FILTMEM, "MPS TRC filter parity error", -1, 1 },
+		{ PKTFIFO, "MPS TRC packet FIFO parity error", -1, 1 },
+		{ MISCPERR, "MPS TRC misc parity error", -1, 1 },
+		{ 0 }
+	};
+	static struct intr_info mps_stat_sram_intr_info[] = {
+		{ 0x1fffff, "MPS statistics SRAM parity error", -1, 1 },
+		{ 0 }
+	};
+	static struct intr_info mps_stat_tx_intr_info[] = {
+		{ 0xfffff, "MPS statistics Tx FIFO parity error", -1, 1 },
+		{ 0 }
+	};
+	static struct intr_info mps_stat_rx_intr_info[] = {
+		{ 0xffffff, "MPS statistics Rx FIFO parity error", -1, 1 },
+		{ 0 }
+	};
+	static struct intr_info mps_cls_intr_info[] = {
+		{ MATCHSRAM, "MPS match SRAM parity error", -1, 1 },
+		{ MATCHTCAM, "MPS match TCAM parity error", -1, 1 },
+		{ HASHSRAM, "MPS hash SRAM parity error", -1, 1 },
+		{ 0 }
+	};
+
+	int fat;
+
+	fat = t4_handle_intr_status(adapter, MPS_RX_PERR_INT_CAUSE,
+				    mps_rx_intr_info) +
+	      t4_handle_intr_status(adapter, MPS_TX_INT_CAUSE,
+				    mps_tx_intr_info) +
+	      t4_handle_intr_status(adapter, MPS_TRC_INT_CAUSE,
+				    mps_trc_intr_info) +
+	      t4_handle_intr_status(adapter, MPS_STAT_PERR_INT_CAUSE_SRAM,
+				    mps_stat_sram_intr_info) +
+	      t4_handle_intr_status(adapter, MPS_STAT_PERR_INT_CAUSE_TX_FIFO,
+				    mps_stat_tx_intr_info) +
+	      t4_handle_intr_status(adapter, MPS_STAT_PERR_INT_CAUSE_RX_FIFO,
+				    mps_stat_rx_intr_info) +
+	      t4_handle_intr_status(adapter, MPS_CLS_INT_CAUSE,
+				    mps_cls_intr_info);
+
+	t4_write_reg(adapter, MPS_INT_CAUSE, CLSINT | TRCINT |
+		     RXINT | TXINT | STATINT);
+	t4_read_reg(adapter, MPS_INT_CAUSE);                    /* flush */
+	if (fat)
+		t4_fatal_err(adapter);
+}
+
+#define MEM_INT_MASK (PERR_INT_CAUSE | ECC_CE_INT_CAUSE | ECC_UE_INT_CAUSE)
+
+/*
+ * EDC/MC interrupt handler.
+ */
+static void mem_intr_handler(struct adapter *adapter, int idx)
+{
+	static const char name[3][5] = { "EDC0", "EDC1", "MC" };
+
+	unsigned int addr, cnt_addr, v;
+
+	if (idx <= MEM_EDC1) {
+		addr = EDC_REG(EDC_INT_CAUSE, idx);
+		cnt_addr = EDC_REG(EDC_ECC_STATUS, idx);
+	} else {
+		addr = MC_INT_CAUSE;
+		cnt_addr = MC_ECC_STATUS;
+	}
+
+	v = t4_read_reg(adapter, addr) & MEM_INT_MASK;
+	if (v & PERR_INT_CAUSE)
+		dev_alert(adapter->pdev_dev, "%s FIFO parity error\n",
+			  name[idx]);
+	if (v & ECC_CE_INT_CAUSE) {
+		u32 cnt = ECC_CECNT_GET(t4_read_reg(adapter, cnt_addr));
+
+		t4_write_reg(adapter, cnt_addr, ECC_CECNT_MASK);
+		if (printk_ratelimit())
+			dev_warn(adapter->pdev_dev,
+				 "%u %s correctable ECC data error%s\n",
+				 cnt, name[idx], cnt > 1 ? "s" : "");
+	}
+	if (v & ECC_UE_INT_CAUSE)
+		dev_alert(adapter->pdev_dev,
+			  "%s uncorrectable ECC data error\n", name[idx]);
+
+	t4_write_reg(adapter, addr, v);
+	if (v & (PERR_INT_CAUSE | ECC_UE_INT_CAUSE))
+		t4_fatal_err(adapter);
+}
+
+/*
+ * MA interrupt handler.
+ */
+static void ma_intr_handler(struct adapter *adap)
+{
+	u32 v, status = t4_read_reg(adap, MA_INT_CAUSE);
+
+	if (status & MEM_PERR_INT_CAUSE)
+		dev_alert(adap->pdev_dev,
+			  "MA parity error, parity status %#x\n",
+			  t4_read_reg(adap, MA_PARITY_ERROR_STATUS));
+	if (status & MEM_WRAP_INT_CAUSE) {
+		v = t4_read_reg(adap, MA_INT_WRAP_STATUS);
+		dev_alert(adap->pdev_dev, "MA address wrap-around error by "
+			  "client %u to address %#x\n",
+			  MEM_WRAP_CLIENT_NUM_GET(v),
+			  MEM_WRAP_ADDRESS_GET(v) << 4);
+	}
+	t4_write_reg(adap, MA_INT_CAUSE, status);
+	t4_fatal_err(adap);
+}
+
+/*
+ * SMB interrupt handler.
+ */
+static void smb_intr_handler(struct adapter *adap)
+{
+	static struct intr_info smb_intr_info[] = {
+		{ MSTTXFIFOPARINT, "SMB master Tx FIFO parity error", -1, 1 },
+		{ MSTRXFIFOPARINT, "SMB master Rx FIFO parity error", -1, 1 },
+		{ SLVFIFOPARINT, "SMB slave FIFO parity error", -1, 1 },
+		{ 0 }
+	};
+
+	if (t4_handle_intr_status(adap, SMB_INT_CAUSE, smb_intr_info))
+		t4_fatal_err(adap);
+}
+
+/*
+ * NC-SI interrupt handler.
+ */
+static void ncsi_intr_handler(struct adapter *adap)
+{
+	static struct intr_info ncsi_intr_info[] = {
+		{ CIM_DM_PRTY_ERR, "NC-SI CIM parity error", -1, 1 },
+		{ MPS_DM_PRTY_ERR, "NC-SI MPS parity error", -1, 1 },
+		{ TXFIFO_PRTY_ERR, "NC-SI Tx FIFO parity error", -1, 1 },
+		{ RXFIFO_PRTY_ERR, "NC-SI Rx FIFO parity error", -1, 1 },
+		{ 0 }
+	};
+
+	if (t4_handle_intr_status(adap, NCSI_INT_CAUSE, ncsi_intr_info))
+		t4_fatal_err(adap);
+}
+
+/*
+ * XGMAC interrupt handler.
+ */
+static void xgmac_intr_handler(struct adapter *adap, int port)
+{
+	u32 v = t4_read_reg(adap, PORT_REG(port, XGMAC_PORT_INT_CAUSE));
+
+	v &= TXFIFO_PRTY_ERR | RXFIFO_PRTY_ERR;
+	if (!v)
+		return;
+
+	if (v & TXFIFO_PRTY_ERR)
+		dev_alert(adap->pdev_dev, "XGMAC %d Tx FIFO parity error\n",
+			  port);
+	if (v & RXFIFO_PRTY_ERR)
+		dev_alert(adap->pdev_dev, "XGMAC %d Rx FIFO parity error\n",
+			  port);
+	t4_write_reg(adap, PORT_REG(port, XGMAC_PORT_INT_CAUSE), v);
+	t4_fatal_err(adap);
+}
+
+/*
+ * PL interrupt handler.
+ */
+static void pl_intr_handler(struct adapter *adap)
+{
+	static struct intr_info pl_intr_info[] = {
+		{ FATALPERR, "T4 fatal parity error", -1, 1 },
+		{ PERRVFID, "PL VFID_MAP parity error", -1, 1 },
+		{ 0 }
+	};
+
+	if (t4_handle_intr_status(adap, PL_PL_INT_CAUSE, pl_intr_info))
+		t4_fatal_err(adap);
+}
+
+#define PF_INTR_MASK (PFSW | PFCIM)
+#define GLBL_INTR_MASK (CIM | MPS | PL | PCIE | MC | EDC0 | \
+		EDC1 | LE | TP | MA | PM_TX | PM_RX | ULP_RX | \
+		CPL_SWITCH | SGE | ULP_TX)
+
+/**
+ *	t4_slow_intr_handler - control path interrupt handler
+ *	@adapter: the adapter
+ *
+ *	T4 interrupt handler for non-data global interrupt events, e.g., errors.
+ *	The designation 'slow' is because it involves register reads, while
+ *	data interrupts typically don't involve any MMIOs.
+ */
+int t4_slow_intr_handler(struct adapter *adapter)
+{
+	u32 cause = t4_read_reg(adapter, PL_INT_CAUSE);
+
+	if (!(cause & GLBL_INTR_MASK))
+		return 0;
+	if (cause & CIM)
+		cim_intr_handler(adapter);
+	if (cause & MPS)
+		mps_intr_handler(adapter);
+	if (cause & NCSI)
+		ncsi_intr_handler(adapter);
+	if (cause & PL)
+		pl_intr_handler(adapter);
+	if (cause & SMB)
+		smb_intr_handler(adapter);
+	if (cause & XGMAC0)
+		xgmac_intr_handler(adapter, 0);
+	if (cause & XGMAC1)
+		xgmac_intr_handler(adapter, 1);
+	if (cause & XGMAC_KR0)
+		xgmac_intr_handler(adapter, 2);
+	if (cause & XGMAC_KR1)
+		xgmac_intr_handler(adapter, 3);
+	if (cause & PCIE)
+		pcie_intr_handler(adapter);
+	if (cause & MC)
+		mem_intr_handler(adapter, MEM_MC);
+	if (cause & EDC0)
+		mem_intr_handler(adapter, MEM_EDC0);
+	if (cause & EDC1)
+		mem_intr_handler(adapter, MEM_EDC1);
+	if (cause & LE)
+		le_intr_handler(adapter);
+	if (cause & TP)
+		tp_intr_handler(adapter);
+	if (cause & MA)
+		ma_intr_handler(adapter);
+	if (cause & PM_TX)
+		pmtx_intr_handler(adapter);
+	if (cause & PM_RX)
+		pmrx_intr_handler(adapter);
+	if (cause & ULP_RX)
+		ulprx_intr_handler(adapter);
+	if (cause & CPL_SWITCH)
+		cplsw_intr_handler(adapter);
+	if (cause & SGE)
+		sge_intr_handler(adapter);
+	if (cause & ULP_TX)
+		ulptx_intr_handler(adapter);
+
+	/* Clear the interrupts just processed for which we are the master. */
+	t4_write_reg(adapter, PL_INT_CAUSE, cause & GLBL_INTR_MASK);
+	(void) t4_read_reg(adapter, PL_INT_CAUSE); /* flush */
+	return 1;
+}
+
+/**
+ *	t4_intr_enable - enable interrupts
+ *	@adapter: the adapter whose interrupts should be enabled
+ *
+ *	Enable PF-specific interrupts for the calling function and the top-level
+ *	interrupt concentrator for global interrupts.  Interrupts are already
+ *	enabled at each module,	here we just enable the roots of the interrupt
+ *	hierarchies.
+ *
+ *	Note: this function should be called only when the driver manages
+ *	non PF-specific interrupts from the various HW modules.  Only one PCI
+ *	function at a time should be doing this.
+ */
+void t4_intr_enable(struct adapter *adapter)
+{
+	u32 pf = SOURCEPF_GET(t4_read_reg(adapter, PL_WHOAMI));
+
+	t4_write_reg(adapter, SGE_INT_ENABLE3, ERR_CPL_EXCEED_IQE_SIZE |
+		     ERR_INVALID_CIDX_INC | ERR_CPL_OPCODE_0 |
+		     ERR_DROPPED_DB | ERR_DATA_CPL_ON_HIGH_QID1 |
+		     ERR_DATA_CPL_ON_HIGH_QID0 | ERR_BAD_DB_PIDX3 |
+		     ERR_BAD_DB_PIDX2 | ERR_BAD_DB_PIDX1 |
+		     ERR_BAD_DB_PIDX0 | ERR_ING_CTXT_PRIO |
+		     ERR_EGR_CTXT_PRIO | INGRESS_SIZE_ERR |
+		     EGRESS_SIZE_ERR);
+	t4_write_reg(adapter, MYPF_REG(PL_PF_INT_ENABLE), PF_INTR_MASK);
+	t4_set_reg_field(adapter, PL_INT_MAP0, 0, 1 << pf);
+}
+
+/**
+ *	t4_intr_disable - disable interrupts
+ *	@adapter: the adapter whose interrupts should be disabled
+ *
+ *	Disable interrupts.  We only disable the top-level interrupt
+ *	concentrators.  The caller must be a PCI function managing global
+ *	interrupts.
+ */
+void t4_intr_disable(struct adapter *adapter)
+{
+	u32 pf = SOURCEPF_GET(t4_read_reg(adapter, PL_WHOAMI));
+
+	t4_write_reg(adapter, MYPF_REG(PL_PF_INT_ENABLE), 0);
+	t4_set_reg_field(adapter, PL_INT_MAP0, 1 << pf, 0);
+}
+
+/**
+ *	t4_intr_clear - clear all interrupts
+ *	@adapter: the adapter whose interrupts should be cleared
+ *
+ *	Clears all interrupts.  The caller must be a PCI function managing
+ *	global interrupts.
+ */
+void t4_intr_clear(struct adapter *adapter)
+{
+	static const unsigned int cause_reg[] = {
+		SGE_INT_CAUSE1, SGE_INT_CAUSE2, SGE_INT_CAUSE3,
+		PCIE_CORE_UTL_SYSTEM_BUS_AGENT_STATUS,
+		PCIE_CORE_UTL_PCI_EXPRESS_PORT_STATUS,
+		PCIE_NONFAT_ERR, PCIE_INT_CAUSE,
+		MC_INT_CAUSE,
+		MA_INT_WRAP_STATUS, MA_PARITY_ERROR_STATUS, MA_INT_CAUSE,
+		EDC_INT_CAUSE, EDC_REG(EDC_INT_CAUSE, 1),
+		CIM_HOST_INT_CAUSE, CIM_HOST_UPACC_INT_CAUSE,
+		MYPF_REG(CIM_PF_HOST_INT_CAUSE),
+		TP_INT_CAUSE,
+		ULP_RX_INT_CAUSE, ULP_TX_INT_CAUSE,
+		PM_RX_INT_CAUSE, PM_TX_INT_CAUSE,
+		MPS_RX_PERR_INT_CAUSE,
+		CPL_INTR_CAUSE,
+		MYPF_REG(PL_PF_INT_CAUSE),
+		PL_PL_INT_CAUSE,
+		LE_DB_INT_CAUSE,
+	};
+
+	unsigned int i;
+
+	for (i = 0; i < ARRAY_SIZE(cause_reg); ++i)
+		t4_write_reg(adapter, cause_reg[i], 0xffffffff);
+
+	t4_write_reg(adapter, PL_INT_CAUSE, GLBL_INTR_MASK);
+	(void) t4_read_reg(adapter, PL_INT_CAUSE);          /* flush */
+}
+
+/**
+ *	hash_mac_addr - return the hash value of a MAC address
+ *	@addr: the 48-bit Ethernet MAC address
+ *
+ *	Hashes a MAC address according to the hash function used by HW inexact
+ *	(hash) address matching.
+ */
+static int hash_mac_addr(const u8 *addr)
+{
+	u32 a = ((u32)addr[0] << 16) | ((u32)addr[1] << 8) | addr[2];
+	u32 b = ((u32)addr[3] << 16) | ((u32)addr[4] << 8) | addr[5];
+	a ^= b;
+	a ^= (a >> 12);
+	a ^= (a >> 6);
+	return a & 0x3f;
+}
+
+/**
+ *	t4_config_rss_range - configure a portion of the RSS mapping table
+ *	@adapter: the adapter
+ *	@mbox: mbox to use for the FW command
+ *	@viid: virtual interface whose RSS subtable is to be written
+ *	@start: start entry in the table to write
+ *	@n: how many table entries to write
+ *	@rspq: values for the response queue lookup table
+ *	@nrspq: number of values in @rspq
+ *
+ *	Programs the selected part of the VI's RSS mapping table with the
+ *	provided values.  If @nrspq < @n the supplied values are used repeatedly
+ *	until the full table range is populated.
+ *
+ *	The caller must ensure the values in @rspq are in the range allowed for
+ *	@viid.
+ */
+int t4_config_rss_range(struct adapter *adapter, int mbox, unsigned int viid,
+			int start, int n, const u16 *rspq, unsigned int nrspq)
+{
+	int ret;
+	const u16 *rsp = rspq;
+	const u16 *rsp_end = rspq + nrspq;
+	struct fw_rss_ind_tbl_cmd cmd;
+
+	memset(&cmd, 0, sizeof(cmd));
+	cmd.op_to_viid = htonl(FW_CMD_OP(FW_RSS_IND_TBL_CMD) |
+			       FW_CMD_REQUEST | FW_CMD_WRITE |
+			       FW_RSS_IND_TBL_CMD_VIID(viid));
+	cmd.retval_len16 = htonl(FW_LEN16(cmd));
+
+	/* each fw_rss_ind_tbl_cmd takes up to 32 entries */
+	while (n > 0) {
+		int nq = min(n, 32);
+		__be32 *qp = &cmd.iq0_to_iq2;
+
+		cmd.niqid = htons(nq);
+		cmd.startidx = htons(start);
+
+		start += nq;
+		n -= nq;
+
+		while (nq > 0) {
+			unsigned int v;
+
+			v = FW_RSS_IND_TBL_CMD_IQ0(*rsp);
+			if (++rsp >= rsp_end)
+				rsp = rspq;
+			v |= FW_RSS_IND_TBL_CMD_IQ1(*rsp);
+			if (++rsp >= rsp_end)
+				rsp = rspq;
+			v |= FW_RSS_IND_TBL_CMD_IQ2(*rsp);
+			if (++rsp >= rsp_end)
+				rsp = rspq;
+
+			*qp++ = htonl(v);
+			nq -= 3;
+		}
+
+		ret = t4_wr_mbox(adapter, mbox, &cmd, sizeof(cmd), NULL);
+		if (ret)
+			return ret;
+	}
+	return 0;
+}
+
+/**
+ *	t4_config_glbl_rss - configure the global RSS mode
+ *	@adapter: the adapter
+ *	@mbox: mbox to use for the FW command
+ *	@mode: global RSS mode
+ *	@flags: mode-specific flags
+ *
+ *	Sets the global RSS mode.
+ */
+int t4_config_glbl_rss(struct adapter *adapter, int mbox, unsigned int mode,
+		       unsigned int flags)
+{
+	struct fw_rss_glb_config_cmd c;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_write = htonl(FW_CMD_OP(FW_RSS_GLB_CONFIG_CMD) |
+			      FW_CMD_REQUEST | FW_CMD_WRITE);
+	c.retval_len16 = htonl(FW_LEN16(c));
+	if (mode == FW_RSS_GLB_CONFIG_CMD_MODE_MANUAL) {
+		c.u.manual.mode_pkd = htonl(FW_RSS_GLB_CONFIG_CMD_MODE(mode));
+	} else if (mode == FW_RSS_GLB_CONFIG_CMD_MODE_BASICVIRTUAL) {
+		c.u.basicvirtual.mode_pkd =
+			htonl(FW_RSS_GLB_CONFIG_CMD_MODE(mode));
+		c.u.basicvirtual.synmapen_to_hashtoeplitz = htonl(flags);
+	} else
+		return -EINVAL;
+	return t4_wr_mbox(adapter, mbox, &c, sizeof(c), NULL);
+}
+
+/* Read an RSS table row */
+static int rd_rss_row(struct adapter *adap, int row, u32 *val)
+{
+	t4_write_reg(adap, TP_RSS_LKP_TABLE, 0xfff00000 | row);
+	return t4_wait_op_done_val(adap, TP_RSS_LKP_TABLE, LKPTBLROWVLD, 1,
+				   5, 0, val);
+}
+
+/**
+ *	t4_read_rss - read the contents of the RSS mapping table
+ *	@adapter: the adapter
+ *	@map: holds the contents of the RSS mapping table
+ *
+ *	Reads the contents of the RSS hash->queue mapping table.
+ */
+int t4_read_rss(struct adapter *adapter, u16 *map)
+{
+	u32 val;
+	int i, ret;
+
+	for (i = 0; i < RSS_NENTRIES / 2; ++i) {
+		ret = rd_rss_row(adapter, i, &val);
+		if (ret)
+			return ret;
+		*map++ = LKPTBLQUEUE0_GET(val);
+		*map++ = LKPTBLQUEUE1_GET(val);
+	}
+	return 0;
+}
+
+/**
+ *	t4_tp_get_tcp_stats - read TP's TCP MIB counters
+ *	@adap: the adapter
+ *	@v4: holds the TCP/IP counter values
+ *	@v6: holds the TCP/IPv6 counter values
+ *
+ *	Returns the values of TP's TCP/IP and TCP/IPv6 MIB counters.
+ *	Either @v4 or @v6 may be %NULL to skip the corresponding stats.
+ */
+void t4_tp_get_tcp_stats(struct adapter *adap, struct tp_tcp_stats *v4,
+			 struct tp_tcp_stats *v6)
+{
+	u32 val[TP_MIB_TCP_RXT_SEG_LO - TP_MIB_TCP_OUT_RST + 1];
+
+#define STAT_IDX(x) ((TP_MIB_TCP_##x) - TP_MIB_TCP_OUT_RST)
+#define STAT(x)     val[STAT_IDX(x)]
+#define STAT64(x)   (((u64)STAT(x##_HI) << 32) | STAT(x##_LO))
+
+	if (v4) {
+		t4_read_indirect(adap, TP_MIB_INDEX, TP_MIB_DATA, val,
+				 ARRAY_SIZE(val), TP_MIB_TCP_OUT_RST);
+		v4->tcpOutRsts = STAT(OUT_RST);
+		v4->tcpInSegs  = STAT64(IN_SEG);
+		v4->tcpOutSegs = STAT64(OUT_SEG);
+		v4->tcpRetransSegs = STAT64(RXT_SEG);
+	}
+	if (v6) {
+		t4_read_indirect(adap, TP_MIB_INDEX, TP_MIB_DATA, val,
+				 ARRAY_SIZE(val), TP_MIB_TCP_V6OUT_RST);
+		v6->tcpOutRsts = STAT(OUT_RST);
+		v6->tcpInSegs  = STAT64(IN_SEG);
+		v6->tcpOutSegs = STAT64(OUT_SEG);
+		v6->tcpRetransSegs = STAT64(RXT_SEG);
+	}
+#undef STAT64
+#undef STAT
+#undef STAT_IDX
+}
+
+/**
+ *	t4_tp_get_err_stats - read TP's error MIB counters
+ *	@adap: the adapter
+ *	@st: holds the counter values
+ *
+ *	Returns the values of TP's error counters.
+ */
+void t4_tp_get_err_stats(struct adapter *adap, struct tp_err_stats *st)
+{
+	t4_read_indirect(adap, TP_MIB_INDEX, TP_MIB_DATA, st->macInErrs,
+			 12, TP_MIB_MAC_IN_ERR_0);
+	t4_read_indirect(adap, TP_MIB_INDEX, TP_MIB_DATA, st->tnlCongDrops,
+			 8, TP_MIB_TNL_CNG_DROP_0);
+	t4_read_indirect(adap, TP_MIB_INDEX, TP_MIB_DATA, st->tnlTxDrops,
+			 4, TP_MIB_TNL_DROP_0);
+	t4_read_indirect(adap, TP_MIB_INDEX, TP_MIB_DATA, st->ofldVlanDrops,
+			 4, TP_MIB_OFD_VLN_DROP_0);
+	t4_read_indirect(adap, TP_MIB_INDEX, TP_MIB_DATA, st->tcp6InErrs,
+			 4, TP_MIB_TCP_V6IN_ERR_0);
+	t4_read_indirect(adap, TP_MIB_INDEX, TP_MIB_DATA, &st->ofldNoNeigh,
+			 2, TP_MIB_OFD_ARP_DROP);
+}
+
+/**
+ *	t4_read_mtu_tbl - returns the values in the HW path MTU table
+ *	@adap: the adapter
+ *	@mtus: where to store the MTU values
+ *	@mtu_log: where to store the MTU base-2 log (may be %NULL)
+ *
+ *	Reads the HW path MTU table.
+ */
+void t4_read_mtu_tbl(struct adapter *adap, u16 *mtus, u8 *mtu_log)
+{
+	u32 v;
+	int i;
+
+	for (i = 0; i < NMTUS; ++i) {
+		t4_write_reg(adap, TP_MTU_TABLE,
+			     MTUINDEX(0xff) | MTUVALUE(i));
+		v = t4_read_reg(adap, TP_MTU_TABLE);
+		mtus[i] = MTUVALUE_GET(v);
+		if (mtu_log)
+			mtu_log[i] = MTUWIDTH_GET(v);
+	}
+}
+
+/**
+ *	init_cong_ctrl - initialize congestion control parameters
+ *	@a: the alpha values for congestion control
+ *	@b: the beta values for congestion control
+ *
+ *	Initialize the congestion control parameters.
+ */
+static void __devinit init_cong_ctrl(unsigned short *a, unsigned short *b)
+{
+	a[0] = a[1] = a[2] = a[3] = a[4] = a[5] = a[6] = a[7] = a[8] = 1;
+	a[9] = 2;
+	a[10] = 3;
+	a[11] = 4;
+	a[12] = 5;
+	a[13] = 6;
+	a[14] = 7;
+	a[15] = 8;
+	a[16] = 9;
+	a[17] = 10;
+	a[18] = 14;
+	a[19] = 17;
+	a[20] = 21;
+	a[21] = 25;
+	a[22] = 30;
+	a[23] = 35;
+	a[24] = 45;
+	a[25] = 60;
+	a[26] = 80;
+	a[27] = 100;
+	a[28] = 200;
+	a[29] = 300;
+	a[30] = 400;
+	a[31] = 500;
+
+	b[0] = b[1] = b[2] = b[3] = b[4] = b[5] = b[6] = b[7] = b[8] = 0;
+	b[9] = b[10] = 1;
+	b[11] = b[12] = 2;
+	b[13] = b[14] = b[15] = b[16] = 3;
+	b[17] = b[18] = b[19] = b[20] = b[21] = 4;
+	b[22] = b[23] = b[24] = b[25] = b[26] = b[27] = 5;
+	b[28] = b[29] = 6;
+	b[30] = b[31] = 7;
+}
+
+/* The minimum additive increment value for the congestion control table */
+#define CC_MIN_INCR 2U
+
+/**
+ *	t4_load_mtus - write the MTU and congestion control HW tables
+ *	@adap: the adapter
+ *	@mtus: the values for the MTU table
+ *	@alpha: the values for the congestion control alpha parameter
+ *	@beta: the values for the congestion control beta parameter
+ *
+ *	Write the HW MTU table with the supplied MTUs and the high-speed
+ *	congestion control table with the supplied alpha, beta, and MTUs.
+ *	We write the two tables together because the additive increments
+ *	depend on the MTUs.
+ */
+void t4_load_mtus(struct adapter *adap, const unsigned short *mtus,
+		  const unsigned short *alpha, const unsigned short *beta)
+{
+	static const unsigned int avg_pkts[NCCTRL_WIN] = {
+		2, 6, 10, 14, 20, 28, 40, 56, 80, 112, 160, 224, 320, 448, 640,
+		896, 1281, 1792, 2560, 3584, 5120, 7168, 10240, 14336, 20480,
+		28672, 40960, 57344, 81920, 114688, 163840, 229376
+	};
+
+	unsigned int i, w;
+
+	for (i = 0; i < NMTUS; ++i) {
+		unsigned int mtu = mtus[i];
+		unsigned int log2 = fls(mtu);
+
+		if (!(mtu & ((1 << log2) >> 2)))     /* round */
+			log2--;
+		t4_write_reg(adap, TP_MTU_TABLE, MTUINDEX(i) |
+			     MTUWIDTH(log2) | MTUVALUE(mtu));
+
+		for (w = 0; w < NCCTRL_WIN; ++w) {
+			unsigned int inc;
+
+			inc = max(((mtu - 40) * alpha[w]) / avg_pkts[w],
+				  CC_MIN_INCR);
+
+			t4_write_reg(adap, TP_CCTRL_TABLE, (i << 21) |
+				     (w << 16) | (beta[w] << 13) | inc);
+		}
+	}
+}
+
+/**
+ *	t4_set_trace_filter - configure one of the tracing filters
+ *	@adap: the adapter
+ *	@tp: the desired trace filter parameters
+ *	@idx: which filter to configure
+ *	@enable: whether to enable or disable the filter
+ *
+ *	Configures one of the tracing filters available in HW.  If @enable is
+ *	%0 @tp is not examined and may be %NULL.
+ */
+int t4_set_trace_filter(struct adapter *adap, const struct trace_params *tp,
+			int idx, int enable)
+{
+	int i, ofst = idx * 4;
+	u32 data_reg, mask_reg, cfg;
+	u32 multitrc = TRCMULTIFILTER;
+
+	if (!enable) {
+		t4_write_reg(adap, MPS_TRC_FILTER_MATCH_CTL_A + ofst, 0);
+		goto out;
+	}
+
+	if (tp->port > 11 || tp->invert > 1 || tp->skip_len > 0x1f ||
+	    tp->skip_ofst > 0x1f || tp->min_len > 0x1ff ||
+	    tp->snap_len > 9600 || (idx && tp->snap_len > 256))
+		return -EINVAL;
+
+	if (tp->snap_len > 256) {            /* must be tracer 0 */
+		if ((t4_read_reg(adap, MPS_TRC_FILTER_MATCH_CTL_A + 4) |
+		     t4_read_reg(adap, MPS_TRC_FILTER_MATCH_CTL_A + 8) |
+		     t4_read_reg(adap, MPS_TRC_FILTER_MATCH_CTL_A + 12)) & TFEN)
+			return -EINVAL;  /* other tracers are enabled */
+		multitrc = 0;
+	} else if (idx) {
+		i = t4_read_reg(adap, MPS_TRC_FILTER_MATCH_CTL_B);
+		if (TFCAPTUREMAX_GET(i) > 256 &&
+		    (t4_read_reg(adap, MPS_TRC_FILTER_MATCH_CTL_A) & TFEN))
+			return -EINVAL;
+	}
+
+	/* stop the tracer we'll be changing */
+	t4_write_reg(adap, MPS_TRC_FILTER_MATCH_CTL_A + ofst, 0);
+
+	/* disable tracing globally if running in the wrong single/multi mode */
+	cfg = t4_read_reg(adap, MPS_TRC_CFG);
+	if ((cfg & TRCEN) && multitrc != (cfg & TRCMULTIFILTER)) {
+		t4_write_reg(adap, MPS_TRC_CFG, cfg ^ TRCEN);
+		t4_read_reg(adap, MPS_TRC_CFG);                  /* flush */
+		msleep(1);
+		if (!(t4_read_reg(adap, MPS_TRC_CFG) & TRCFIFOEMPTY))
+			return -ETIMEDOUT;
+	}
+	/*
+	 * At this point either the tracing is enabled and in the right mode or
+	 * disabled.
+	 */
+
+	idx *= (MPS_TRC_FILTER1_MATCH - MPS_TRC_FILTER0_MATCH);
+	data_reg = MPS_TRC_FILTER0_MATCH + idx;
+	mask_reg = MPS_TRC_FILTER0_DONT_CARE + idx;
+
+	for (i = 0; i < TRACE_LEN / 4; i++, data_reg += 4, mask_reg += 4) {
+		t4_write_reg(adap, data_reg, tp->data[i]);
+		t4_write_reg(adap, mask_reg, ~tp->mask[i]);
+	}
+	t4_write_reg(adap, MPS_TRC_FILTER_MATCH_CTL_B + ofst,
+		     TFCAPTUREMAX(tp->snap_len) |
+		     TFMINPKTSIZE(tp->min_len));
+	t4_write_reg(adap, MPS_TRC_FILTER_MATCH_CTL_A + ofst,
+		     TFOFFSET(tp->skip_ofst) | TFLENGTH(tp->skip_len) |
+		     TFPORT(tp->port) | TFEN |
+		     (tp->invert ? TFINVERTMATCH : 0));
+
+	cfg &= ~TRCMULTIFILTER;
+	t4_write_reg(adap, MPS_TRC_CFG, cfg | TRCEN | multitrc);
+out:	t4_read_reg(adap, MPS_TRC_CFG);  /* flush */
+	return 0;
+}
+
+/**
+ *	t4_get_trace_filter - query one of the tracing filters
+ *	@adap: the adapter
+ *	@tp: the current trace filter parameters
+ *	@idx: which trace filter to query
+ *	@enabled: non-zero if the filter is enabled
+ *
+ *	Returns the current settings of one of the HW tracing filters.
+ */
+void t4_get_trace_filter(struct adapter *adap, struct trace_params *tp, int idx,
+			 int *enabled)
+{
+	u32 ctla, ctlb;
+	int i, ofst = idx * 4;
+	u32 data_reg, mask_reg;
+
+	ctla = t4_read_reg(adap, MPS_TRC_FILTER_MATCH_CTL_A + ofst);
+	ctlb = t4_read_reg(adap, MPS_TRC_FILTER_MATCH_CTL_B + ofst);
+
+	*enabled = !!(ctla & TFEN);
+	tp->snap_len = TFCAPTUREMAX_GET(ctlb);
+	tp->min_len = TFMINPKTSIZE_GET(ctlb);
+	tp->skip_ofst = TFOFFSET_GET(ctla);
+	tp->skip_len = TFLENGTH_GET(ctla);
+	tp->invert = !!(ctla & TFINVERTMATCH);
+	tp->port = TFPORT_GET(ctla);
+
+	ofst = (MPS_TRC_FILTER1_MATCH - MPS_TRC_FILTER0_MATCH) * idx;
+	data_reg = MPS_TRC_FILTER0_MATCH + ofst;
+	mask_reg = MPS_TRC_FILTER0_DONT_CARE + ofst;
+
+	for (i = 0; i < TRACE_LEN / 4; i++, data_reg += 4, mask_reg += 4) {
+		tp->mask[i] = ~t4_read_reg(adap, mask_reg);
+		tp->data[i] = t4_read_reg(adap, data_reg) & tp->mask[i];
+	}
+}
+
+/**
+ *	get_mps_bg_map - return the buffer groups associated with a port
+ *	@adap: the adapter
+ *	@idx: the port index
+ *
+ *	Returns a bitmap indicating which MPS buffer groups are associated
+ *	with the given port.  Bit i is set if buffer group i is used by the
+ *	port.
+ */
+static unsigned int get_mps_bg_map(struct adapter *adap, int idx)
+{
+	u32 n = NUMPORTS_GET(t4_read_reg(adap, MPS_CMN_CTL));
+
+	if (n == 0)
+		return idx == 0 ? 0xf : 0;
+	if (n == 1)
+		return idx < 2 ? (3 << (2 * idx)) : 0;
+	return 1 << idx;
+}
+
+/**
+ *	t4_get_port_stats - collect port statistics
+ *	@adap: the adapter
+ *	@idx: the port index
+ *	@p: the stats structure to fill
+ *
+ *	Collect statistics related to the given port from HW.
+ */
+void t4_get_port_stats(struct adapter *adap, int idx, struct port_stats *p)
+{
+	u32 bgmap = get_mps_bg_map(adap, idx);
+
+#define GET_STAT(name) \
+	t4_read_reg64(adap, PORT_REG(idx, MPS_PORT_STAT_##name##_L))
+#define GET_STAT_COM(name) t4_read_reg64(adap, MPS_STAT_##name##_L)
+
+	p->tx_octets           = GET_STAT(TX_PORT_BYTES);
+	p->tx_frames           = GET_STAT(TX_PORT_FRAMES);
+	p->tx_bcast_frames     = GET_STAT(TX_PORT_BCAST);
+	p->tx_mcast_frames     = GET_STAT(TX_PORT_MCAST);
+	p->tx_ucast_frames     = GET_STAT(TX_PORT_UCAST);
+	p->tx_error_frames     = GET_STAT(TX_PORT_ERROR);
+	p->tx_frames_64        = GET_STAT(TX_PORT_64B);
+	p->tx_frames_65_127    = GET_STAT(TX_PORT_65B_127B);
+	p->tx_frames_128_255   = GET_STAT(TX_PORT_128B_255B);
+	p->tx_frames_256_511   = GET_STAT(TX_PORT_256B_511B);
+	p->tx_frames_512_1023  = GET_STAT(TX_PORT_512B_1023B);
+	p->tx_frames_1024_1518 = GET_STAT(TX_PORT_1024B_1518B);
+	p->tx_frames_1519_max  = GET_STAT(TX_PORT_1519B_MAX);
+	p->tx_drop             = GET_STAT(TX_PORT_DROP);
+	p->tx_pause            = GET_STAT(TX_PORT_PAUSE);
+	p->tx_ppp0             = GET_STAT(TX_PORT_PPP0);
+	p->tx_ppp1             = GET_STAT(TX_PORT_PPP1);
+	p->tx_ppp2             = GET_STAT(TX_PORT_PPP2);
+	p->tx_ppp3             = GET_STAT(TX_PORT_PPP3);
+	p->tx_ppp4             = GET_STAT(TX_PORT_PPP4);
+	p->tx_ppp5             = GET_STAT(TX_PORT_PPP5);
+	p->tx_ppp6             = GET_STAT(TX_PORT_PPP6);
+	p->tx_ppp7             = GET_STAT(TX_PORT_PPP7);
+
+	p->rx_octets           = GET_STAT(RX_PORT_BYTES);
+	p->rx_frames           = GET_STAT(RX_PORT_FRAMES);
+	p->rx_bcast_frames     = GET_STAT(RX_PORT_BCAST);
+	p->rx_mcast_frames     = GET_STAT(RX_PORT_MCAST);
+	p->rx_ucast_frames     = GET_STAT(RX_PORT_UCAST);
+	p->rx_too_long         = GET_STAT(RX_PORT_MTU_ERROR);
+	p->rx_jabber           = GET_STAT(RX_PORT_MTU_CRC_ERROR);
+	p->rx_fcs_err          = GET_STAT(RX_PORT_CRC_ERROR);
+	p->rx_len_err          = GET_STAT(RX_PORT_LEN_ERROR);
+	p->rx_symbol_err       = GET_STAT(RX_PORT_SYM_ERROR);
+	p->rx_runt             = GET_STAT(RX_PORT_LESS_64B);
+	p->rx_frames_64        = GET_STAT(RX_PORT_64B);
+	p->rx_frames_65_127    = GET_STAT(RX_PORT_65B_127B);
+	p->rx_frames_128_255   = GET_STAT(RX_PORT_128B_255B);
+	p->rx_frames_256_511   = GET_STAT(RX_PORT_256B_511B);
+	p->rx_frames_512_1023  = GET_STAT(RX_PORT_512B_1023B);
+	p->rx_frames_1024_1518 = GET_STAT(RX_PORT_1024B_1518B);
+	p->rx_frames_1519_max  = GET_STAT(RX_PORT_1519B_MAX);
+	p->rx_pause            = GET_STAT(RX_PORT_PAUSE);
+	p->rx_ppp0             = GET_STAT(RX_PORT_PPP0);
+	p->rx_ppp1             = GET_STAT(RX_PORT_PPP1);
+	p->rx_ppp2             = GET_STAT(RX_PORT_PPP2);
+	p->rx_ppp3             = GET_STAT(RX_PORT_PPP3);
+	p->rx_ppp4             = GET_STAT(RX_PORT_PPP4);
+	p->rx_ppp5             = GET_STAT(RX_PORT_PPP5);
+	p->rx_ppp6             = GET_STAT(RX_PORT_PPP6);
+	p->rx_ppp7             = GET_STAT(RX_PORT_PPP7);
+
+	p->rx_ovflow0 = (bgmap & 1) ? GET_STAT_COM(RX_BG_0_MAC_DROP_FRAME) : 0;
+	p->rx_ovflow1 = (bgmap & 2) ? GET_STAT_COM(RX_BG_1_MAC_DROP_FRAME) : 0;
+	p->rx_ovflow2 = (bgmap & 4) ? GET_STAT_COM(RX_BG_2_MAC_DROP_FRAME) : 0;
+	p->rx_ovflow3 = (bgmap & 8) ? GET_STAT_COM(RX_BG_3_MAC_DROP_FRAME) : 0;
+	p->rx_trunc0 = (bgmap & 1) ? GET_STAT_COM(RX_BG_0_MAC_TRUNC_FRAME) : 0;
+	p->rx_trunc1 = (bgmap & 2) ? GET_STAT_COM(RX_BG_1_MAC_TRUNC_FRAME) : 0;
+	p->rx_trunc2 = (bgmap & 4) ? GET_STAT_COM(RX_BG_2_MAC_TRUNC_FRAME) : 0;
+	p->rx_trunc3 = (bgmap & 8) ? GET_STAT_COM(RX_BG_3_MAC_TRUNC_FRAME) : 0;
+
+#undef GET_STAT
+#undef GET_STAT_COM
+}
+
+/**
+ *	t4_get_lb_stats - collect loopback port statistics
+ *	@adap: the adapter
+ *	@idx: the loopback port index
+ *	@p: the stats structure to fill
+ *
+ *	Return HW statistics for the given loopback port.
+ */
+void t4_get_lb_stats(struct adapter *adap, int idx, struct lb_port_stats *p)
+{
+	u32 bgmap = get_mps_bg_map(adap, idx);
+
+#define GET_STAT(name) \
+	t4_read_reg64(adap, PORT_REG(idx, MPS_PORT_STAT_LB_PORT_##name##_L))
+#define GET_STAT_COM(name) t4_read_reg64(adap, MPS_STAT_##name##_L)
+
+	p->octets           = GET_STAT(BYTES);
+	p->frames           = GET_STAT(FRAMES);
+	p->bcast_frames     = GET_STAT(BCAST);
+	p->mcast_frames     = GET_STAT(MCAST);
+	p->ucast_frames     = GET_STAT(UCAST);
+	p->error_frames     = GET_STAT(ERROR);
+
+	p->frames_64        = GET_STAT(64B);
+	p->frames_65_127    = GET_STAT(65B_127B);
+	p->frames_128_255   = GET_STAT(128B_255B);
+	p->frames_256_511   = GET_STAT(256B_511B);
+	p->frames_512_1023  = GET_STAT(512B_1023B);
+	p->frames_1024_1518 = GET_STAT(1024B_1518B);
+	p->frames_1519_max  = GET_STAT(1519B_MAX);
+	p->drop             = t4_read_reg(adap, PORT_REG(idx,
+					  MPS_PORT_STAT_LB_PORT_DROP_FRAMES));
+
+	p->ovflow0 = (bgmap & 1) ? GET_STAT_COM(RX_BG_0_LB_DROP_FRAME) : 0;
+	p->ovflow1 = (bgmap & 2) ? GET_STAT_COM(RX_BG_1_LB_DROP_FRAME) : 0;
+	p->ovflow2 = (bgmap & 4) ? GET_STAT_COM(RX_BG_2_LB_DROP_FRAME) : 0;
+	p->ovflow3 = (bgmap & 8) ? GET_STAT_COM(RX_BG_3_LB_DROP_FRAME) : 0;
+	p->trunc0 = (bgmap & 1) ? GET_STAT_COM(RX_BG_0_LB_TRUNC_FRAME) : 0;
+	p->trunc1 = (bgmap & 2) ? GET_STAT_COM(RX_BG_1_LB_TRUNC_FRAME) : 0;
+	p->trunc2 = (bgmap & 4) ? GET_STAT_COM(RX_BG_2_LB_TRUNC_FRAME) : 0;
+	p->trunc3 = (bgmap & 8) ? GET_STAT_COM(RX_BG_3_LB_TRUNC_FRAME) : 0;
+
+#undef GET_STAT
+#undef GET_STAT_COM
+}
+
+/**
+ *	t4_wol_magic_enable - enable/disable magic packet WoL
+ *	@adap: the adapter
+ *	@port: the physical port index
+ *	@addr: MAC address expected in magic packets, %NULL to disable
+ *
+ *	Enables/disables magic packet wake-on-LAN for the selected port.
+ */
+void t4_wol_magic_enable(struct adapter *adap, unsigned int port,
+			 const u8 *addr)
+{
+	if (addr) {
+		t4_write_reg(adap, PORT_REG(port, XGMAC_PORT_MAGIC_MACID_LO),
+			     (addr[2] << 24) | (addr[3] << 16) |
+			     (addr[4] << 8) | addr[5]);
+		t4_write_reg(adap, PORT_REG(port, XGMAC_PORT_MAGIC_MACID_HI),
+			     (addr[0] << 8) | addr[1]);
+	}
+	t4_set_reg_field(adap, PORT_REG(port, XGMAC_PORT_CFG2), MAGICEN,
+			 addr ? MAGICEN : 0);
+}
+
+/**
+ *	t4_wol_pat_enable - enable/disable pattern-based WoL
+ *	@adap: the adapter
+ *	@port: the physical port index
+ *	@map: bitmap of which HW pattern filters to set
+ *	@mask0: byte mask for bytes 0-63 of a packet
+ *	@mask1: byte mask for bytes 64-127 of a packet
+ *	@crc: Ethernet CRC for selected bytes
+ *	@enable: enable/disable switch
+ *
+ *	Sets the pattern filters indicated in @map to mask out the bytes
+ *	specified in @mask0/@mask1 in received packets and compare the CRC of
+ *	the resulting packet against @crc.  If @enable is %true pattern-based
+ *	WoL is enabled, otherwise disabled.
+ */
+int t4_wol_pat_enable(struct adapter *adap, unsigned int port, unsigned int map,
+		      u64 mask0, u64 mask1, unsigned int crc, bool enable)
+{
+	int i;
+
+	if (!enable) {
+		t4_set_reg_field(adap, PORT_REG(port, XGMAC_PORT_CFG2),
+				 PATEN, 0);
+		return 0;
+	}
+	if (map > 0xff)
+		return -EINVAL;
+
+#define EPIO_REG(name) PORT_REG(port, XGMAC_PORT_EPIO_##name)
+
+	t4_write_reg(adap, EPIO_REG(DATA1), mask0 >> 32);
+	t4_write_reg(adap, EPIO_REG(DATA2), mask1);
+	t4_write_reg(adap, EPIO_REG(DATA3), mask1 >> 32);
+
+	for (i = 0; i < NWOL_PAT; i++, map >>= 1) {
+		if (!(map & 1))
+			continue;
+
+		/* write byte masks */
+		t4_write_reg(adap, EPIO_REG(DATA0), mask0);
+		t4_write_reg(adap, EPIO_REG(OP), ADDRESS(i) | EPIOWR);
+		t4_read_reg(adap, EPIO_REG(OP));                /* flush */
+		if (t4_read_reg(adap, EPIO_REG(OP)) & BUSY)
+			return -ETIMEDOUT;
+
+		/* write CRC */
+		t4_write_reg(adap, EPIO_REG(DATA0), crc);
+		t4_write_reg(adap, EPIO_REG(OP), ADDRESS(i + 32) | EPIOWR);
+		t4_read_reg(adap, EPIO_REG(OP));                /* flush */
+		if (t4_read_reg(adap, EPIO_REG(OP)) & BUSY)
+			return -ETIMEDOUT;
+	}
+#undef EPIO_REG
+
+	t4_set_reg_field(adap, PORT_REG(port, XGMAC_PORT_CFG2), 0, PATEN);
+	return 0;
+}
+
+#define INIT_CMD(var, cmd, rd_wr) do { \
+	(var).op_to_write = htonl(FW_CMD_OP(FW_##cmd##_CMD) | \
+				  FW_CMD_REQUEST | FW_CMD_##rd_wr); \
+	(var).retval_len16 = htonl(FW_LEN16(var)); \
+} while (0)
+
+/**
+ *	t4_mdio_rd - read a PHY register through MDIO
+ *	@adap: the adapter
+ *	@mbox: mailbox to use for the FW command
+ *	@phy_addr: the PHY address
+ *	@mmd: the PHY MMD to access (0 for clause 22 PHYs)
+ *	@reg: the register to read
+ *	@valp: where to store the value
+ *
+ *	Issues a FW command through the given mailbox to read a PHY register.
+ */
+int t4_mdio_rd(struct adapter *adap, unsigned int mbox, unsigned int phy_addr,
+	       unsigned int mmd, unsigned int reg, u16 *valp)
+{
+	int ret;
+	struct fw_ldst_cmd c;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_addrspace = htonl(FW_CMD_OP(FW_LDST_CMD) | FW_CMD_REQUEST |
+		FW_CMD_READ | FW_LDST_CMD_ADDRSPACE(FW_LDST_ADDRSPC_MDIO));
+	c.cycles_to_len16 = htonl(FW_LEN16(c));
+	c.u.mdio.paddr_mmd = htons(FW_LDST_CMD_PADDR(phy_addr) |
+				   FW_LDST_CMD_MMD(mmd));
+	c.u.mdio.raddr = htons(reg);
+
+	ret = t4_wr_mbox(adap, mbox, &c, sizeof(c), &c);
+	if (ret == 0)
+		*valp = ntohs(c.u.mdio.rval);
+	return ret;
+}
+
+/**
+ *	t4_mdio_wr - write a PHY register through MDIO
+ *	@adap: the adapter
+ *	@mbox: mailbox to use for the FW command
+ *	@phy_addr: the PHY address
+ *	@mmd: the PHY MMD to access (0 for clause 22 PHYs)
+ *	@reg: the register to write
+ *	@valp: value to write
+ *
+ *	Issues a FW command through the given mailbox to write a PHY register.
+ */
+int t4_mdio_wr(struct adapter *adap, unsigned int mbox, unsigned int phy_addr,
+	       unsigned int mmd, unsigned int reg, u16 val)
+{
+	struct fw_ldst_cmd c;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_addrspace = htonl(FW_CMD_OP(FW_LDST_CMD) | FW_CMD_REQUEST |
+		FW_CMD_WRITE | FW_LDST_CMD_ADDRSPACE(FW_LDST_ADDRSPC_MDIO));
+	c.cycles_to_len16 = htonl(FW_LEN16(c));
+	c.u.mdio.paddr_mmd = htons(FW_LDST_CMD_PADDR(phy_addr) |
+				   FW_LDST_CMD_MMD(mmd));
+	c.u.mdio.raddr = htons(reg);
+	c.u.mdio.rval = htons(val);
+
+	return t4_wr_mbox(adap, mbox, &c, sizeof(c), NULL);
+}
+
+/**
+ *	t4_fw_hello - establish communication with FW
+ *	@adap: the adapter
+ *	@mbox: mailbox to use for the FW command
+ *	@evt_mbox: mailbox to receive async FW events
+ *	@master: specifies the caller's willingness to be the device master
+ *	@state: returns the current device state
+ *
+ *	Issues a command to establish communication with FW.
+ */
+int t4_fw_hello(struct adapter *adap, unsigned int mbox, unsigned int evt_mbox,
+		enum dev_master master, enum dev_state *state)
+{
+	int ret;
+	struct fw_hello_cmd c;
+
+	INIT_CMD(c, HELLO, WRITE);
+	c.err_to_mbasyncnot = htonl(
+		FW_HELLO_CMD_MASTERDIS(master == MASTER_CANT) |
+		FW_HELLO_CMD_MASTERFORCE(master == MASTER_MUST) |
+		FW_HELLO_CMD_MBMASTER(master == MASTER_MUST ? mbox : 0xff) |
+		FW_HELLO_CMD_MBASYNCNOT(evt_mbox));
+
+	ret = t4_wr_mbox(adap, mbox, &c, sizeof(c), &c);
+	if (ret == 0 && state) {
+		u32 v = ntohl(c.err_to_mbasyncnot);
+		if (v & FW_HELLO_CMD_INIT)
+			*state = DEV_STATE_INIT;
+		else if (v & FW_HELLO_CMD_ERR)
+			*state = DEV_STATE_ERR;
+		else
+			*state = DEV_STATE_UNINIT;
+	}
+	return ret;
+}
+
+/**
+ *	t4_fw_bye - end communication with FW
+ *	@adap: the adapter
+ *	@mbox: mailbox to use for the FW command
+ *
+ *	Issues a command to terminate communication with FW.
+ */
+int t4_fw_bye(struct adapter *adap, unsigned int mbox)
+{
+	struct fw_bye_cmd c;
+
+	INIT_CMD(c, BYE, WRITE);
+	return t4_wr_mbox(adap, mbox, &c, sizeof(c), NULL);
+}
+
+/**
+ *	t4_init_cmd - ask FW to initialize the device
+ *	@adap: the adapter
+ *	@mbox: mailbox to use for the FW command
+ *
+ *	Issues a command to FW to partially initialize the device.  This
+ *	performs initialization that generally doesn't depend on user input.
+ */
+int t4_early_init(struct adapter *adap, unsigned int mbox)
+{
+	struct fw_initialize_cmd c;
+
+	INIT_CMD(c, INITIALIZE, WRITE);
+	return t4_wr_mbox(adap, mbox, &c, sizeof(c), NULL);
+}
+
+/**
+ *	t4_fw_reset - issue a reset to FW
+ *	@adap: the adapter
+ *	@mbox: mailbox to use for the FW command
+ *	@reset: specifies the type of reset to perform
+ *
+ *	Issues a reset command of the specified type to FW.
+ */
+int t4_fw_reset(struct adapter *adap, unsigned int mbox, int reset)
+{
+	struct fw_reset_cmd c;
+
+	INIT_CMD(c, RESET, WRITE);
+	c.val = htonl(reset);
+	return t4_wr_mbox(adap, mbox, &c, sizeof(c), NULL);
+}
+
+/**
+ *	t4_query_params - query FW or device parameters
+ *	@adap: the adapter
+ *	@mbox: mailbox to use for the FW command
+ *	@pf: the PF
+ *	@vf: the VF
+ *	@nparams: the number of parameters
+ *	@params: the parameter names
+ *	@val: the parameter values
+ *
+ *	Reads the value of FW or device parameters.  Up to 7 parameters can be
+ *	queried at once.
+ */
+int t4_query_params(struct adapter *adap, unsigned int mbox, unsigned int pf,
+		    unsigned int vf, unsigned int nparams, const u32 *params,
+		    u32 *val)
+{
+	int i, ret;
+	struct fw_params_cmd c;
+	__be32 *p = &c.param[0].mnem;
+
+	if (nparams > 7)
+		return -EINVAL;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_vfn = htonl(FW_CMD_OP(FW_PARAMS_CMD) | FW_CMD_REQUEST |
+			    FW_CMD_READ | FW_PARAMS_CMD_PFN(pf) |
+			    FW_PARAMS_CMD_VFN(vf));
+	c.retval_len16 = htonl(FW_LEN16(c));
+	for (i = 0; i < nparams; i++, p += 2)
+		*p = htonl(*params++);
+
+	ret = t4_wr_mbox(adap, mbox, &c, sizeof(c), &c);
+	if (ret == 0)
+		for (i = 0, p = &c.param[0].val; i < nparams; i++, p += 2)
+			*val++ = ntohl(*p);
+	return ret;
+}
+
+/**
+ *	t4_set_params - sets FW or device parameters
+ *	@adap: the adapter
+ *	@mbox: mailbox to use for the FW command
+ *	@pf: the PF
+ *	@vf: the VF
+ *	@nparams: the number of parameters
+ *	@params: the parameter names
+ *	@val: the parameter values
+ *
+ *	Sets the value of FW or device parameters.  Up to 7 parameters can be
+ *	specified at once.
+ */
+int t4_set_params(struct adapter *adap, unsigned int mbox, unsigned int pf,
+		  unsigned int vf, unsigned int nparams, const u32 *params,
+		  const u32 *val)
+{
+	struct fw_params_cmd c;
+	__be32 *p = &c.param[0].mnem;
+
+	if (nparams > 7)
+		return -EINVAL;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_vfn = htonl(FW_CMD_OP(FW_PARAMS_CMD) | FW_CMD_REQUEST |
+			    FW_CMD_WRITE | FW_PARAMS_CMD_PFN(pf) |
+			    FW_PARAMS_CMD_VFN(vf));
+	c.retval_len16 = htonl(FW_LEN16(c));
+	while (nparams--) {
+		*p++ = htonl(*params++);
+		*p++ = htonl(*val++);
+	}
+
+	return t4_wr_mbox(adap, mbox, &c, sizeof(c), NULL);
+}
+
+/**
+ *	t4_cfg_pfvf - configure PF/VF resource limits
+ *	@adap: the adapter
+ *	@mbox: mailbox to use for the FW command
+ *	@pf: the PF being configured
+ *	@vf: the VF being configured
+ *	@txq: the max number of egress queues
+ *	@txq_eth_ctrl: the max number of egress Ethernet or control queues
+ *	@rxqi: the max number of interrupt-capable ingress queues
+ *	@rxq: the max number of interruptless ingress queues
+ *	@tc: the PCI traffic class
+ *	@vi: the max number of virtual interfaces
+ *	@cmask: the channel access rights mask for the PF/VF
+ *	@pmask: the port access rights mask for the PF/VF
+ *	@nexact: the maximum number of exact MPS filters
+ *	@rcaps: read capabilities
+ *	@wxcaps: write/execute capabilities
+ *
+ *	Configures resource limits and capabilities for a physical or virtual
+ *	function.
+ */
+int t4_cfg_pfvf(struct adapter *adap, unsigned int mbox, unsigned int pf,
+		unsigned int vf, unsigned int txq, unsigned int txq_eth_ctrl,
+		unsigned int rxqi, unsigned int rxq, unsigned int tc,
+		unsigned int vi, unsigned int cmask, unsigned int pmask,
+		unsigned int nexact, unsigned int rcaps, unsigned int wxcaps)
+{
+	struct fw_pfvf_cmd c;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_vfn = htonl(FW_CMD_OP(FW_PFVF_CMD) | FW_CMD_REQUEST |
+			    FW_CMD_WRITE | FW_PFVF_CMD_PFN(pf) |
+			    FW_PFVF_CMD_VFN(vf));
+	c.retval_len16 = htonl(FW_LEN16(c));
+	c.niqflint_niq = htonl(FW_PFVF_CMD_NIQFLINT(rxqi) |
+			       FW_PFVF_CMD_NIQ(rxq));
+	c.cmask_to_neq = htonl(FW_PFVF_CMD_CMASK(cmask) |
+			       FW_PFVF_CMD_PMASK(pmask) |
+			       FW_PFVF_CMD_NEQ(txq));
+	c.tc_to_nexactf = htonl(FW_PFVF_CMD_TC(tc) | FW_PFVF_CMD_NVI(vi) |
+				FW_PFVF_CMD_NEXACTF(nexact));
+	c.r_caps_to_nethctrl = htonl(FW_PFVF_CMD_R_CAPS(rcaps) |
+				     FW_PFVF_CMD_WX_CAPS(wxcaps) |
+				     FW_PFVF_CMD_NETHCTRL(txq_eth_ctrl));
+	return t4_wr_mbox(adap, mbox, &c, sizeof(c), NULL);
+}
+
+/**
+ *	t4_alloc_vi - allocate a virtual interface
+ *	@adap: the adapter
+ *	@mbox: mailbox to use for the FW command
+ *	@port: physical port associated with the VI
+ *	@pf: the PF owning the VI
+ *	@vf: the VF owning the VI
+ *	@nmac: number of MAC addresses needed (1 to 5)
+ *	@mac: the MAC addresses of the VI
+ *	@rss_size: size of RSS table slice associated with this VI
+ *
+ *	Allocates a virtual interface for the given physical port.  If @mac is
+ *	not %NULL it contains the MAC addresses of the VI as assigned by FW.
+ *	@mac should be large enough to hold @nmac Ethernet addresses, they are
+ *	stored consecutively so the space needed is @nmac * 6 bytes.
+ *	Returns a negative error number or the non-negative VI id.
+ */
+int t4_alloc_vi(struct adapter *adap, unsigned int mbox, unsigned int port,
+		unsigned int pf, unsigned int vf, unsigned int nmac, u8 *mac,
+		unsigned int *rss_size)
+{
+	int ret;
+	struct fw_vi_cmd c;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_vfn = htonl(FW_CMD_OP(FW_VI_CMD) | FW_CMD_REQUEST |
+			    FW_CMD_WRITE | FW_CMD_EXEC |
+			    FW_VI_CMD_PFN(pf) | FW_VI_CMD_VFN(vf));
+	c.alloc_to_len16 = htonl(FW_VI_CMD_ALLOC | FW_LEN16(c));
+	c.portid_pkd = FW_VI_CMD_PORTID(port);
+	c.nmac = nmac - 1;
+
+	ret = t4_wr_mbox(adap, mbox, &c, sizeof(c), &c);
+	if (ret)
+		return ret;
+
+	if (mac) {
+		memcpy(mac, c.mac, sizeof(c.mac));
+		switch (nmac) {
+		case 5:
+			memcpy(mac + 24, c.nmac3, sizeof(c.nmac3));
+		case 4:
+			memcpy(mac + 18, c.nmac2, sizeof(c.nmac2));
+		case 3:
+			memcpy(mac + 12, c.nmac1, sizeof(c.nmac1));
+		case 2:
+			memcpy(mac + 6,  c.nmac0, sizeof(c.nmac0));
+		}
+	}
+	if (rss_size)
+		*rss_size = FW_VI_CMD_RSSSIZE_GET(ntohs(c.rsssize_pkd));
+	return ntohs(c.viid_pkd);
+}
+
+/**
+ *	t4_free_vi - free a virtual interface
+ *	@adap: the adapter
+ *	@mbox: mailbox to use for the FW command
+ *	@pf: the PF owning the VI
+ *	@vf: the VF owning the VI
+ *	@viid: virtual interface identifiler
+ *
+ *	Free a previously allocated virtual interface.
+ */
+int t4_free_vi(struct adapter *adap, unsigned int mbox, unsigned int pf,
+	       unsigned int vf, unsigned int viid)
+{
+	struct fw_vi_cmd c;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_vfn = htonl(FW_CMD_OP(FW_VI_CMD) | FW_CMD_REQUEST |
+			    FW_CMD_EXEC | FW_VI_CMD_PFN(pf) |
+			    FW_VI_CMD_VFN(vf));
+	c.alloc_to_len16 = htonl(FW_VI_CMD_FREE | FW_LEN16(c));
+	c.viid_pkd = htons(FW_VI_CMD_VIID(viid));
+	return t4_wr_mbox(adap, mbox, &c, sizeof(c), &c);
+}
+
+/**
+ *	t4_set_rxmode - set Rx properties of a virtual interface
+ *	@adap: the adapter
+ *	@mbox: mailbox to use for the FW command
+ *	@viid: the VI id
+ *	@mtu: the new MTU or -1
+ *	@promisc: 1 to enable promiscuous mode, 0 to disable it, -1 no change
+ *	@all_multi: 1 to enable all-multi mode, 0 to disable it, -1 no change
+ *	@bcast: 1 to enable broadcast Rx, 0 to disable it, -1 no change
+ *	@sleep_ok: if true we may sleep while awaiting command completion
+ *
+ *	Sets Rx properties of a virtual interface.
+ */
+int t4_set_rxmode(struct adapter *adap, unsigned int mbox, unsigned int viid,
+		  int mtu, int promisc, int all_multi, int bcast, bool sleep_ok)
+{
+	struct fw_vi_rxmode_cmd c;
+
+	/* convert to FW values */
+	if (mtu < 0)
+		mtu = FW_RXMODE_MTU_NO_CHG;
+	if (promisc < 0)
+		promisc = FW_VI_RXMODE_CMD_PROMISCEN_MASK;
+	if (all_multi < 0)
+		all_multi = FW_VI_RXMODE_CMD_ALLMULTIEN_MASK;
+	if (bcast < 0)
+		bcast = FW_VI_RXMODE_CMD_BROADCASTEN_MASK;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_viid = htonl(FW_CMD_OP(FW_VI_RXMODE_CMD) | FW_CMD_REQUEST |
+			     FW_CMD_WRITE | FW_VI_RXMODE_CMD_VIID(viid));
+	c.retval_len16 = htonl(FW_LEN16(c));
+	c.mtu_to_broadcasten = htonl(FW_VI_RXMODE_CMD_MTU(mtu) |
+				     FW_VI_RXMODE_CMD_PROMISCEN(promisc) |
+				     FW_VI_RXMODE_CMD_ALLMULTIEN(all_multi) |
+				     FW_VI_RXMODE_CMD_BROADCASTEN(bcast));
+	return t4_wr_mbox_meat(adap, mbox, &c, sizeof(c), NULL, sleep_ok);
+}
+
+/**
+ *	t4_alloc_mac_filt - allocates exact-match filters for MAC addresses
+ *	@adap: the adapter
+ *	@mbox: mailbox to use for the FW command
+ *	@viid: the VI id
+ *	@free: if true any existing filters for this VI id are first removed
+ *	@naddr: the number of MAC addresses to allocate filters for (up to 7)
+ *	@addr: the MAC address(es)
+ *	@idx: where to store the index of each allocated filter
+ *	@hash: pointer to hash address filter bitmap
+ *	@sleep_ok: call is allowed to sleep
+ *
+ *	Allocates an exact-match filter for each of the supplied addresses and
+ *	sets it to the corresponding address.  If @idx is not %NULL it should
+ *	have at least @naddr entries, each of which will be set to the index of
+ *	the filter allocated for the corresponding MAC address.  If a filter
+ *	could not be allocated for an address its index is set to 0xffff.
+ *	If @hash is not %NULL addresses that fail to allocate an exact filter
+ *	are hashed and update the hash filter bitmap pointed at by @hash.
+ *
+ *	Returns a negative error number or the number of filters allocated.
+ */
+int t4_alloc_mac_filt(struct adapter *adap, unsigned int mbox,
+		      unsigned int viid, bool free, unsigned int naddr,
+		      const u8 **addr, u16 *idx, u64 *hash, bool sleep_ok)
+{
+	int i, ret;
+	struct fw_vi_mac_cmd c;
+	struct fw_vi_mac_exact *p;
+
+	if (naddr > 7)
+		return -EINVAL;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_viid = htonl(FW_CMD_OP(FW_VI_MAC_CMD) | FW_CMD_REQUEST |
+			     FW_CMD_WRITE | (free ? FW_CMD_EXEC : 0) |
+			     FW_VI_MAC_CMD_VIID(viid));
+	c.freemacs_to_len16 = htonl(FW_VI_MAC_CMD_FREEMACS(free) |
+				    FW_CMD_LEN16((naddr + 2) / 2));
+
+	for (i = 0, p = c.u.exact; i < naddr; i++, p++) {
+		p->valid_to_idx = htons(FW_VI_MAC_CMD_VALID |
+				      FW_VI_MAC_CMD_IDX(FW_VI_MAC_ADD_MAC));
+		memcpy(p->macaddr, addr[i], sizeof(p->macaddr));
+	}
+
+	ret = t4_wr_mbox_meat(adap, mbox, &c, sizeof(c), &c, sleep_ok);
+	if (ret)
+		return ret;
+
+	for (i = 0, p = c.u.exact; i < naddr; i++, p++) {
+		u16 index = FW_VI_MAC_CMD_IDX_GET(ntohs(p->valid_to_idx));
+
+		if (idx)
+			idx[i] = index >= NEXACT_MAC ? 0xffff : index;
+		if (index < NEXACT_MAC)
+			ret++;
+		else if (hash)
+			*hash |= (1 << hash_mac_addr(addr[i]));
+	}
+	return ret;
+}
+
+/**
+ *	t4_change_mac - modifies the exact-match filter for a MAC address
+ *	@adap: the adapter
+ *	@mbox: mailbox to use for the FW command
+ *	@viid: the VI id
+ *	@idx: index of existing filter for old value of MAC address, or -1
+ *	@addr: the new MAC address value
+ *	@persist: whether a new MAC allocation should be persistent
+ *	@add_smt: if true also add the address to the HW SMT
+ *
+ *	Modifies an exact-match filter and sets it to the new MAC address.
+ *	Note that in general it is not possible to modify the value of a given
+ *	filter so the generic way to modify an address filter is to free the one
+ *	being used by the old address value and allocate a new filter for the
+ *	new address value.  @idx can be -1 if the address is a new addition.
+ *
+ *	Returns a negative error number or the index of the filter with the new
+ *	MAC value.
+ */
+int t4_change_mac(struct adapter *adap, unsigned int mbox, unsigned int viid,
+		  int idx, const u8 *addr, bool persist, bool add_smt)
+{
+	int ret, mode;
+	struct fw_vi_mac_cmd c;
+	struct fw_vi_mac_exact *p = c.u.exact;
+
+	if (idx < 0)                             /* new allocation */
+		idx = persist ? FW_VI_MAC_ADD_PERSIST_MAC : FW_VI_MAC_ADD_MAC;
+	mode = add_smt ? FW_VI_MAC_SMT_AND_MPSTCAM : FW_VI_MAC_MPS_TCAM_ENTRY;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_viid = htonl(FW_CMD_OP(FW_VI_MAC_CMD) | FW_CMD_REQUEST |
+			     FW_CMD_WRITE | FW_VI_MAC_CMD_VIID(viid));
+	c.freemacs_to_len16 = htonl(FW_CMD_LEN16(1));
+	p->valid_to_idx = htons(FW_VI_MAC_CMD_VALID |
+				FW_VI_MAC_CMD_SMAC_RESULT(mode) |
+				FW_VI_MAC_CMD_IDX(idx));
+	memcpy(p->macaddr, addr, sizeof(p->macaddr));
+
+	ret = t4_wr_mbox(adap, mbox, &c, sizeof(c), &c);
+	if (ret == 0) {
+		ret = FW_VI_MAC_CMD_IDX_GET(ntohs(p->valid_to_idx));
+		if (ret >= NEXACT_MAC)
+			ret = -ENOMEM;
+	}
+	return ret;
+}
+
+/**
+ *	t4_set_addr_hash - program the MAC inexact-match hash filter
+ *	@adap: the adapter
+ *	@mbox: mailbox to use for the FW command
+ *	@viid: the VI id
+ *	@ucast: whether the hash filter should also match unicast addresses
+ *	@vec: the value to be written to the hash filter
+ *	@sleep_ok: call is allowed to sleep
+ *
+ *	Sets the 64-bit inexact-match hash filter for a virtual interface.
+ */
+int t4_set_addr_hash(struct adapter *adap, unsigned int mbox, unsigned int viid,
+		     bool ucast, u64 vec, bool sleep_ok)
+{
+	struct fw_vi_mac_cmd c;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_viid = htonl(FW_CMD_OP(FW_VI_MAC_CMD) | FW_CMD_REQUEST |
+			     FW_CMD_WRITE | FW_VI_ENABLE_CMD_VIID(viid));
+	c.freemacs_to_len16 = htonl(FW_VI_MAC_CMD_HASHVECEN |
+				    FW_VI_MAC_CMD_HASHUNIEN(ucast) |
+				    FW_CMD_LEN16(1));
+	c.u.hash.hashvec = cpu_to_be64(vec);
+	return t4_wr_mbox_meat(adap, mbox, &c, sizeof(c), NULL, sleep_ok);
+}
+
+/**
+ *	t4_enable_vi - enable/disable a virtual interface
+ *	@adap: the adapter
+ *	@mbox: mailbox to use for the FW command
+ *	@viid: the VI id
+ *	@rx_en: 1=enable Rx, 0=disable Rx
+ *	@tx_en: 1=enable Tx, 0=disable Tx
+ *
+ *	Enables/disables a virtual interface.
+ */
+int t4_enable_vi(struct adapter *adap, unsigned int mbox, unsigned int viid,
+		 bool rx_en, bool tx_en)
+{
+	struct fw_vi_enable_cmd c;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_viid = htonl(FW_CMD_OP(FW_VI_ENABLE_CMD) | FW_CMD_REQUEST |
+			     FW_CMD_EXEC | FW_VI_ENABLE_CMD_VIID(viid));
+	c.ien_to_len16 = htonl(FW_VI_ENABLE_CMD_IEN(rx_en) |
+			       FW_VI_ENABLE_CMD_EEN(tx_en) | FW_LEN16(c));
+	return t4_wr_mbox(adap, mbox, &c, sizeof(c), NULL);
+}
+
+/**
+ *	t4_identify_port - identify a VI's port by blinking its LED
+ *	@adap: the adapter
+ *	@mbox: mailbox to use for the FW command
+ *	@viid: the VI id
+ *	@nblinks: how many times to blink LED at 2.5 Hz
+ *
+ *	Identifies a VI's port by blinking its LED.
+ */
+int t4_identify_port(struct adapter *adap, unsigned int mbox, unsigned int viid,
+		     unsigned int nblinks)
+{
+	struct fw_vi_enable_cmd c;
+
+	c.op_to_viid = htonl(FW_CMD_OP(FW_VI_ENABLE_CMD) | FW_CMD_REQUEST |
+			     FW_CMD_EXEC | FW_VI_ENABLE_CMD_VIID(viid));
+	c.ien_to_len16 = htonl(FW_VI_ENABLE_CMD_LED | FW_LEN16(c));
+	c.blinkdur = htons(nblinks);
+	return t4_wr_mbox(adap, mbox, &c, sizeof(c), NULL);
+}
+
+/**
+ *	t4_iq_start_stop - enable/disable an ingress queue and its FLs
+ *	@adap: the adapter
+ *	@mbox: mailbox to use for the FW command
+ *	@start: %true to enable the queues, %false to disable them
+ *	@pf: the PF owning the queues
+ *	@vf: the VF owning the queues
+ *	@iqid: ingress queue id
+ *	@fl0id: FL0 queue id or 0xffff if no attached FL0
+ *	@fl1id: FL1 queue id or 0xffff if no attached FL1
+ *
+ *	Starts or stops an ingress queue and its associated FLs, if any.
+ */
+int t4_iq_start_stop(struct adapter *adap, unsigned int mbox, bool start,
+		     unsigned int pf, unsigned int vf, unsigned int iqid,
+		     unsigned int fl0id, unsigned int fl1id)
+{
+	struct fw_iq_cmd c;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_vfn = htonl(FW_CMD_OP(FW_IQ_CMD) | FW_CMD_REQUEST |
+			    FW_CMD_EXEC | FW_IQ_CMD_PFN(pf) |
+			    FW_IQ_CMD_VFN(vf));
+	c.alloc_to_len16 = htonl(FW_IQ_CMD_IQSTART(start) |
+				 FW_IQ_CMD_IQSTOP(!start) | FW_LEN16(c));
+	c.iqid = htons(iqid);
+	c.fl0id = htons(fl0id);
+	c.fl1id = htons(fl1id);
+	return t4_wr_mbox(adap, mbox, &c, sizeof(c), NULL);
+}
+
+/**
+ *	t4_iq_free - free an ingress queue and its FLs
+ *	@adap: the adapter
+ *	@mbox: mailbox to use for the FW command
+ *	@pf: the PF owning the queues
+ *	@vf: the VF owning the queues
+ *	@iqtype: the ingress queue type
+ *	@iqid: ingress queue id
+ *	@fl0id: FL0 queue id or 0xffff if no attached FL0
+ *	@fl1id: FL1 queue id or 0xffff if no attached FL1
+ *
+ *	Frees an ingress queue and its associated FLs, if any.
+ */
+int t4_iq_free(struct adapter *adap, unsigned int mbox, unsigned int pf,
+	       unsigned int vf, unsigned int iqtype, unsigned int iqid,
+	       unsigned int fl0id, unsigned int fl1id)
+{
+	struct fw_iq_cmd c;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_vfn = htonl(FW_CMD_OP(FW_IQ_CMD) | FW_CMD_REQUEST |
+			    FW_CMD_EXEC | FW_IQ_CMD_PFN(pf) |
+			    FW_IQ_CMD_VFN(vf));
+	c.alloc_to_len16 = htonl(FW_IQ_CMD_FREE | FW_LEN16(c));
+	c.type_to_iqandstindex = htonl(FW_IQ_CMD_TYPE(iqtype));
+	c.iqid = htons(iqid);
+	c.fl0id = htons(fl0id);
+	c.fl1id = htons(fl1id);
+	return t4_wr_mbox(adap, mbox, &c, sizeof(c), NULL);
+}
+
+/**
+ *	t4_eth_eq_free - free an Ethernet egress queue
+ *	@adap: the adapter
+ *	@mbox: mailbox to use for the FW command
+ *	@pf: the PF owning the queue
+ *	@vf: the VF owning the queue
+ *	@eqid: egress queue id
+ *
+ *	Frees an Ethernet egress queue.
+ */
+int t4_eth_eq_free(struct adapter *adap, unsigned int mbox, unsigned int pf,
+		   unsigned int vf, unsigned int eqid)
+{
+	struct fw_eq_eth_cmd c;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_vfn = htonl(FW_CMD_OP(FW_EQ_ETH_CMD) | FW_CMD_REQUEST |
+			    FW_CMD_EXEC | FW_EQ_ETH_CMD_PFN(pf) |
+			    FW_EQ_ETH_CMD_VFN(vf));
+	c.alloc_to_len16 = htonl(FW_EQ_ETH_CMD_FREE | FW_LEN16(c));
+	c.eqid_pkd = htonl(FW_EQ_ETH_CMD_EQID(eqid));
+	return t4_wr_mbox(adap, mbox, &c, sizeof(c), NULL);
+}
+
+/**
+ *	t4_ctrl_eq_free - free a control egress queue
+ *	@adap: the adapter
+ *	@mbox: mailbox to use for the FW command
+ *	@pf: the PF owning the queue
+ *	@vf: the VF owning the queue
+ *	@eqid: egress queue id
+ *
+ *	Frees a control egress queue.
+ */
+int t4_ctrl_eq_free(struct adapter *adap, unsigned int mbox, unsigned int pf,
+		    unsigned int vf, unsigned int eqid)
+{
+	struct fw_eq_ctrl_cmd c;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_vfn = htonl(FW_CMD_OP(FW_EQ_CTRL_CMD) | FW_CMD_REQUEST |
+			    FW_CMD_EXEC | FW_EQ_CTRL_CMD_PFN(pf) |
+			    FW_EQ_CTRL_CMD_VFN(vf));
+	c.alloc_to_len16 = htonl(FW_EQ_CTRL_CMD_FREE | FW_LEN16(c));
+	c.cmpliqid_eqid = htonl(FW_EQ_CTRL_CMD_EQID(eqid));
+	return t4_wr_mbox(adap, mbox, &c, sizeof(c), NULL);
+}
+
+/**
+ *	t4_ofld_eq_free - free an offload egress queue
+ *	@adap: the adapter
+ *	@mbox: mailbox to use for the FW command
+ *	@pf: the PF owning the queue
+ *	@vf: the VF owning the queue
+ *	@eqid: egress queue id
+ *
+ *	Frees a control egress queue.
+ */
+int t4_ofld_eq_free(struct adapter *adap, unsigned int mbox, unsigned int pf,
+		    unsigned int vf, unsigned int eqid)
+{
+	struct fw_eq_ofld_cmd c;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_vfn = htonl(FW_CMD_OP(FW_EQ_OFLD_CMD) | FW_CMD_REQUEST |
+			    FW_CMD_EXEC | FW_EQ_OFLD_CMD_PFN(pf) |
+			    FW_EQ_OFLD_CMD_VFN(vf));
+	c.alloc_to_len16 = htonl(FW_EQ_OFLD_CMD_FREE | FW_LEN16(c));
+	c.eqid_pkd = htonl(FW_EQ_OFLD_CMD_EQID(eqid));
+	return t4_wr_mbox(adap, mbox, &c, sizeof(c), NULL);
+}
+
+/**
+ *	t4_handle_fw_rpl - process a FW reply message
+ *	@adap: the adapter
+ *	@rpl: start of the FW message
+ *
+ *	Processes a FW message, such as link state change messages.
+ */
+int t4_handle_fw_rpl(struct adapter *adap, const __be64 *rpl)
+{
+	u8 opcode = *(const u8 *)rpl;
+
+	if (opcode == FW_PORT_CMD) {    /* link/module state change message */
+		int speed = 0, fc = 0;
+		const struct fw_port_cmd *p = (void *)rpl;
+		int chan = FW_PORT_CMD_PORTID_GET(ntohl(p->op_to_portid));
+		int port = adap->chan_map[chan];
+		struct port_info *pi = adap2pinfo(adap, port);
+		struct link_config *lc = &pi->link_cfg;
+		u32 stat = ntohl(p->u.info.lstatus_to_modtype);
+		int link_ok = (stat & FW_PORT_CMD_LSTATUS) != 0;
+		u32 mod = FW_PORT_CMD_MODTYPE_GET(stat);
+
+		if (stat & FW_PORT_CMD_RXPAUSE)
+			fc |= PAUSE_RX;
+		if (stat & FW_PORT_CMD_TXPAUSE)
+			fc |= PAUSE_TX;
+		if (stat & FW_PORT_CMD_LSPEED(FW_PORT_CAP_SPEED_100M))
+			speed = SPEED_100;
+		else if (stat & FW_PORT_CMD_LSPEED(FW_PORT_CAP_SPEED_1G))
+			speed = SPEED_1000;
+		else if (stat & FW_PORT_CMD_LSPEED(FW_PORT_CAP_SPEED_10G))
+			speed = SPEED_10000;
+
+		if (link_ok != lc->link_ok || speed != lc->speed ||
+		    fc != lc->fc) {                    /* something changed */
+			lc->link_ok = link_ok;
+			lc->speed = speed;
+			lc->fc = fc;
+			t4_os_link_changed(adap, port, link_ok);
+		}
+		if (mod != pi->mod_type) {
+			pi->mod_type = mod;
+			t4_os_portmod_changed(adap, port);
+		}
+	}
[… more lines in the raw message]

Lines of this file are not shown. View the raw message.

drivers/net/cxgb4/t4_hw.h new file +100 -0

100 lines not on this page · raw message

The rest of this diff is not shown. View the raw message.

Signature · 1 line
1.5.4

Part of the diff is not shown here. View the raw message.

#3 lore raw

PATCH 3/6 cxgb4: Add packet queues and packet DMA code

---
 drivers/net/cxgb4/sge.c | 2431 +++++++++++++++++++++++++++++++++++++++++++++++
 1 files changed, 2431 insertions(+), 0 deletions(-)
 create mode 100644 drivers/net/cxgb4/sge.c
drivers/net/cxgb4/sge.c new file +2431 -0
new file mode 100644
index 0000000..14adc58
@@ -0,0 +1,2431 @@
+/*
+ * This file is part of the Chelsio T4 Ethernet driver for Linux.
+ *
+ * Copyright (c) 2003-2010 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/skbuff.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/if_vlan.h>
+#include <linux/ip.h>
+#include <linux/dma-mapping.h>
+#include <linux/jiffies.h>
+#include <net/ipv6.h>
+#include <net/tcp.h>
+#include "cxgb4.h"
+#include "t4_regs.h"
+#include "t4_msg.h"
+#include "t4fw_api.h"
+
+/*
+ * Rx buffer size.  We use largish buffers if possible but settle for single
+ * pages under memory shortage.
+ */
+#if PAGE_SHIFT >= 16
+# define FL_PG_ORDER 0
+#else
+# define FL_PG_ORDER (16 - PAGE_SHIFT)
+#endif
+
+/* RX_PULL_LEN should be <= RX_COPY_THRES */
+#define RX_COPY_THRES    256
+#define RX_PULL_LEN      128
+
+/*
+ * Main body length for sk_buffs used for Rx Ethernet packets with fragments.
+ * Should be >= RX_PULL_LEN but possibly bigger to give pskb_may_pull some room.
+ */
+#define RX_PKT_SKB_LEN   512
+
+/* Ethernet header padding prepended to RX_PKTs */
+#define RX_PKT_PAD 2
+
+/*
+ * Max number of Tx descriptors we clean up at a time.  Should be modest as
+ * freeing skbs isn't cheap and it happens while holding locks.  We just need
+ * to free packets faster than they arrive, we eventually catch up and keep
+ * the amortized cost reasonable.  Must be >= 2 * TXQ_STOP_THRES.
+ */
+#define MAX_TX_RECLAIM 16
+
+/*
+ * Max number of Rx buffers we replenish at a time.  Again keep this modest,
+ * allocating buffers isn't cheap either.
+ */
+#define MAX_RX_REFILL 16U
+
+/*
+ * Period of the Rx queue check timer.  This timer is infrequent as it has
+ * something to do only when the system experiences severe memory shortage.
+ */
+#define RX_QCHECK_PERIOD (HZ / 2)
+
+/*
+ * Period of the Tx queue check timer.
+ */
+#define TX_QCHECK_PERIOD (HZ / 2)
+
+/*
+ * Max number of Tx descriptors to be reclaimed by the Tx timer.
+ */
+#define MAX_TIMER_TX_RECLAIM 100
+
+/*
+ * Timer index used when backing off due to memory shortage.
+ */
+#define NOMEM_TMR_IDX (SGE_NTIMERS - 1)
+
+/*
+ * An FL with <= FL_STARVE_THRES buffers is starving and a periodic timer will
+ * attempt to refill it.
+ */
+#define FL_STARVE_THRES 4
+
+/*
+ * Suspend an Ethernet Tx queue with fewer available descriptors than this.
+ * This is the same as calc_tx_descs() for a TSO packet with
+ * nr_frags == MAX_SKB_FRAGS.
+ */
+#define ETHTXQ_STOP_THRES \
+	(1 + DIV_ROUND_UP((3 * MAX_SKB_FRAGS) / 2 + (MAX_SKB_FRAGS & 1), 8))
+
+/*
+ * Suspension threshold for non-Ethernet Tx queues.  We require enough room
+ * for a full sized WR.
+ */
+#define TXQ_STOP_THRES (SGE_MAX_WR_LEN / sizeof(struct tx_desc))
+
+/*
+ * Max Tx descriptor space we allow for an Ethernet packet to be inlined
+ * into a WR.
+ */
+#define MAX_IMM_TX_PKT_LEN 128
+
+/*
+ * Max size of a WR sent through a control Tx queue.
+ */
+#define MAX_CTRL_WR_LEN SGE_MAX_WR_LEN
+
+enum {
+	/* packet alignment in FL buffers */
+	FL_ALIGN = L1_CACHE_BYTES < 32 ? 32 : L1_CACHE_BYTES,
+	/* egress status entry size */
+	STAT_LEN = L1_CACHE_BYTES > 64 ? 128 : 64
+};
+
+struct tx_sw_desc {                /* SW state per Tx descriptor */
+	struct sk_buff *skb;
+	struct ulptx_sgl *sgl;
+};
+
+struct rx_sw_desc {                /* SW state per Rx descriptor */
+	struct page *page;
+	dma_addr_t dma_addr;
+};
+
+/*
+ * The low bits of rx_sw_desc.dma_addr have special meaning.
+ */
+enum {
+	RX_LARGE_BUF    = 1 << 0, /* buffer is larger than PAGE_SIZE */
+	RX_UNMAPPED_BUF = 1 << 1, /* buffer is not mapped */
+};
+
+static inline dma_addr_t get_buf_addr(const struct rx_sw_desc *d)
+{
+	return d->dma_addr & ~(dma_addr_t)(RX_LARGE_BUF | RX_UNMAPPED_BUF);
+}
+
+static inline bool is_buf_mapped(const struct rx_sw_desc *d)
+{
+	return !(d->dma_addr & RX_UNMAPPED_BUF);
+}
+
+/**
+ *	txq_avail - return the number of available slots in a Tx queue
+ *	@q: the Tx queue
+ *
+ *	Returns the number of descriptors in a Tx queue available to write new
+ *	packets.
+ */
+static inline unsigned int txq_avail(const struct sge_txq *q)
+{
+	return q->size - 1 - q->in_use;
+}
+
+/**
+ *	fl_cap - return the capacity of a free-buffer list
+ *	@fl: the FL
+ *
+ *	Returns the capacity of a free-buffer list.  The capacity is less than
+ *	the size because one descriptor needs to be left unpopulated, otherwise
+ *	HW will think the FL is empty.
+ */
+static inline unsigned int fl_cap(const struct sge_fl *fl)
+{
+	return fl->size - 8;   /* 1 descriptor = 8 buffers */
+}
+
+static inline bool fl_starving(const struct sge_fl *fl)
+{
+	return fl->avail - fl->pend_cred <= FL_STARVE_THRES;
+}
+
+static int map_skb(struct device *dev, const struct sk_buff *skb,
+		   dma_addr_t *addr)
+{
+	const skb_frag_t *fp, *end;
+	const struct skb_shared_info *si;
+
+	*addr = dma_map_single(dev, skb->data, skb_headlen(skb), DMA_TO_DEVICE);
+	if (dma_mapping_error(dev, *addr))
+		goto out_err;
+
+	si = skb_shinfo(skb);
+	end = &si->frags[si->nr_frags];
+
+	for (fp = si->frags; fp < end; fp++) {
+		*++addr = dma_map_page(dev, fp->page, fp->page_offset, fp->size,
+				       DMA_TO_DEVICE);
+		if (dma_mapping_error(dev, *addr))
+			goto unwind;
+	}
+	return 0;
+
+unwind:
+	while (fp-- > si->frags)
+		dma_unmap_page(dev, *--addr, fp->size, DMA_TO_DEVICE);
+
+	dma_unmap_single(dev, addr[-1], skb_headlen(skb), DMA_TO_DEVICE);
+out_err:
+	return -ENOMEM;
+}
+
+#ifdef CONFIG_NEED_DMA_MAP_STATE
+static void unmap_skb(struct device *dev, const struct sk_buff *skb,
+		      const dma_addr_t *addr)
+{
+	const skb_frag_t *fp, *end;
+	const struct skb_shared_info *si;
+
+	dma_unmap_single(dev, *addr++, skb_headlen(skb), DMA_TO_DEVICE);
+
+	si = skb_shinfo(skb);
+	end = &si->frags[si->nr_frags];
+	for (fp = si->frags; fp < end; fp++)
+		dma_unmap_page(dev, *addr++, fp->size, DMA_TO_DEVICE);
+}
+
+/**
+ *	deferred_unmap_destructor - unmap a packet when it is freed
+ *	@skb: the packet
+ *
+ *	This is the packet destructor used for Tx packets that need to remain
+ *	mapped until they are freed rather than until their Tx descriptors are
+ *	freed.
+ */
+static void deferred_unmap_destructor(struct sk_buff *skb)
+{
+	unmap_skb(skb->dev->dev.parent, skb, (dma_addr_t *)skb->head);
+}
+#endif
+
+static void unmap_sgl(struct device *dev, const struct sk_buff *skb,
+		      const struct ulptx_sgl *sgl, const struct sge_txq *q)
+{
+	const struct ulptx_sge_pair *p;
+	unsigned int nfrags = skb_shinfo(skb)->nr_frags;
+
+	if (likely(skb_headlen(skb)))
+		dma_unmap_single(dev, be64_to_cpu(sgl->addr0), ntohl(sgl->len0),
+				 DMA_TO_DEVICE);
+	else {
+		dma_unmap_page(dev, be64_to_cpu(sgl->addr0), ntohl(sgl->len0),
+			       DMA_TO_DEVICE);
+		nfrags--;
+	}
+
+	/*
+	 * the complexity below is because of the possibility of a wrap-around
+	 * in the middle of an SGL
+	 */
+	for (p = sgl->sge; nfrags >= 2; nfrags -= 2) {
+		if (likely((u8 *)(p + 1) <= (u8 *)q->stat)) {
+unmap:			dma_unmap_page(dev, be64_to_cpu(p->addr[0]),
+				       ntohl(p->len[0]), DMA_TO_DEVICE);
+			dma_unmap_page(dev, be64_to_cpu(p->addr[1]),
+				       ntohl(p->len[1]), DMA_TO_DEVICE);
+			p++;
+		} else if ((u8 *)p == (u8 *)q->stat) {
+			p = (const struct ulptx_sge_pair *)q->desc;
+			goto unmap;
+		} else if ((u8 *)p + 8 == (u8 *)q->stat) {
+			const __be64 *addr = (const __be64 *)q->desc;
+
+			dma_unmap_page(dev, be64_to_cpu(addr[0]),
+				       ntohl(p->len[0]), DMA_TO_DEVICE);
+			dma_unmap_page(dev, be64_to_cpu(addr[1]),
+				       ntohl(p->len[1]), DMA_TO_DEVICE);
+			p = (const struct ulptx_sge_pair *)&addr[2];
+		} else {
+			const __be64 *addr = (const __be64 *)q->desc;
+
+			dma_unmap_page(dev, be64_to_cpu(p->addr[0]),
+				       ntohl(p->len[0]), DMA_TO_DEVICE);
+			dma_unmap_page(dev, be64_to_cpu(addr[0]),
+				       ntohl(p->len[1]), DMA_TO_DEVICE);
+			p = (const struct ulptx_sge_pair *)&addr[1];
+		}
+	}
+	if (nfrags) {
+		__be64 addr;
+
+		if ((u8 *)p == (u8 *)q->stat)
+			p = (const struct ulptx_sge_pair *)q->desc;
+		addr = (u8 *)p + 16 <= (u8 *)q->stat ? p->addr[0] :
+						       *(const __be64 *)q->desc;
+		dma_unmap_page(dev, be64_to_cpu(addr), ntohl(p->len[0]),
+			       DMA_TO_DEVICE);
+	}
+}
+
+/**
+ *	free_tx_desc - reclaims Tx descriptors and their buffers
+ *	@adapter: the adapter
+ *	@q: the Tx queue to reclaim descriptors from
+ *	@n: the number of descriptors to reclaim
+ *	@unmap: whether the buffers should be unmapped for DMA
+ *
+ *	Reclaims Tx descriptors from an SGE Tx queue and frees the associated
+ *	Tx buffers.  Called with the Tx queue lock held.
+ */
+static void free_tx_desc(struct adapter *adap, struct sge_txq *q,
+			 unsigned int n, bool unmap)
+{
+	struct tx_sw_desc *d;
+	unsigned int cidx = q->cidx;
+	struct device *dev = adap->pdev_dev;
+
+	d = &q->sdesc[cidx];
+	while (n--) {
+		if (d->skb) {                       /* an SGL is present */
+			if (unmap)
+				unmap_sgl(dev, d->skb, d->sgl, q);
+			kfree_skb(d->skb);
+			d->skb = NULL;
+		}
+		++d;
+		if (++cidx == q->size) {
+			cidx = 0;
+			d = q->sdesc;
+		}
+	}
+	q->cidx = cidx;
+}
+
+/*
+ * Return the number of reclaimable descriptors in a Tx queue.
+ */
+static inline int reclaimable(const struct sge_txq *q)
+{
+	int hw_cidx = ntohs(q->stat->cidx);
+	hw_cidx -= q->cidx;
+	return hw_cidx < 0 ? hw_cidx + q->size : hw_cidx;
+}
+
+/**
+ *	reclaim_completed_tx - reclaims completed Tx descriptors
+ *	@adap: the adapter
+ *	@q: the Tx queue to reclaim completed descriptors from
+ *	@unmap: whether the buffers should be unmapped for DMA
+ *
+ *	Reclaims Tx descriptors that the SGE has indicated it has processed,
+ *	and frees the associated buffers if possible.  Called with the Tx
+ *	queue locked.
+ */
+static inline void reclaim_completed_tx(struct adapter *adap, struct sge_txq *q,
+					bool unmap)
+{
+	int avail = reclaimable(q);
+
+	if (avail) {
+		/*
+		 * Limit the amount of clean up work we do at a time to keep
+		 * the Tx lock hold time O(1).
+		 */
+		if (avail > MAX_TX_RECLAIM)
+			avail = MAX_TX_RECLAIM;
+
+		free_tx_desc(adap, q, avail, unmap);
+		q->in_use -= avail;
+	}
+}
+
+static inline int get_buf_size(const struct rx_sw_desc *d)
+{
+#if FL_PG_ORDER > 0
+	return (d->dma_addr & RX_LARGE_BUF) ? (PAGE_SIZE << FL_PG_ORDER) :
+					      PAGE_SIZE;
+#else
+	return PAGE_SIZE;
+#endif
+}
+
+/**
+ *	free_rx_bufs - free the Rx buffers on an SGE free list
+ *	@adap: the adapter
+ *	@q: the SGE free list to free buffers from
+ *	@n: how many buffers to free
+ *
+ *	Release the next @n buffers on an SGE free-buffer Rx queue.   The
+ *	buffers must be made inaccessible to HW before calling this function.
+ */
+static void free_rx_bufs(struct adapter *adap, struct sge_fl *q, int n)
+{
+	while (n--) {
+		struct rx_sw_desc *d = &q->sdesc[q->cidx];
+
+		if (is_buf_mapped(d))
+			dma_unmap_page(adap->pdev_dev, get_buf_addr(d),
+				       get_buf_size(d), PCI_DMA_FROMDEVICE);
+		put_page(d->page);
+		d->page = NULL;
+		if (++q->cidx == q->size)
+			q->cidx = 0;
+		q->avail--;
+	}
+}
+
+/**
+ *	unmap_rx_buf - unmap the current Rx buffer on an SGE free list
+ *	@adap: the adapter
+ *	@q: the SGE free list
+ *
+ *	Unmap the current buffer on an SGE free-buffer Rx queue.   The
+ *	buffer must be made inaccessible to HW before calling this function.
+ *
+ *	This is similar to @free_rx_bufs above but does not free the buffer.
+ *	Do note that the FL still loses any further access to the buffer.
+ */
+static void unmap_rx_buf(struct adapter *adap, struct sge_fl *q)
+{
+	struct rx_sw_desc *d = &q->sdesc[q->cidx];
+
+	if (is_buf_mapped(d))
+		dma_unmap_page(adap->pdev_dev, get_buf_addr(d),
+			       get_buf_size(d), PCI_DMA_FROMDEVICE);
+	d->page = NULL;
+	if (++q->cidx == q->size)
+		q->cidx = 0;
+	q->avail--;
+}
+
+static inline void ring_fl_db(struct adapter *adap, struct sge_fl *q)
+{
+	if (q->pend_cred >= 8) {
+		wmb();
+		t4_write_reg(adap, MYPF_REG(SGE_PF_KDOORBELL), DBPRIO |
+			     QID(q->cntxt_id) | PIDX(q->pend_cred / 8));
+		q->pend_cred &= 7;
+	}
+}
+
+static inline void set_rx_sw_desc(struct rx_sw_desc *sd, struct page *pg,
+				  dma_addr_t mapping)
+{
+	sd->page = pg;
+	sd->dma_addr = mapping;      /* includes size low bits */
+}
+
+/**
+ *	refill_fl - refill an SGE Rx buffer ring
+ *	@adap: the adapter
+ *	@q: the ring to refill
+ *	@n: the number of new buffers to allocate
+ *	@gfp: the gfp flags for the allocations
+ *
+ *	(Re)populate an SGE free-buffer queue with up to @n new packet buffers,
+ *	allocated with the supplied gfp flags.  The caller must assure that
+ *	@n does not exceed the queue's capacity.  If afterwards the queue is
+ *	found critically low mark it as starving in the bitmap of starving FLs.
+ *
+ *	Returns the number of buffers allocated.
+ */
+static unsigned int refill_fl(struct adapter *adap, struct sge_fl *q, int n,
+			      gfp_t gfp)
+{
+	struct page *pg;
+	dma_addr_t mapping;
+	unsigned int cred = q->avail;
+	__be64 *d = &q->desc[q->pidx];
+	struct rx_sw_desc *sd = &q->sdesc[q->pidx];
+
+	gfp |= __GFP_NOWARN;         /* failures are expected */
+
+#if FL_PG_ORDER > 0
+	/*
+	 * Prefer large buffers
+	 */
+	while (n) {
+		pg = alloc_pages(gfp | __GFP_COMP, FL_PG_ORDER);
+		if (unlikely(!pg)) {
+			q->large_alloc_failed++;
+			break;       /* fall back to single pages */
+		}
+
+		mapping = dma_map_page(adap->pdev_dev, pg, 0,
+				       PAGE_SIZE << FL_PG_ORDER,
+				       PCI_DMA_FROMDEVICE);
+		if (unlikely(dma_mapping_error(adap->pdev_dev, mapping))) {
+			__free_pages(pg, FL_PG_ORDER);
+			goto out;   /* do not try small pages for this error */
+		}
+		mapping |= RX_LARGE_BUF;
+		*d++ = cpu_to_be64(mapping);
+
+		set_rx_sw_desc(sd, pg, mapping);
+		sd++;
+
+		q->avail++;
+		if (++q->pidx == q->size) {
+			q->pidx = 0;
+			sd = q->sdesc;
+			d = q->desc;
+		}
+		n--;
+	}
+#endif
+
+	while (n--) {
+		pg = __netdev_alloc_page(adap->port[0], gfp);
+		if (unlikely(!pg)) {
+			q->alloc_failed++;
+			break;
+		}
+
+		mapping = dma_map_page(adap->pdev_dev, pg, 0, PAGE_SIZE,
+				       PCI_DMA_FROMDEVICE);
+		if (unlikely(dma_mapping_error(adap->pdev_dev, mapping))) {
+			netdev_free_page(adap->port[0], pg);
+			goto out;
+		}
+		*d++ = cpu_to_be64(mapping);
+
+		set_rx_sw_desc(sd, pg, mapping);
+		sd++;
+
+		q->avail++;
+		if (++q->pidx == q->size) {
+			q->pidx = 0;
+			sd = q->sdesc;
+			d = q->desc;
+		}
+	}
+
+out:	cred = q->avail - cred;
+	q->pend_cred += cred;
+	ring_fl_db(adap, q);
+
+	if (unlikely(fl_starving(q))) {
+		smp_wmb();
+		set_bit(q->cntxt_id, adap->sge.starving_fl);
+	}
+
+	return cred;
+}
+
+static inline void __refill_fl(struct adapter *adap, struct sge_fl *fl)
+{
+	refill_fl(adap, fl, min(MAX_RX_REFILL, fl_cap(fl) - fl->avail),
+		  GFP_ATOMIC);
+}
+
+/**
+ *	alloc_ring - allocate resources for an SGE descriptor ring
+ *	@dev: the PCI device's core device
+ *	@nelem: the number of descriptors
+ *	@elem_size: the size of each descriptor
+ *	@sw_size: the size of the SW state associated with each ring element
+ *	@phys: the physical address of the allocated ring
+ *	@metadata: address of the array holding the SW state for the ring
+ *	@stat_size: extra space in HW ring for status information
+ *
+ *	Allocates resources for an SGE descriptor ring, such as Tx queues,
+ *	free buffer lists, or response queues.  Each SGE ring requires
+ *	space for its HW descriptors plus, optionally, space for the SW state
+ *	associated with each HW entry (the metadata).  The function returns
+ *	three values: the virtual address for the HW ring (the return value
+ *	of the function), the bus address of the HW ring, and the address
+ *	of the SW ring.
+ */
+static void *alloc_ring(struct device *dev, size_t nelem, size_t elem_size,
+			size_t sw_size, dma_addr_t *phys, void *metadata,
+			size_t stat_size)
+{
+	size_t len = nelem * elem_size + stat_size;
+	void *s = NULL;
+	void *p = dma_alloc_coherent(dev, len, phys, GFP_KERNEL);
+
+	if (!p)
+		return NULL;
+	if (sw_size) {
+		s = kcalloc(nelem, sw_size, GFP_KERNEL);
+
+		if (!s) {
+			dma_free_coherent(dev, len, p, *phys);
+			return NULL;
+		}
+	}
+	if (metadata)
+		*(void **)metadata = s;
+	memset(p, 0, len);
+	return p;
+}
+
+/**
+ *	sgl_len - calculates the size of an SGL of the given capacity
+ *	@n: the number of SGL entries
+ *
+ *	Calculates the number of flits needed for a scatter/gather list that
+ *	can hold the given number of entries.
+ */
+static inline unsigned int sgl_len(unsigned int n)
+{
+	n--;
+	return (3 * n) / 2 + (n & 1) + 2;
+}
+
+/**
+ *	flits_to_desc - returns the num of Tx descriptors for the given flits
+ *	@n: the number of flits
+ *
+ *	Returns the number of Tx descriptors needed for the supplied number
+ *	of flits.
+ */
+static inline unsigned int flits_to_desc(unsigned int n)
+{
+	BUG_ON(n > SGE_MAX_WR_LEN / 8);
+	return DIV_ROUND_UP(n, 8);
+}
+
+/**
+ *	is_eth_imm - can an Ethernet packet be sent as immediate data?
+ *	@skb: the packet
+ *
+ *	Returns whether an Ethernet packet is small enough to fit as
+ *	immediate data.
+ */
+static inline int is_eth_imm(const struct sk_buff *skb)
+{
+	return skb->len <= MAX_IMM_TX_PKT_LEN - sizeof(struct cpl_tx_pkt);
+}
+
+/**
+ *	calc_tx_flits - calculate the number of flits for a packet Tx WR
+ *	@skb: the packet
+ *
+ *	Returns the number of flits needed for a Tx WR for the given Ethernet
+ *	packet, including the needed WR and CPL headers.
+ */
+static inline unsigned int calc_tx_flits(const struct sk_buff *skb)
+{
+	unsigned int flits;
+
+	if (is_eth_imm(skb))
+		return DIV_ROUND_UP(skb->len + sizeof(struct cpl_tx_pkt), 8);
+
+	flits = sgl_len(skb_shinfo(skb)->nr_frags + 1) + 4;
+	if (skb_shinfo(skb)->gso_size)
+		flits += 2;
+	return flits;
+}
+
+/**
+ *	calc_tx_descs - calculate the number of Tx descriptors for a packet
+ *	@skb: the packet
+ *
+ *	Returns the number of Tx descriptors needed for the given Ethernet
+ *	packet, including the needed WR and CPL headers.
+ */
+static inline unsigned int calc_tx_descs(const struct sk_buff *skb)
+{
+	return flits_to_desc(calc_tx_flits(skb));
+}
+
+/**
+ *	write_sgl - populate a scatter/gather list for a packet
+ *	@skb: the packet
+ *	@q: the Tx queue we are writing into
+ *	@sgl: starting location for writing the SGL
+ *	@end: points right after the end of the SGL
+ *	@start: start offset into skb main-body data to include in the SGL
+ *	@addr: the list of bus addresses for the SGL elements
+ *
+ *	Generates a gather list for the buffers that make up a packet.
+ *	The caller must provide adequate space for the SGL that will be written.
+ *	The SGL includes all of the packet's page fragments and the data in its
+ *	main body except for the first @start bytes.  @sgl must be 16-byte
+ *	aligned and within a Tx descriptor with available space.  @end points
+ *	right after the end of the SGL but does not account for any potential
+ *	wrap around, i.e., @end > @sgl.
+ */
+static void write_sgl(const struct sk_buff *skb, struct sge_txq *q,
+		      struct ulptx_sgl *sgl, u64 *end, unsigned int start,
+		      const dma_addr_t *addr)
+{
+	unsigned int i, len;
+	struct ulptx_sge_pair *to;
+	const struct skb_shared_info *si = skb_shinfo(skb);
+	unsigned int nfrags = si->nr_frags;
+	struct ulptx_sge_pair buf[MAX_SKB_FRAGS / 2 + 1];
+
+	len = skb_headlen(skb) - start;
+	if (likely(len)) {
+		sgl->len0 = htonl(len);
+		sgl->addr0 = cpu_to_be64(addr[0] + start);
+		nfrags++;
+	} else {
+		sgl->len0 = htonl(si->frags[0].size);
+		sgl->addr0 = cpu_to_be64(addr[1]);
+	}
+
+	sgl->cmd_nsge = htonl(ULPTX_CMD(ULP_TX_SC_DSGL) | ULPTX_NSGE(nfrags));
+	if (likely(--nfrags == 0))
+		return;
+	/*
+	 * Most of the complexity below deals with the possibility we hit the
+	 * end of the queue in the middle of writing the SGL.  For this case
+	 * only we create the SGL in a temporary buffer and then copy it.
+	 */
+	to = (u8 *)end > (u8 *)q->stat ? buf : sgl->sge;
+
+	for (i = (nfrags != si->nr_frags); nfrags >= 2; nfrags -= 2, to++) {
+		to->len[0] = cpu_to_be32(si->frags[i].size);
+		to->len[1] = cpu_to_be32(si->frags[++i].size);
+		to->addr[0] = cpu_to_be64(addr[i]);
+		to->addr[1] = cpu_to_be64(addr[++i]);
+	}
+	if (nfrags) {
+		to->len[0] = cpu_to_be32(si->frags[i].size);
+		to->len[1] = cpu_to_be32(0);
+		to->addr[0] = cpu_to_be64(addr[i + 1]);
+	}
+	if (unlikely((u8 *)end > (u8 *)q->stat)) {
+		unsigned int part0 = (u8 *)q->stat - (u8 *)sgl->sge, part1;
+
+		if (likely(part0))
+			memcpy(sgl->sge, buf, part0);
+		part1 = (u8 *)end - (u8 *)q->stat;
+		memcpy(q->desc, (u8 *)buf + part0, part1);
+		end = (void *)q->desc + part1;
+	}
+	if ((uintptr_t)end & 8)           /* 0-pad to multiple of 16 */
+		*(u64 *)end = 0;
+}
+
+/**
+ *	ring_tx_db - check and potentially ring a Tx queue's doorbell
+ *	@adap: the adapter
+ *	@q: the Tx queue
+ *	@n: number of new descriptors to give to HW
+ *
+ *	Ring the doorbel for a Tx queue.
+ */
+static inline void ring_tx_db(struct adapter *adap, struct sge_txq *q, int n)
+{
+	wmb();            /* write descriptors before telling HW */
+	t4_write_reg(adap, MYPF_REG(SGE_PF_KDOORBELL),
+		     QID(q->cntxt_id) | PIDX(n));
+}
+
+/**
+ *	inline_tx_skb - inline a packet's data into Tx descriptors
+ *	@skb: the packet
+ *	@q: the Tx queue where the packet will be inlined
+ *	@pos: starting position in the Tx queue where to inline the packet
+ *
+ *	Inline a packet's contents directly into Tx descriptors, starting at
+ *	the given position within the Tx DMA ring.
+ *	Most of the complexity of this operation is dealing with wrap arounds
+ *	in the middle of the packet we want to inline.
+ */
+static void inline_tx_skb(const struct sk_buff *skb, const struct sge_txq *q,
+			  void *pos)
+{
+	u64 *p;
+	int left = (void *)q->stat - pos;
+
+	if (likely(skb->len <= left)) {
+		if (likely(!skb->data_len))
+			skb_copy_from_linear_data(skb, pos, skb->len);
+		else
+			skb_copy_bits(skb, 0, pos, skb->len);
+		pos += skb->len;
+	} else {
+		skb_copy_bits(skb, 0, pos, left);
+		skb_copy_bits(skb, left, q->desc, skb->len - left);
+		pos = (void *)q->desc + (skb->len - left);
+	}
+
+	/* 0-pad to multiple of 16 */
+	p = PTR_ALIGN(pos, 8);
+	if ((uintptr_t)p & 8)
+		*p = 0;
+}
+
+/*
+ * Figure out what HW csum a packet wants and return the appropriate control
+ * bits.
+ */
+static u64 hwcsum(const struct sk_buff *skb)
+{
+	int csum_type;
+	const struct iphdr *iph = ip_hdr(skb);
+
+	if (iph->version == 4) {
+		if (iph->protocol == IPPROTO_TCP)
+			csum_type = TX_CSUM_TCPIP;
+		else if (iph->protocol == IPPROTO_UDP)
+			csum_type = TX_CSUM_UDPIP;
+		else {
+nocsum:			/*
+			 * unknown protocol, disable HW csum
+			 * and hope a bad packet is detected
+			 */
+			return TXPKT_L4CSUM_DIS;
+		}
+	} else {
+		/*
+		 * this doesn't work with extension headers
+		 */
+		const struct ipv6hdr *ip6h = (const struct ipv6hdr *)iph;
+
+		if (ip6h->nexthdr == IPPROTO_TCP)
+			csum_type = TX_CSUM_TCPIP6;
+		else if (ip6h->nexthdr == IPPROTO_UDP)
+			csum_type = TX_CSUM_UDPIP6;
+		else
+			goto nocsum;
+	}
+
+	if (likely(csum_type >= TX_CSUM_TCPIP))
+		return TXPKT_CSUM_TYPE(csum_type) |
+			TXPKT_IPHDR_LEN(skb_network_header_len(skb)) |
+			TXPKT_ETHHDR_LEN(skb_network_offset(skb) - ETH_HLEN);
+	else {
+		int start = skb_transport_offset(skb);
+
+		return TXPKT_CSUM_TYPE(csum_type) | TXPKT_CSUM_START(start) |
+			TXPKT_CSUM_LOC(start + skb->csum_offset);
+	}
+}
+
+static void eth_txq_stop(struct sge_eth_txq *q)
+{
+	netif_tx_stop_queue(q->txq);
+	q->q.stops++;
+}
+
+static inline void txq_advance(struct sge_txq *q, unsigned int n)
+{
+	q->in_use += n;
+	q->pidx += n;
+	if (q->pidx >= q->size)
+		q->pidx -= q->size;
+}
+
+/**
+ *	t4_eth_xmit - add a packet to an Ethernet Tx queue
+ *	@skb: the packet
+ *	@dev: the egress net device
+ *
+ *	Add a packet to an SGE Ethernet Tx queue.  Runs with softirqs disabled.
+ */
+netdev_tx_t t4_eth_xmit(struct sk_buff *skb, struct net_device *dev)
+{
+	u32 wr_mid;
+	u64 cntrl, *end;
+	int qidx, credits;
+	unsigned int flits, ndesc;
+	struct adapter *adap;
+	struct sge_eth_txq *q;
+	const struct port_info *pi;
+	struct fw_eth_tx_pkt_wr *wr;
+	struct cpl_tx_pkt_core *cpl;
+	const struct skb_shared_info *ssi;
+	dma_addr_t addr[MAX_SKB_FRAGS + 1];
+
+	/*
+	 * The chip min packet length is 10 octets but play safe and reject
+	 * anything shorter than an Ethernet header.
+	 */
+	if (unlikely(skb->len < ETH_HLEN)) {
+out_free:	dev_kfree_skb(skb);
+		return NETDEV_TX_OK;
+	}
+
+	pi = netdev_priv(dev);
+	adap = pi->adapter;
+	qidx = skb_get_queue_mapping(skb);
+	q = &adap->sge.ethtxq[qidx + pi->first_qset];
+
+	reclaim_completed_tx(adap, &q->q, true);
+
+	flits = calc_tx_flits(skb);
+	ndesc = flits_to_desc(flits);
+	credits = txq_avail(&q->q) - ndesc;
+
+	if (unlikely(credits < 0)) {
+		eth_txq_stop(q);
+		dev_err(adap->pdev_dev,
+			"%s: Tx ring %u full while queue awake!\n",
+			dev->name, qidx);
+		return NETDEV_TX_BUSY;
+	}
+
+	if (!is_eth_imm(skb) &&
+	    unlikely(map_skb(adap->pdev_dev, skb, addr) < 0)) {
+		q->mapping_err++;
+		goto out_free;
+	}
+
+	wr_mid = FW_WR_LEN16(DIV_ROUND_UP(flits, 2));
+	if (unlikely(credits < ETHTXQ_STOP_THRES)) {
+		eth_txq_stop(q);
+		wr_mid |= FW_WR_EQUEQ | FW_WR_EQUIQ;
+	}
+
+	wr = (void *)&q->q.desc[q->q.pidx];
+	wr->equiq_to_len16 = htonl(wr_mid);
+	wr->r3 = cpu_to_be64(0);
+	end = (u64 *)wr + flits;
+
+	ssi = skb_shinfo(skb);
+	if (ssi->gso_size) {
+		struct cpl_tx_pkt_lso *lso = (void *)wr;
+		bool v6 = (ssi->gso_type & SKB_GSO_TCPV6) != 0;
+		int l3hdr_len = skb_network_header_len(skb);
+		int eth_xtra_len = skb_network_offset(skb) - ETH_HLEN;
+
+		wr->op_immdlen = htonl(FW_WR_OP(FW_ETH_TX_PKT_WR) |
+				       FW_WR_IMMDLEN(sizeof(*lso)));
+		lso->lso_ctrl = htonl(LSO_OPCODE(CPL_TX_PKT_LSO) |
+				      LSO_FIRST_SLICE | LSO_LAST_SLICE |
+				      LSO_IPV6(v6) |
+				      LSO_ETHHDR_LEN(eth_xtra_len / 4) |
+				      LSO_IPHDR_LEN(l3hdr_len / 4) |
+				      LSO_TCPHDR_LEN(tcp_hdr(skb)->doff));
+		lso->ipid_ofst = htons(0);
+		lso->mss = htons(ssi->gso_size);
+		lso->seqno_offset = htonl(0);
+		lso->len = htonl(skb->len);
+		cpl = (void *)(lso + 1);
+		cntrl = TXPKT_CSUM_TYPE(v6 ? TX_CSUM_TCPIP6 : TX_CSUM_TCPIP) |
+			TXPKT_IPHDR_LEN(l3hdr_len) |
+			TXPKT_ETHHDR_LEN(eth_xtra_len);
+		q->tso++;
+		q->tx_cso += ssi->gso_segs;
+	} else {
+		int len;
+
+		len = is_eth_imm(skb) ? skb->len + sizeof(*cpl) : sizeof(*cpl);
+		wr->op_immdlen = htonl(FW_WR_OP(FW_ETH_TX_PKT_WR) |
+				       FW_WR_IMMDLEN(len));
+		cpl = (void *)(wr + 1);
+		if (skb->ip_summed == CHECKSUM_PARTIAL) {
+			cntrl = hwcsum(skb) | TXPKT_IPCSUM_DIS;
+			q->tx_cso++;
+		} else
+			cntrl = TXPKT_L4CSUM_DIS | TXPKT_IPCSUM_DIS;
+	}
+
+	if (vlan_tx_tag_present(skb)) {
+		q->vlan_ins++;
+		cntrl |= TXPKT_VLAN_VLD | TXPKT_VLAN(vlan_tx_tag_get(skb));
+	}
+
+	cpl->ctrl0 = htonl(TXPKT_OPCODE(CPL_TX_PKT_XT) |
+			   TXPKT_INTF(pi->tx_chan) | TXPKT_PF(0));
+	cpl->pack = htons(0);
+	cpl->len = htons(skb->len);
+	cpl->ctrl1 = cpu_to_be64(cntrl);
+
+	if (is_eth_imm(skb)) {
+		inline_tx_skb(skb, &q->q, cpl + 1);
+		dev_kfree_skb(skb);
+	} else {
+		int last_desc;
+
+		write_sgl(skb, &q->q, (struct ulptx_sgl *)(cpl + 1), end, 0,
+			  addr);
+		skb_orphan(skb);
+
+		last_desc = q->q.pidx + ndesc - 1;
+		if (last_desc >= q->q.size)
+			last_desc -= q->q.size;
+		q->q.sdesc[last_desc].skb = skb;
+		q->q.sdesc[last_desc].sgl = (struct ulptx_sgl *)(cpl + 1);
+	}
+
+	txq_advance(&q->q, ndesc);
+
+	ring_tx_db(adap, &q->q, ndesc);
+	return NETDEV_TX_OK;
+}
+
+/**
+ *	reclaim_completed_tx_imm - reclaim completed control-queue Tx descs
+ *	@q: the SGE control Tx queue
+ *
+ *	This is a variant of reclaim_completed_tx() that is used for Tx queues
+ *	that send only immediate data (presently just the control queues) and
+ *	thus do not have any sk_buffs to release.
+ */
+static inline void reclaim_completed_tx_imm(struct sge_txq *q)
+{
+	int hw_cidx = ntohs(q->stat->cidx);
+	int reclaim = hw_cidx - q->cidx;
+
+	if (reclaim < 0)
+		reclaim += q->size;
+
+	q->in_use -= reclaim;
+	q->cidx = hw_cidx;
+}
+
+/**
+ *	is_imm - check whether a packet can be sent as immediate data
+ *	@skb: the packet
+ *
+ *	Returns true if a packet can be sent as a WR with immediate data.
+ */
+static inline int is_imm(const struct sk_buff *skb)
+{
+	return skb->len <= MAX_CTRL_WR_LEN;
+}
+
+/**
+ *	ctrlq_check_stop - check if a control queue is full and should stop
+ *	@q: the queue
+ *	@wr: most recent WR written to the queue
+ *
+ *	Check if a control queue has become full and should be stopped.
+ *	We clean up control queue descriptors very lazily, only when we are out.
+ *	If the queue is still full after reclaiming any completed descriptors
+ *	we suspend it and have the last WR wake it up.
+ */
+static void ctrlq_check_stop(struct sge_ctrl_txq *q, struct fw_wr_hdr *wr)
+{
+	reclaim_completed_tx_imm(&q->q);
+	if (unlikely(txq_avail(&q->q) < TXQ_STOP_THRES)) {
+		wr->lo |= htonl(FW_WR_EQUEQ | FW_WR_EQUIQ);
+		q->q.stops++;
+		q->full = 1;
+	}
+}
+
+/**
+ *	ctrl_xmit - send a packet through an SGE control Tx queue
+ *	@q: the control queue
+ *	@skb: the packet
+ *
+ *	Send a packet through an SGE control Tx queue.  Packets sent through
+ *	a control queue must fit entirely as immediate data.
+ */
+static int ctrl_xmit(struct sge_ctrl_txq *q, struct sk_buff *skb)
+{
+	unsigned int ndesc;
+	struct fw_wr_hdr *wr;
+
+	if (unlikely(!is_imm(skb))) {
+		WARN_ON(1);
+		dev_kfree_skb(skb);
+		return NET_XMIT_DROP;
+	}
+
+	ndesc = DIV_ROUND_UP(skb->len, sizeof(struct tx_desc));
+	spin_lock(&q->sendq.lock);
+
+	if (unlikely(q->full)) {
+		skb->priority = ndesc;                  /* save for restart */
+		__skb_queue_tail(&q->sendq, skb);
+		spin_unlock(&q->sendq.lock);
+		return NET_XMIT_CN;
+	}
+
+	wr = (struct fw_wr_hdr *)&q->q.desc[q->q.pidx];
+	inline_tx_skb(skb, &q->q, wr);
+
+	txq_advance(&q->q, ndesc);
+	if (unlikely(txq_avail(&q->q) < TXQ_STOP_THRES))
+		ctrlq_check_stop(q, wr);
+
+	ring_tx_db(q->adap, &q->q, ndesc);
+	spin_unlock(&q->sendq.lock);
+
+	kfree_skb(skb);
+	return NET_XMIT_SUCCESS;
+}
+
+/**
+ *	restart_ctrlq - restart a suspended control queue
+ *	@data: the control queue to restart
+ *
+ *	Resumes transmission on a suspended Tx control queue.
+ */
+static void restart_ctrlq(unsigned long data)
+{
+	struct sk_buff *skb;
+	unsigned int written = 0;
+	struct sge_ctrl_txq *q = (struct sge_ctrl_txq *)data;
+
+	spin_lock(&q->sendq.lock);
+	reclaim_completed_tx_imm(&q->q);
+	BUG_ON(txq_avail(&q->q) < TXQ_STOP_THRES);  /* q should be empty */
+
+	while ((skb = __skb_dequeue(&q->sendq)) != NULL) {
+		struct fw_wr_hdr *wr;
+		unsigned int ndesc = skb->priority;     /* previously saved */
+
+		/*
+		 * Write descriptors and free skbs outside the lock to limit
+		 * wait times.  q->full is still set so new skbs will be queued.
+		 */
+		spin_unlock(&q->sendq.lock);
+
+		wr = (struct fw_wr_hdr *)&q->q.desc[q->q.pidx];
+		inline_tx_skb(skb, &q->q, wr);
+		kfree_skb(skb);
+
+		written += ndesc;
+		txq_advance(&q->q, ndesc);
+		if (unlikely(txq_avail(&q->q) < TXQ_STOP_THRES)) {
+			unsigned long old = q->q.stops;
+
+			ctrlq_check_stop(q, wr);
+			if (q->q.stops != old) {          /* suspended anew */
+				spin_lock(&q->sendq.lock);
+				goto ringdb;
+			}
+		}
+		if (written > 16) {
+			ring_tx_db(q->adap, &q->q, written);
+			written = 0;
+		}
+		spin_lock(&q->sendq.lock);
+	}
+	q->full = 0;
+ringdb: if (written)
+		ring_tx_db(q->adap, &q->q, written);
+	spin_unlock(&q->sendq.lock);
+}
+
+/**
+ *	t4_mgmt_tx - send a management message
+ *	@adap: the adapter
+ *	@skb: the packet containing the management message
+ *
+ *	Send a management message through control queue 0.
+ */
+int t4_mgmt_tx(struct adapter *adap, struct sk_buff *skb)
+{
+	int ret;
+
+	local_bh_disable();
+	ret = ctrl_xmit(&adap->sge.ctrlq[0], skb);
+	local_bh_enable();
+	return ret;
+}
+
+/**
+ *	is_ofld_imm - check whether a packet can be sent as immediate data
+ *	@skb: the packet
+ *
+ *	Returns true if a packet can be sent as an offload WR with immediate
+ *	data.  We currently use the same limit as for Ethernet packets.
+ */
+static inline int is_ofld_imm(const struct sk_buff *skb)
+{
+	return skb->len <= MAX_IMM_TX_PKT_LEN;
+}
+
+/**
+ *	calc_tx_flits_ofld - calculate # of flits for an offload packet
+ *	@skb: the packet
+ *
+ *	Returns the number of flits needed for the given offload packet.
+ *	These packets are already fully constructed and no additional headers
+ *	will be added.
+ */
+static inline unsigned int calc_tx_flits_ofld(const struct sk_buff *skb)
+{
+	unsigned int flits, cnt;
+
+	if (is_ofld_imm(skb))
+		return DIV_ROUND_UP(skb->len, 8);
+
+	flits = skb_transport_offset(skb) / 8U;   /* headers */
+	cnt = skb_shinfo(skb)->nr_frags;
+	if (skb->tail != skb->transport_header)
+		cnt++;
+	return flits + sgl_len(cnt);
+}
+
+/**
+ *	txq_stop_maperr - stop a Tx queue due to I/O MMU exhaustion
+ *	@adap: the adapter
+ *	@q: the queue to stop
+ *
+ *	Mark a Tx queue stopped due to I/O MMU exhaustion and resulting
+ *	inability to map packets.  A periodic timer attempts to restart
+ *	queues so marked.
+ */
+static void txq_stop_maperr(struct sge_ofld_txq *q)
+{
+	q->mapping_err++;
+	q->q.stops++;
+	set_bit(q->q.cntxt_id, q->adap->sge.txq_maperr);
+}
+
+/**
+ *	ofldtxq_stop - stop an offload Tx queue that has become full
+ *	@q: the queue to stop
+ *	@skb: the packet causing the queue to become full
+ *
+ *	Stops an offload Tx queue that has become full and modifies the packet
+ *	being written to request a wakeup.
+ */
+static void ofldtxq_stop(struct sge_ofld_txq *q, struct sk_buff *skb)
+{
+	struct fw_wr_hdr *wr = (struct fw_wr_hdr *)skb->data;
+
+	wr->lo |= htonl(FW_WR_EQUEQ | FW_WR_EQUIQ);
+	q->q.stops++;
+	q->full = 1;
+}
+
+/**
+ *	service_ofldq - restart a suspended offload queue
+ *	@q: the offload queue
+ *
+ *	Services an offload Tx queue by moving packets from its packet queue
+ *	to the HW Tx ring.  The function starts and ends with the queue locked.
+ */
+static void service_ofldq(struct sge_ofld_txq *q)
+{
+	u64 *pos;
+	int credits;
+	struct sk_buff *skb;
+	unsigned int written = 0;
+	unsigned int flits, ndesc;
+
+	while ((skb = skb_peek(&q->sendq)) != NULL && !q->full) {
+		/*
+		 * We drop the lock but leave skb on sendq, thus retaining
+		 * exclusive access to the state of the queue.
+		 */
+		spin_unlock(&q->sendq.lock);
+
+		reclaim_completed_tx(q->adap, &q->q, false);
+
+		flits = skb->priority;                /* previously saved */
+		ndesc = flits_to_desc(flits);
+		credits = txq_avail(&q->q) - ndesc;
+		BUG_ON(credits < 0);
+		if (unlikely(credits < TXQ_STOP_THRES))
+			ofldtxq_stop(q, skb);
+
+		pos = (u64 *)&q->q.desc[q->q.pidx];
+		if (is_ofld_imm(skb))
+			inline_tx_skb(skb, &q->q, pos);
+		else if (map_skb(q->adap->pdev_dev, skb,
+				 (dma_addr_t *)skb->head)) {
+			txq_stop_maperr(q);
+			spin_lock(&q->sendq.lock);
+			break;
+		} else {
+			int last_desc, hdr_len = skb_transport_offset(skb);
+
+			memcpy(pos, skb->data, hdr_len);
+			write_sgl(skb, &q->q, (void *)pos + hdr_len,
+				  pos + flits, hdr_len,
+				  (dma_addr_t *)skb->head);
+#ifdef CONFIG_NEED_DMA_MAP_STATE
+			skb->dev = q->adap->port[0];
+			skb->destructor = deferred_unmap_destructor;
+#endif
+			last_desc = q->q.pidx + ndesc - 1;
+			if (last_desc >= q->q.size)
+				last_desc -= q->q.size;
+			q->q.sdesc[last_desc].skb = skb;
+		}
+
+		txq_advance(&q->q, ndesc);
+		written += ndesc;
+		if (unlikely(written > 32)) {
+			ring_tx_db(q->adap, &q->q, written);
+			written = 0;
+		}
+
+		spin_lock(&q->sendq.lock);
+		__skb_unlink(skb, &q->sendq);
+		if (is_ofld_imm(skb))
+			kfree_skb(skb);
+	}
+	if (likely(written))
+		ring_tx_db(q->adap, &q->q, written);
+}
+
+/**
+ *	ofld_xmit - send a packet through an offload queue
+ *	@q: the Tx offload queue
+ *	@skb: the packet
+ *
+ *	Send an offload packet through an SGE offload queue.
+ */
+static int ofld_xmit(struct sge_ofld_txq *q, struct sk_buff *skb)
+{
+	skb->priority = calc_tx_flits_ofld(skb);       /* save for restart */
+	spin_lock(&q->sendq.lock);
+	__skb_queue_tail(&q->sendq, skb);
+	if (q->sendq.qlen == 1)
+		service_ofldq(q);
+	spin_unlock(&q->sendq.lock);
+	return NET_XMIT_SUCCESS;
+}
+
+/**
+ *	restart_ofldq - restart a suspended offload queue
+ *	@data: the offload queue to restart
+ *
+ *	Resumes transmission on a suspended Tx offload queue.
+ */
+static void restart_ofldq(unsigned long data)
+{
+	struct sge_ofld_txq *q = (struct sge_ofld_txq *)data;
+
+	spin_lock(&q->sendq.lock);
+	q->full = 0;            /* the queue actually is completely empty now */
+	service_ofldq(q);
+	spin_unlock(&q->sendq.lock);
+}
+
+/**
+ *	skb_txq - return the Tx queue an offload packet should use
+ *	@skb: the packet
+ *
+ *	Returns the Tx queue an offload packet should use as indicated by bits
+ *	1-15 in the packet's queue_mapping.
+ */
+static inline unsigned int skb_txq(const struct sk_buff *skb)
+{
+	return skb->queue_mapping >> 1;
+}
+
+/**
+ *	is_ctrl_pkt - return whether an offload packet is a control packet
+ *	@skb: the packet
+ *
+ *	Returns whether an offload packet should use an OFLD or a CTRL
+ *	Tx queue as indicated by bit 0 in the packet's queue_mapping.
+ */
+static inline unsigned int is_ctrl_pkt(const struct sk_buff *skb)
+{
+	return skb->queue_mapping & 1;
+}
+
+static inline int ofld_send(struct adapter *adap, struct sk_buff *skb)
+{
+	unsigned int idx = skb_txq(skb);
+
+	if (unlikely(is_ctrl_pkt(skb)))
+		return ctrl_xmit(&adap->sge.ctrlq[idx], skb);
+	return ofld_xmit(&adap->sge.ofldtxq[idx], skb);
+}
+
+/**
+ *	t4_ofld_send - send an offload packet
+ *	@adap: the adapter
+ *	@skb: the packet
+ *
+ *	Sends an offload packet.  We use the packet queue_mapping to select the
+ *	appropriate Tx queue as follows: bit 0 indicates whether the packet
+ *	should be sent as regular or control, bits 1-15 select the queue.
+ */
+int t4_ofld_send(struct adapter *adap, struct sk_buff *skb)
+{
+	int ret;
+
+	local_bh_disable();
+	ret = ofld_send(adap, skb);
+	local_bh_enable();
+	return ret;
+}
+
+/**
+ *	cxgb4_ofld_send - send an offload packet
+ *	@dev: the net device
+ *	@skb: the packet
+ *
+ *	Sends an offload packet.  This is an exported version of @t4_ofld_send,
+ *	intended for ULDs.
+ */
+int cxgb4_ofld_send(struct net_device *dev, struct sk_buff *skb)
+{
+	return t4_ofld_send(netdev2adap(dev), skb);
+}
+EXPORT_SYMBOL(cxgb4_ofld_send);
+
+static inline void copy_frags(struct skb_shared_info *ssi,
+			      const struct pkt_gl *gl, unsigned int offset)
+{
+	unsigned int n;
+
+	/* usually there's just one frag */
+	ssi->frags[0].page = gl->frags[0].page;
+	ssi->frags[0].page_offset = gl->frags[0].page_offset + offset;
+	ssi->frags[0].size = gl->frags[0].size - offset;
+	ssi->nr_frags = gl->nfrags;
+	n = gl->nfrags - 1;
+	if (n)
+		memcpy(&ssi->frags[1], &gl->frags[1], n * sizeof(skb_frag_t));
+
+	/* get a reference to the last page, we don't own it */
+	get_page(gl->frags[n].page);
+}
+
+/**
+ *	cxgb4_pktgl_to_skb - build an sk_buff from a packet gather list
+ *	@gl: the gather list
+ *	@skb_len: size of sk_buff main body if it carries fragments
+ *	@pull_len: amount of data to move to the sk_buff's main body
+ *
+ *	Builds an sk_buff from the given packet gather list.  Returns the
+ *	sk_buff or %NULL if sk_buff allocation failed.
+ */
+struct sk_buff *cxgb4_pktgl_to_skb(const struct pkt_gl *gl,
+				   unsigned int skb_len, unsigned int pull_len)
+{
+	struct sk_buff *skb;
+
+	/*
+	 * Below we rely on RX_COPY_THRES being less than the smallest Rx buffer
+	 * size, which is expected since buffers are at least PAGE_SIZEd.
+	 * In this case packets up to RX_COPY_THRES have only one fragment.
+	 */
+	if (gl->tot_len <= RX_COPY_THRES) {
+		skb = dev_alloc_skb(gl->tot_len);
+		if (unlikely(!skb))
+			goto out;
+		__skb_put(skb, gl->tot_len);
+		skb_copy_to_linear_data(skb, gl->va, gl->tot_len);
+	} else {
+		skb = dev_alloc_skb(skb_len);
+		if (unlikely(!skb))
+			goto out;
+		__skb_put(skb, pull_len);
+		skb_copy_to_linear_data(skb, gl->va, pull_len);
+
+		copy_frags(skb_shinfo(skb), gl, pull_len);
+		skb->len = gl->tot_len;
+		skb->data_len = skb->len - pull_len;
+		skb->truesize += skb->data_len;
+	}
+out:	return skb;
+}
+EXPORT_SYMBOL(cxgb4_pktgl_to_skb);
+
+/**
+ *	t4_pktgl_free - free a packet gather list
+ *	@gl: the gather list
+ *
+ *	Releases the pages of a packet gather list.  We do not own the last
+ *	page on the list and do not free it.
+ */
+void t4_pktgl_free(const struct pkt_gl *gl)
+{
+	int n;
+	const skb_frag_t *p;
+
+	for (p = gl->frags, n = gl->nfrags - 1; n--; p++)
+		put_page(p->page);
+}
+
+/*
+ * Process an MPS trace packet.  Give it an unused protocol number so it won't
+ * be delivered to anyone and send it to the stack for capture.
+ */
+static noinline int handle_trace_pkt(struct adapter *adap,
+				     const struct pkt_gl *gl)
+{
+	struct sk_buff *skb;
+	struct cpl_trace_pkt *p;
+
+	skb = cxgb4_pktgl_to_skb(gl, RX_PULL_LEN, RX_PULL_LEN);
+	if (unlikely(!skb)) {
+		t4_pktgl_free(gl);
+		return 0;
+	}
+
+	p = (struct cpl_trace_pkt *)skb->data;
+	__skb_pull(skb, sizeof(*p));
+	skb_reset_mac_header(skb);
+	skb->protocol = htons(0xffff);
+	skb->dev = adap->port[0];
+	netif_receive_skb(skb);
+	return 0;
+}
+
+static void do_gro(struct sge_eth_rxq *rxq, const struct pkt_gl *gl,
+		   const struct cpl_rx_pkt *pkt)
+{
+	int ret;
+	struct sk_buff *skb;
+
+	skb = napi_get_frags(&rxq->rspq.napi);
+	if (unlikely(!skb)) {
+		t4_pktgl_free(gl);
+		rxq->stats.rx_drops++;
+		return;
+	}
+
+	copy_frags(skb_shinfo(skb), gl, RX_PKT_PAD);
+	skb->len = gl->tot_len - RX_PKT_PAD;
+	skb->data_len = skb->len;
+	skb->truesize += skb->data_len;
+	skb->ip_summed = CHECKSUM_UNNECESSARY;
+	skb_record_rx_queue(skb, rxq->rspq.idx);
+
+	if (unlikely(pkt->vlan_ex)) {
+		struct port_info *pi = netdev_priv(rxq->rspq.netdev);
+		struct vlan_group *grp = pi->vlan_grp;
+
+		rxq->stats.vlan_ex++;
+		if (likely(grp)) {
+			ret = vlan_gro_frags(&rxq->rspq.napi, grp,
+					     ntohs(pkt->vlan));
+			goto stats;
+		}
+	}
+	ret = napi_gro_frags(&rxq->rspq.napi);
+stats:	if (ret == GRO_HELD)
+		rxq->stats.lro_pkts++;
+	else if (ret == GRO_MERGED || ret == GRO_MERGED_FREE)
+		rxq->stats.lro_merged++;
+	rxq->stats.pkts++;
+	rxq->stats.rx_cso++;
+}
+
+/**
+ *	t4_ethrx_handler - process an ingress ethernet packet
+ *	@q: the response queue that received the packet
+ *	@rsp: the response queue descriptor holding the RX_PKT message
+ *	@si: the gather list of packet fragments
+ *
+ *	Process an ingress ethernet packet and deliver it to the stack.
+ */
+int t4_ethrx_handler(struct sge_rspq *q, const __be64 *rsp,
+		     const struct pkt_gl *si)
+{
+	bool csum_ok;
+	struct sk_buff *skb;
+	struct port_info *pi;
+	const struct cpl_rx_pkt *pkt;
+	struct sge_eth_rxq *rxq = container_of(q, struct sge_eth_rxq, rspq);
+
+	if (unlikely(*(u8 *)rsp == CPL_TRACE_PKT))
+		return handle_trace_pkt(q->adap, si);
+
+	pkt = (void *)&rsp[1];
+	csum_ok = pkt->csum_calc && !pkt->err_vec;
+	if ((pkt->l2info & htonl(RXF_TCP)) &&
+	    (q->netdev->features & NETIF_F_GRO) && csum_ok && !pkt->ip_frag) {
+		do_gro(rxq, si, pkt);
+		return 0;
+	}
+
+	skb = cxgb4_pktgl_to_skb(si, RX_PKT_SKB_LEN, RX_PULL_LEN);
+	if (unlikely(!skb)) {
+		t4_pktgl_free(si);
+		rxq->stats.rx_drops++;
+		return 0;
+	}
+
+	__skb_pull(skb, RX_PKT_PAD);      /* remove ethernet header padding */
+	skb->protocol = eth_type_trans(skb, q->netdev);
+	skb_record_rx_queue(skb, q->idx);
+	pi = netdev_priv(skb->dev);
+	rxq->stats.pkts++;
+
+	if (csum_ok && (pi->rx_offload & RX_CSO) &&
+	    (pkt->l2info & htonl(RXF_UDP | RXF_TCP))) {
+		if (!pkt->ip_frag)
+			skb->ip_summed = CHECKSUM_UNNECESSARY;
+		else {
+			__sum16 c = (__force __sum16)pkt->csum;
+			skb->csum = csum_unfold(c);
+			skb->ip_summed = CHECKSUM_COMPLETE;
+		}
+		rxq->stats.rx_cso++;
+	} else
+		skb->ip_summed = CHECKSUM_NONE;
+
+	if (unlikely(pkt->vlan_ex)) {
+		struct vlan_group *grp = pi->vlan_grp;
+
+		rxq->stats.vlan_ex++;
+		if (likely(grp))
+			vlan_hwaccel_receive_skb(skb, grp, ntohs(pkt->vlan));
+		else
+			dev_kfree_skb_any(skb);
+	} else
+		netif_receive_skb(skb);
+
+	return 0;
+}
+
+/**
+ *	restore_rx_bufs - put back a packet's Rx buffers
+ *	@si: the packet gather list
+ *	@q: the SGE free list
+ *	@frags: number of FL buffers to restore
+ *
+ *	Puts back on an FL the Rx buffers associated with @si.  The buffers
+ *	have already been unmapped and are left unmapped, we mark them so to
+ *	prevent further unmapping attempts.
+ *
+ *	This function undoes a series of @unmap_rx_buf calls when we find out
+ *	that the current packet can't be processed right away afterall and we
+ *	need to come back to it later.  This is a very rare event and there's
+ *	no effort to make this particularly efficient.
+ */
+static void restore_rx_bufs(const struct pkt_gl *si, struct sge_fl *q,
+			    int frags)
+{
+	struct rx_sw_desc *d;
+
+	while (frags--) {
+		if (q->cidx == 0)
+			q->cidx = q->size - 1;
+		else
+			q->cidx--;
+		d = &q->sdesc[q->cidx];
+		d->page = si->frags[frags].page;
+		d->dma_addr |= RX_UNMAPPED_BUF;
+		q->avail++;
+	}
+}
+
+/**
+ *	is_new_response - check if a response is newly written
+ *	@r: the response descriptor
+ *	@q: the response queue
+ *
+ *	Returns true if a response descriptor contains a yet unprocessed
+ *	response.
+ */
+static inline bool is_new_response(const struct rsp_ctrl *r,
+				   const struct sge_rspq *q)
+{
+	return RSPD_GEN(r->type_gen) == q->gen;
+}
+
+/**
+ *	rspq_next - advance to the next entry in a response queue
+ *	@q: the queue
+ *
+ *	Updates the state of a response queue to advance it to the next entry.
+ */
+static inline void rspq_next(struct sge_rspq *q)
+{
+	q->cur_desc = (void *)q->cur_desc + q->iqe_len;
+	if (unlikely(++q->cidx == q->size)) {
+		q->cidx = 0;
+		q->gen ^= 1;
+		q->cur_desc = q->desc;
+	}
+}
+
+/**
+ *	process_responses - process responses from an SGE response queue
+ *	@q: the ingress queue to process
+ *	@budget: how many responses can be processed in this round
+ *
+ *	Process responses from an SGE response queue up to the supplied budget.
+ *	Responses include received packets as well as control messages from FW
+ *	or HW.
+ *
+ *	Additionally choose the interrupt holdoff time for the next interrupt
+ *	on this queue.  If the system is under memory shortage use a fairly
+ *	long delay to help recovery.
+ */
+static int process_responses(struct sge_rspq *q, int budget)
+{
+	int ret, rsp_type;
+	int budget_left = budget;
+	const struct rsp_ctrl *rc;
+	struct sge_eth_rxq *rxq = container_of(q, struct sge_eth_rxq, rspq);
+
+	while (likely(budget_left)) {
+		rc = (void *)q->cur_desc + (q->iqe_len - sizeof(*rc));
+		if (!is_new_response(rc, q))
+			break;
+
+		rmb();
+		rsp_type = RSPD_TYPE(rc->type_gen);
+		if (likely(rsp_type == RSP_TYPE_FLBUF)) {
+			skb_frag_t *fp;
+			struct pkt_gl si;
+			const struct rx_sw_desc *rsd;
+			u32 len = ntohl(rc->pldbuflen_qid), bufsz, frags;
+
+			if (len & RSPD_NEWBUF) {
+				if (likely(q->offset > 0)) {
+					free_rx_bufs(q->adap, &rxq->fl, 1);
+					q->offset = 0;
+				}
+				len &= RSPD_LEN;
+			}
+			si.tot_len = len;
+
+			/* gather packet fragments */
+			for (frags = 0, fp = si.frags; ; frags++, fp++) {
+				rsd = &rxq->fl.sdesc[rxq->fl.cidx];
+				bufsz = get_buf_size(rsd);
+				fp->page = rsd->page;
+				fp->page_offset = q->offset;
+				fp->size = min(bufsz, len);
+				len -= fp->size;
+				if (!len)
+					break;
+				unmap_rx_buf(q->adap, &rxq->fl);
+			}
+
+			/*
+			 * Last buffer remains mapped so explicitly make it
+			 * coherent for CPU access.
+			 */
+			dma_sync_single_for_cpu(q->adap->pdev_dev,
+						get_buf_addr(rsd),
+						fp->size, DMA_FROM_DEVICE);
+
+			si.va = page_address(si.frags[0].page) +
+				si.frags[0].page_offset;
+			prefetch(si.va);
+
+			si.nfrags = frags + 1;
+			ret = q->handler(q, q->cur_desc, &si);
+			if (likely(ret == 0))
+				q->offset += ALIGN(fp->size, FL_ALIGN);
+			else
+				restore_rx_bufs(&si, &rxq->fl, frags);
+		} else if (likely(rsp_type == RSP_TYPE_CPL)) {
+			ret = q->handler(q, q->cur_desc, NULL);
+		} else {
+			ret = q->handler(q, (const __be64 *)rc, CXGB4_MSG_AN);
+		}
+
+		if (unlikely(ret)) {
+			/* couldn't process descriptor, back off for recovery */
+			q->next_intr_params = QINTR_TIMER_IDX(NOMEM_TMR_IDX);
+			break;
+		}
+
+		rspq_next(q);
+		budget_left--;
+	}
+
+	if (q->offset >= 0 && rxq->fl.size - rxq->fl.avail >= 16)
+		__refill_fl(q->adap, &rxq->fl);
+	return budget - budget_left;
+}
+
+/**
+ *	napi_rx_handler - the NAPI handler for Rx processing
+ *	@napi: the napi instance
+ *	@budget: how many packets we can process in this round
+ *
+ *	Handler for new data events when using NAPI.  This does not need any
+ *	locking or protection from interrupts as data interrupts are off at
+ *	this point and other adapter interrupts do not interfere (the latter
+ *	in not a concern at all with MSI-X as non-data interrupts then have
+ *	a separate handler).
+ */
+static int napi_rx_handler(struct napi_struct *napi, int budget)
+{
+	unsigned int params;
+	struct sge_rspq *q = container_of(napi, struct sge_rspq, napi);
+	int work_done = process_responses(q, budget);
+
+	if (likely(work_done < budget)) {
+		napi_complete(napi);
+		params = q->next_intr_params;
+		q->next_intr_params = q->intr_params;
+	} else
+		params = QINTR_TIMER_IDX(7);
+
+	t4_write_reg(q->adap, MYPF_REG(SGE_PF_GTS), CIDXINC(work_done) |
+		     INGRESSQID((u32)q->cntxt_id) | SEINTARM(params));
+	return work_done;
+}
+
+/*
+ * The MSI-X interrupt handler for an SGE response queue.
+ */
+irqreturn_t t4_sge_intr_msix(int irq, void *cookie)
+{
+	struct sge_rspq *q = cookie;
+
+	napi_schedule(&q->napi);
+	return IRQ_HANDLED;
+}
+
+/*
+ * Process the indirect interrupt entries in the interrupt queue and kick off
+ * NAPI for each queue that has generated an entry.
+ */
+static unsigned int process_intrq(struct adapter *adap)
+{
+	unsigned int credits;
+	const struct rsp_ctrl *rc;
+	struct sge_rspq *q = &adap->sge.intrq;
+
+	spin_lock(&adap->sge.intrq_lock);
+	for (credits = 0; ; credits++) {
+		rc = (void *)q->cur_desc + (q->iqe_len - sizeof(*rc));
+		if (!is_new_response(rc, q))
+			break;
+
+		rmb();
+		if (RSPD_TYPE(rc->type_gen) == RSP_TYPE_INTR) {
+			unsigned int qid = ntohl(rc->pldbuflen_qid);
+
+			napi_schedule(&adap->sge.ingr_map[qid]->napi);
+		}
+
+		rspq_next(q);
+	}
+
+	t4_write_reg(adap, MYPF_REG(SGE_PF_GTS), CIDXINC(credits) |
+		     INGRESSQID(q->cntxt_id) | SEINTARM(q->intr_params));
+	spin_unlock(&adap->sge.intrq_lock);
+	return credits;
+}
+
+/*
+ * The MSI interrupt handler, which handles data events from SGE response queues
+ * as well as error and other async events as they all use the same MSI vector.
+ */
+static irqreturn_t t4_intr_msi(int irq, void *cookie)
+{
+	struct adapter *adap = cookie;
+
+	t4_slow_intr_handler(adap);
+	process_intrq(adap);
+	return IRQ_HANDLED;
+}
+
+/*
+ * Interrupt handler for legacy INTx interrupts.
+ * Handles data events from SGE response queues as well as error and other
+ * async events as they all use the same interrupt line.
+ */
+static irqreturn_t t4_intr_intx(int irq, void *cookie)
+{
+	struct adapter *adap = cookie;
+
+	t4_write_reg(adap, MYPF_REG(PCIE_PF_CLI), 0);
+	if (t4_slow_intr_handler(adap) | process_intrq(adap))
+		return IRQ_HANDLED;
+	return IRQ_NONE;             /* probably shared interrupt */
+}
+
+/**
+ *	t4_intr_handler - select the top-level interrupt handler
+ *	@adap: the adapter
+ *
+ *	Selects the top-level interrupt handler based on the type of interrupts
+ *	(MSI-X, MSI, or INTx).
+ */
+irq_handler_t t4_intr_handler(struct adapter *adap)
+{
+	if (adap->flags & USING_MSIX)
+		return t4_sge_intr_msix;
+	if (adap->flags & USING_MSI)
+		return t4_intr_msi;
+	return t4_intr_intx;
+}
+
+static void sge_rx_timer_cb(unsigned long data)
+{
+	unsigned long m;
+	unsigned int i, cnt[2];
+	struct adapter *adap = (struct adapter *)data;
+	struct sge *s = &adap->sge;
+
+	for (i = 0; i < ARRAY_SIZE(s->starving_fl); i++)
+		for (m = s->starving_fl[i]; m; m &= m - 1) {
+			struct sge_eth_rxq *rxq;
+			unsigned int id = __ffs(m) + i * BITS_PER_LONG;
+			struct sge_fl *fl = s->egr_map[id];
+
+			clear_bit(id, s->starving_fl);
+			smp_mb__after_clear_bit();
+
+			if (fl_starving(fl)) {
+				rxq = container_of(fl, struct sge_eth_rxq, fl);
+				if (napi_reschedule(&rxq->rspq.napi))
+					fl->starving++;
+				else
+					set_bit(id, s->starving_fl);
+			}
+		}
+
+	t4_write_reg(adap, SGE_DEBUG_INDEX, 13);
+	cnt[0] = t4_read_reg(adap, SGE_DEBUG_DATA_HIGH);
+	cnt[1] = t4_read_reg(adap, SGE_DEBUG_DATA_LOW);
+
+	for (i = 0; i < 2; i++)
+		if (cnt[i] >= s->starve_thres) {
+			if (s->idma_state[i] || cnt[i] == 0xffffffff)
+				continue;
+			s->idma_state[i] = 1;
+			t4_write_reg(adap, SGE_DEBUG_INDEX, 11);
+			m = t4_read_reg(adap, SGE_DEBUG_DATA_LOW) >> (i * 16);
+			dev_warn(adap->pdev_dev,
+				 "SGE idma%u starvation detected for "
+				 "queue %lu\n", i, m & 0xffff);
+		} else if (s->idma_state[i])
+			s->idma_state[i] = 0;
+
+	mod_timer(&s->rx_timer, jiffies + RX_QCHECK_PERIOD);
+}
+
+static void sge_tx_timer_cb(unsigned long data)
+{
+	unsigned long m;
+	unsigned int i, budget;
+	struct adapter *adap = (struct adapter *)data;
+	struct sge *s = &adap->sge;
+
+	for (i = 0; i < ARRAY_SIZE(s->txq_maperr); i++)
+		for (m = s->txq_maperr[i]; m; m &= m - 1) {
+			unsigned long id = __ffs(m) + i * BITS_PER_LONG;
+			struct sge_ofld_txq *txq = s->egr_map[id];
+
+			clear_bit(id, s->txq_maperr);
+			tasklet_schedule(&txq->qresume_tsk);
+		}
+
+	budget = MAX_TIMER_TX_RECLAIM;
+	i = s->ethtxq_rover;
+	do {
+		struct sge_eth_txq *q = &s->ethtxq[i];
+
+		if (q->q.in_use &&
+		    time_after_eq(jiffies, q->txq->trans_start + HZ / 100) &&
+		    __netif_tx_trylock(q->txq)) {
+			int avail = reclaimable(&q->q);
+
+			if (avail) {
+				if (avail > budget)
+					avail = budget;
+
+				free_tx_desc(adap, &q->q, avail, true);
+				q->q.in_use -= avail;
+				budget -= avail;
+			}
+			__netif_tx_unlock(q->txq);
+		}
+
+		if (++i >= s->ethqsets)
+			i = 0;
+	} while (budget && i != s->ethtxq_rover);
+	s->ethtxq_rover = i;
+	mod_timer(&s->tx_timer, jiffies + (budget ? TX_QCHECK_PERIOD : 2));
+}
+
+int t4_sge_alloc_rxq(struct adapter *adap, struct sge_rspq *iq, bool fwevtq,
+		     struct net_device *dev, int intr_idx,
+		     struct sge_fl *fl, rspq_handler_t hnd)
+{
+	int ret, flsz = 0;
+	struct fw_iq_cmd c;
+	struct port_info *pi = netdev_priv(dev);
+
+	/* Size needs to be multiple of 16, including status entry. */
+	iq->size = roundup(iq->size, 16);
+
+	iq->desc = alloc_ring(adap->pdev_dev, iq->size, iq->iqe_len, 0,
+			      &iq->phys_addr, NULL, 0);
+	if (!iq->desc)
+		return -ENOMEM;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_vfn = htonl(FW_CMD_OP(FW_IQ_CMD) | FW_CMD_REQUEST |
+			    FW_CMD_WRITE | FW_CMD_EXEC |
+			    FW_IQ_CMD_PFN(0) | FW_IQ_CMD_VFN(0));
+	c.alloc_to_len16 = htonl(FW_IQ_CMD_ALLOC | FW_IQ_CMD_IQSTART(1) |
+				 FW_LEN16(c));
+	c.type_to_iqandstindex = htonl(FW_IQ_CMD_TYPE(FW_IQ_TYPE_FL_INT_CAP) |
+		FW_IQ_CMD_IQASYNCH(fwevtq) | FW_IQ_CMD_VIID(pi->viid) |
+		FW_IQ_CMD_IQANDST(intr_idx < 0) | FW_IQ_CMD_IQANUD(1) |
+		FW_IQ_CMD_IQANDSTINDEX(intr_idx >= 0 ? intr_idx :
+							-intr_idx - 1));
+	c.iqdroprss_to_iqesize = htons(FW_IQ_CMD_IQPCIECH(pi->tx_chan) |
+		FW_IQ_CMD_IQGTSMODE |
+		FW_IQ_CMD_IQINTCNTTHRESH(iq->pktcnt_idx) |
+		FW_IQ_CMD_IQESIZE(ilog2(iq->iqe_len) - 4));
+	c.iqsize = htons(iq->size);
+	c.iqaddr = cpu_to_be64(iq->phys_addr);
+
+	if (fl) {
+		fl->size = roundup(fl->size, 8);
+		fl->desc = alloc_ring(adap->pdev_dev, fl->size, sizeof(__be64),
+				      sizeof(struct rx_sw_desc), &fl->addr,
+				      &fl->sdesc, STAT_LEN);
+		if (!fl->desc)
+			goto fl_nomem;
+
+		flsz = fl->size / 8 + STAT_LEN / sizeof(struct tx_desc);
+		c.iqns_to_fl0congen = htonl(FW_IQ_CMD_FL0PACKEN |
+					    FW_IQ_CMD_FL0PADEN);
+		c.fl0dcaen_to_fl0cidxfthresh = htons(FW_IQ_CMD_FL0FBMIN(2) |
+				FW_IQ_CMD_FL0FBMAX(3));
+		c.fl0size = htons(flsz);
+		c.fl0addr = cpu_to_be64(fl->addr);
+	}
+
+	ret = t4_wr_mbox(adap, 0, &c, sizeof(c), &c);
+	if (ret)
+		goto err;
+
+	netif_napi_add(dev, &iq->napi, napi_rx_handler, 64);
+	iq->cur_desc = iq->desc;
+	iq->cidx = 0;
+	iq->gen = 1;
+	iq->next_intr_params = iq->intr_params;
+	iq->cntxt_id = ntohs(c.iqid);
+	iq->abs_id = ntohs(c.physiqid);
+	iq->size--;                           /* subtract status entry */
+	iq->adap = adap;
+	iq->netdev = dev;
+	iq->handler = hnd;
+
+	/* set offset to -1 to distinguish ingress queues without FL */
+	iq->offset = fl ? 0 : -1;
+
+	adap->sge.ingr_map[iq->cntxt_id] = iq;
+
+	if (fl) {
+		fl->cntxt_id = htons(c.fl0id);
+		fl->avail = fl->pend_cred = 0;
+		fl->pidx = fl->cidx = 0;
+		fl->alloc_failed = fl->large_alloc_failed = fl->starving = 0;
+		adap->sge.egr_map[fl->cntxt_id] = fl;
+		refill_fl(adap, fl, fl_cap(fl), GFP_KERNEL);
+	}
+	return 0;
+
+fl_nomem:
+	ret = -ENOMEM;
+err:
+	if (iq->desc) {
+		dma_free_coherent(adap->pdev_dev, iq->size * iq->iqe_len,
+				  iq->desc, iq->phys_addr);
+		iq->desc = NULL;
+	}
+	if (fl && fl->desc) {
+		kfree(fl->sdesc);
+		fl->sdesc = NULL;
+		dma_free_coherent(adap->pdev_dev, flsz * sizeof(struct tx_desc),
+				  fl->desc, fl->addr);
+		fl->desc = NULL;
+	}
+	return ret;
+}
+
+static void init_txq(struct adapter *adap, struct sge_txq *q, unsigned int id)
+{
+	q->in_use = 0;
+	q->cidx = q->pidx = 0;
+	q->stops = q->restarts = 0;
+	q->stat = (void *)&q->desc[q->size];
+	q->cntxt_id = id;
+	adap->sge.egr_map[id] = q;
+}
+
+int t4_sge_alloc_eth_txq(struct adapter *adap, struct sge_eth_txq *txq,
+			 struct net_device *dev, struct netdev_queue *netdevq,
+			 unsigned int iqid)
+{
+	int ret, nentries;
+	struct fw_eq_eth_cmd c;
+	struct port_info *pi = netdev_priv(dev);
+
+	/* Add status entries */
+	nentries = txq->q.size + STAT_LEN / sizeof(struct tx_desc);
+
+	txq->q.desc = alloc_ring(adap->pdev_dev, txq->q.size,
+			sizeof(struct tx_desc), sizeof(struct tx_sw_desc),
+			&txq->q.phys_addr, &txq->q.sdesc, STAT_LEN);
+	if (!txq->q.desc)
+		return -ENOMEM;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_vfn = htonl(FW_CMD_OP(FW_EQ_ETH_CMD) | FW_CMD_REQUEST |
+			    FW_CMD_WRITE | FW_CMD_EXEC |
+			    FW_EQ_ETH_CMD_PFN(0) | FW_EQ_ETH_CMD_VFN(0));
+	c.alloc_to_len16 = htonl(FW_EQ_ETH_CMD_ALLOC |
+				 FW_EQ_ETH_CMD_EQSTART | FW_LEN16(c));
+	c.viid_pkd = htonl(FW_EQ_ETH_CMD_VIID(pi->viid));
+	c.fetchszm_to_iqid = htonl(FW_EQ_ETH_CMD_HOSTFCMODE(2) |
+				   FW_EQ_ETH_CMD_PCIECHN(pi->tx_chan) |
+				   FW_EQ_ETH_CMD_IQID(iqid));
+	c.dcaen_to_eqsize = htonl(FW_EQ_ETH_CMD_FBMIN(2) |
+				  FW_EQ_ETH_CMD_FBMAX(3) |
+				  FW_EQ_ETH_CMD_CIDXFTHRESH(5) |
+				  FW_EQ_ETH_CMD_EQSIZE(nentries));
+	c.eqaddr = cpu_to_be64(txq->q.phys_addr);
+
+	ret = t4_wr_mbox(adap, 0, &c, sizeof(c), &c);
+	if (ret) {
+		kfree(txq->q.sdesc);
+		txq->q.sdesc = NULL;
+		dma_free_coherent(adap->pdev_dev,
+				  nentries * sizeof(struct tx_desc),
+				  txq->q.desc, txq->q.phys_addr);
+		txq->q.desc = NULL;
+		return ret;
+	}
+
+	init_txq(adap, &txq->q, FW_EQ_ETH_CMD_EQID_GET(ntohl(c.eqid_pkd)));
+	txq->txq = netdevq;
+	txq->tso = txq->tx_cso = txq->vlan_ins = 0;
+	txq->mapping_err = 0;
+	return 0;
+}
+
+int t4_sge_alloc_ctrl_txq(struct adapter *adap, struct sge_ctrl_txq *txq,
+			  struct net_device *dev, unsigned int iqid,
+			  unsigned int cmplqid)
+{
+	int ret, nentries;
+	struct fw_eq_ctrl_cmd c;
+	struct port_info *pi = netdev_priv(dev);
+
+	/* Add status entries */
+	nentries = txq->q.size + STAT_LEN / sizeof(struct tx_desc);
+
+	txq->q.desc = alloc_ring(adap->pdev_dev, nentries,
+				 sizeof(struct tx_desc), 0, &txq->q.phys_addr,
+				 NULL, 0);
+	if (!txq->q.desc)
+		return -ENOMEM;
+
+	c.op_to_vfn = htonl(FW_CMD_OP(FW_EQ_CTRL_CMD) | FW_CMD_REQUEST |
+			    FW_CMD_WRITE | FW_CMD_EXEC |
+			    FW_EQ_CTRL_CMD_PFN(0) | FW_EQ_CTRL_CMD_VFN(0));
+	c.alloc_to_len16 = htonl(FW_EQ_CTRL_CMD_ALLOC |
+				 FW_EQ_CTRL_CMD_EQSTART | FW_LEN16(c));
+	c.cmpliqid_eqid = htonl(FW_EQ_CTRL_CMD_CMPLIQID(cmplqid));
+	c.physeqid_pkd = htonl(0);
+	c.fetchszm_to_iqid = htonl(FW_EQ_CTRL_CMD_HOSTFCMODE(2) |
+				   FW_EQ_CTRL_CMD_PCIECHN(pi->tx_chan) |
+				   FW_EQ_CTRL_CMD_IQID(iqid));
+	c.dcaen_to_eqsize = htonl(FW_EQ_CTRL_CMD_FBMIN(2) |
+				  FW_EQ_CTRL_CMD_FBMAX(3) |
+				  FW_EQ_CTRL_CMD_CIDXFTHRESH(5) |
+				  FW_EQ_CTRL_CMD_EQSIZE(nentries));
+	c.eqaddr = cpu_to_be64(txq->q.phys_addr);
+
+	ret = t4_wr_mbox(adap, 0, &c, sizeof(c), &c);
+	if (ret) {
+		dma_free_coherent(adap->pdev_dev,
+				  nentries * sizeof(struct tx_desc),
+				  txq->q.desc, txq->q.phys_addr);
+		txq->q.desc = NULL;
+		return ret;
+	}
+
+	init_txq(adap, &txq->q, FW_EQ_CTRL_CMD_EQID_GET(ntohl(c.cmpliqid_eqid)));
+	txq->adap = adap;
+	skb_queue_head_init(&txq->sendq);
+	tasklet_init(&txq->qresume_tsk, restart_ctrlq, (unsigned long)txq);
+	txq->full = 0;
+	return 0;
+}
+
+int t4_sge_alloc_ofld_txq(struct adapter *adap, struct sge_ofld_txq *txq,
+			  struct net_device *dev, unsigned int iqid)
+{
+	int ret, nentries;
+	struct fw_eq_ofld_cmd c;
+	struct port_info *pi = netdev_priv(dev);
+
+	/* Add status entries */
+	nentries = txq->q.size + STAT_LEN / sizeof(struct tx_desc);
+
+	txq->q.desc = alloc_ring(adap->pdev_dev, txq->q.size,
+			sizeof(struct tx_desc), sizeof(struct tx_sw_desc),
+			&txq->q.phys_addr, &txq->q.sdesc, STAT_LEN);
+	if (!txq->q.desc)
+		return -ENOMEM;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_vfn = htonl(FW_CMD_OP(FW_EQ_OFLD_CMD) | FW_CMD_REQUEST |
+			    FW_CMD_WRITE | FW_CMD_EXEC |
+			    FW_EQ_OFLD_CMD_PFN(0) | FW_EQ_OFLD_CMD_VFN(0));
+	c.alloc_to_len16 = htonl(FW_EQ_OFLD_CMD_ALLOC |
+				 FW_EQ_OFLD_CMD_EQSTART | FW_LEN16(c));
+	c.fetchszm_to_iqid = htonl(FW_EQ_OFLD_CMD_HOSTFCMODE(2) |
+				   FW_EQ_OFLD_CMD_PCIECHN(pi->tx_chan) |
+				   FW_EQ_OFLD_CMD_IQID(iqid));
+	c.dcaen_to_eqsize = htonl(FW_EQ_OFLD_CMD_FBMIN(2) |
+				  FW_EQ_OFLD_CMD_FBMAX(3) |
+				  FW_EQ_OFLD_CMD_CIDXFTHRESH(5) |
+				  FW_EQ_OFLD_CMD_EQSIZE(nentries));
+	c.eqaddr = cpu_to_be64(txq->q.phys_addr);
+
+	ret = t4_wr_mbox(adap, 0, &c, sizeof(c), &c);
+	if (ret) {
+		kfree(txq->q.sdesc);
+		txq->q.sdesc = NULL;
+		dma_free_coherent(adap->pdev_dev,
+				  nentries * sizeof(struct tx_desc),
+				  txq->q.desc, txq->q.phys_addr);
+		txq->q.desc = NULL;
+		return ret;
+	}
+
+	init_txq(adap, &txq->q, FW_EQ_OFLD_CMD_EQID_GET(ntohl(c.eqid_pkd)));
+	txq->adap = adap;
+	skb_queue_head_init(&txq->sendq);
+	tasklet_init(&txq->qresume_tsk, restart_ofldq, (unsigned long)txq);
+	txq->full = 0;
+	txq->mapping_err = 0;
+	return 0;
+}
+
+static void free_txq(struct adapter *adap, struct sge_txq *q)
+{
+	dma_free_coherent(adap->pdev_dev,
+			  q->size * sizeof(struct tx_desc) + STAT_LEN,
+			  q->desc, q->phys_addr);
+	q->cntxt_id = 0;
+	q->sdesc = NULL;
+	q->desc = NULL;
+}
+
+static void free_rspq_fl(struct adapter *adap, struct sge_rspq *rq,
+			 struct sge_fl *fl)
+{
+	unsigned int fl_id = fl ? fl->cntxt_id : 0xffff;
+
+	adap->sge.ingr_map[rq->cntxt_id] = NULL;
+	t4_iq_free(adap, 0, 0, 0, FW_IQ_TYPE_FL_INT_CAP, rq->cntxt_id, fl_id,
+		   0xffff);
+	dma_free_coherent(adap->pdev_dev, (rq->size + 1) * rq->iqe_len,
+			  rq->desc, rq->phys_addr);
+	netif_napi_del(&rq->napi);
+	rq->netdev = NULL;
+	rq->cntxt_id = rq->abs_id = 0;
+	rq->desc = NULL;
+
+	if (fl) {
+		free_rx_bufs(adap, fl, fl->avail);
+		dma_free_coherent(adap->pdev_dev, fl->size * 8 + STAT_LEN,
+				  fl->desc, fl->addr);
+		kfree(fl->sdesc);
+		fl->sdesc = NULL;
+		fl->cntxt_id = 0;
+		fl->desc = NULL;
+	}
+}
+
+/**
+ *	t4_free_sge_resources - free SGE resources
+ *	@adap: the adapter
+ *
+ *	Frees resources used by the SGE queue sets.
+ */
+void t4_free_sge_resources(struct adapter *adap)
+{
+	int i;
+	struct sge_eth_rxq *eq = adap->sge.ethrxq;
+	struct sge_eth_txq *etq = adap->sge.ethtxq;
+	struct sge_ofld_rxq *oq = adap->sge.ofldrxq;
+
+	/* clean up Ethernet Tx/Rx queues */
+	for (i = 0; i < adap->sge.ethqsets; i++, eq++, etq++) {
+		if (eq->rspq.desc)
+			free_rspq_fl(adap, &eq->rspq, &eq->fl);
+		if (etq->q.desc) {
+			t4_eth_eq_free(adap, 0, 0, 0, etq->q.cntxt_id);
+			free_tx_desc(adap, &etq->q, etq->q.in_use, true);
+			kfree(etq->q.sdesc);
+			free_txq(adap, &etq->q);
+		}
+	}
+
+	/* clean up RDMA and iSCSI Rx queues */
+	for (i = 0; i < adap->sge.ofldqsets; i++, oq++) {
+		if (oq->rspq.desc)
+			free_rspq_fl(adap, &oq->rspq, &oq->fl);
+	}
+	for (i = 0, oq = adap->sge.rdmarxq; i < adap->sge.rdmaqs; i++, oq++) {
+		if (oq->rspq.desc)
+			free_rspq_fl(adap, &oq->rspq, &oq->fl);
+	}
+
+	/* clean up offload Tx queues */
+	for (i = 0; i < ARRAY_SIZE(adap->sge.ofldtxq); i++) {
+		struct sge_ofld_txq *q = &adap->sge.ofldtxq[i];
+
+		if (q->q.desc) {
+			tasklet_kill(&q->qresume_tsk);
+			t4_ofld_eq_free(adap, 0, 0, 0, q->q.cntxt_id);
+			free_tx_desc(adap, &q->q, q->q.in_use, false);
+			kfree(q->q.sdesc);
+			__skb_queue_purge(&q->sendq);
+			free_txq(adap, &q->q);
+		}
+	}
+
+	/* clean up control Tx queues */
+	for (i = 0; i < ARRAY_SIZE(adap->sge.ctrlq); i++) {
+		struct sge_ctrl_txq *cq = &adap->sge.ctrlq[i];
+
+		if (cq->q.desc) {
+			tasklet_kill(&cq->qresume_tsk);
+			t4_ctrl_eq_free(adap, 0, 0, 0, cq->q.cntxt_id);
+			__skb_queue_purge(&cq->sendq);
+			free_txq(adap, &cq->q);
+		}
+	}
+
+	if (adap->sge.fw_evtq.desc)
+		free_rspq_fl(adap, &adap->sge.fw_evtq, NULL);
+
+	if (adap->sge.intrq.desc)
+		free_rspq_fl(adap, &adap->sge.intrq, NULL);
+
+	/* clear the reverse egress queue map */
+	memset(adap->sge.egr_map, 0, sizeof(adap->sge.egr_map));
+}
+
+void t4_sge_start(struct adapter *adap)
+{
+	adap->sge.ethtxq_rover = 0;
+	mod_timer(&adap->sge.rx_timer, jiffies + RX_QCHECK_PERIOD);
+	mod_timer(&adap->sge.tx_timer, jiffies + TX_QCHECK_PERIOD);
+}
+
+/**
+ *	t4_sge_stop - disable SGE operation
+ *	@adap: the adapter
+ *
+ *	Stop tasklets and timers associated with the DMA engine.  Note that
+ *	this is effective only if measures have been taken to disable any HW
+ *	events that may restart them.
+ */
+void t4_sge_stop(struct adapter *adap)
+{
+	int i;
+	struct sge *s = &adap->sge;
+
+	if (in_interrupt())  /* actions below require waiting */
+		return;
+
+	if (s->rx_timer.function)
+		del_timer_sync(&s->rx_timer);
+	if (s->tx_timer.function)
+		del_timer_sync(&s->tx_timer);
+
+	for (i = 0; i < ARRAY_SIZE(s->ofldtxq); i++) {
+		struct sge_ofld_txq *q = &s->ofldtxq[i];
+
+		if (q->q.desc)
+			tasklet_kill(&q->qresume_tsk);
+	}
+	for (i = 0; i < ARRAY_SIZE(s->ctrlq); i++) {
+		struct sge_ctrl_txq *cq = &s->ctrlq[i];
+
+		if (cq->q.desc)
+			tasklet_kill(&cq->qresume_tsk);
+	}
+}
+
+/**
+ *	t4_sge_init - initialize SGE
+ *	@adap: the adapter
+ *
+ *	Performs SGE initialization needed every time after a chip reset.
+ *	We do not initialize any of the queues here, instead the driver
+ *	top-level must request them individually.
+ */
+void t4_sge_init(struct adapter *adap)
+{
+	struct sge *s = &adap->sge;
+	unsigned int fl_align_log = ilog2(FL_ALIGN);
+
+	t4_set_reg_field(adap, SGE_CONTROL, PKTSHIFT_MASK |
+			 INGPADBOUNDARY_MASK | EGRSTATUSPAGESIZE,
+			 INGPADBOUNDARY(fl_align_log - 5) | PKTSHIFT(2) |
+			 RXPKTCPLMODE |
+			 (STAT_LEN == 128 ? EGRSTATUSPAGESIZE : 0));
+	t4_set_reg_field(adap, SGE_HOST_PAGE_SIZE, HOSTPAGESIZEPF0_MASK,
+			 HOSTPAGESIZEPF0(PAGE_SHIFT - 10));
+	t4_write_reg(adap, SGE_FL_BUFFER_SIZE0, PAGE_SIZE);
+#if FL_PG_ORDER > 0
+	t4_write_reg(adap, SGE_FL_BUFFER_SIZE1, PAGE_SIZE << FL_PG_ORDER);
+#endif
+	t4_write_reg(adap, SGE_INGRESS_RX_THRESHOLD,
+		     THRESHOLD_0(s->counter_val[0]) |
+		     THRESHOLD_1(s->counter_val[1]) |
+		     THRESHOLD_2(s->counter_val[2]) |
+		     THRESHOLD_3(s->counter_val[3]));
+	t4_write_reg(adap, SGE_TIMER_VALUE_0_AND_1,
+		     TIMERVALUE0(us_to_core_ticks(adap, s->timer_val[0])) |
+		     TIMERVALUE1(us_to_core_ticks(adap, s->timer_val[1])));
+	t4_write_reg(adap, SGE_TIMER_VALUE_2_AND_3,
+		     TIMERVALUE0(us_to_core_ticks(adap, s->timer_val[2])) |
+		     TIMERVALUE1(us_to_core_ticks(adap, s->timer_val[3])));
+	t4_write_reg(adap, SGE_TIMER_VALUE_4_AND_5,
+		     TIMERVALUE0(us_to_core_ticks(adap, s->timer_val[4])) |
+		     TIMERVALUE1(us_to_core_ticks(adap, s->timer_val[5])));
+	setup_timer(&s->rx_timer, sge_rx_timer_cb, (unsigned long)adap);
+	setup_timer(&s->tx_timer, sge_tx_timer_cb, (unsigned long)adap);
+	s->starve_thres = core_ticks_per_usec(adap) * 1000000;  /* 1 s */
+	s->idma_state[0] = s->idma_state[1] = 0;
+	spin_lock_init(&s->intrq_lock);
+}
Signature · 1 line
1.5.4
#4 lore raw

PATCH 4/6 cxgb4: Add remaining driver headers and L2T management

---
 drivers/net/cxgb4/cxgb4.h     |  743 +++++++++++++++++++++++++++++++++++++++++
 drivers/net/cxgb4/cxgb4_uld.h |  235 +++++++++++++
 drivers/net/cxgb4/l2t.c       |  626 ++++++++++++++++++++++++++++++++++
 drivers/net/cxgb4/l2t.h       |  110 ++++++
 4 files changed, 1714 insertions(+), 0 deletions(-)
 create mode 100644 drivers/net/cxgb4/cxgb4.h
 create mode 100644 drivers/net/cxgb4/cxgb4_uld.h
 create mode 100644 drivers/net/cxgb4/l2t.c
 create mode 100644 drivers/net/cxgb4/l2t.h
drivers/net/cxgb4/cxgb4.h new file +743 -0
new file mode 100644
index 0000000..ca162ef
@@ -0,0 +1,743 @@
+/*
+ * This file is part of the Chelsio T4 Ethernet driver for Linux.
+ *
+ * Copyright (c) 2003-2010 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.
+ */
+
+#ifndef __CXGB4_H__
+#define __CXGB4_H__
+
+#include <linux/bitops.h>
+#include <linux/cache.h>
+#include <linux/interrupt.h>
+#include <linux/list.h>
+#include <linux/netdevice.h>
+#include <linux/pci.h>
+#include <linux/spinlock.h>
+#include <linux/timer.h>
+#include <asm/io.h>
+#include "cxgb4_uld.h"
+#include "t4_hw.h"
+
+#define FW_VERSION_MAJOR 1
+#define FW_VERSION_MINOR 1
+#define FW_VERSION_MICRO 0
+
+enum {
+	MAX_NPORTS = 4,     /* max # of ports */
+	SERNUM_LEN = 16,    /* Serial # length */
+	EC_LEN     = 16,    /* E/C length */
+	ID_LEN     = 16,    /* ID length */
+};
+
+enum {
+	MEM_EDC0,
+	MEM_EDC1,
+	MEM_MC
+};
+
+enum dev_master {
+	MASTER_CANT,
+	MASTER_MAY,
+	MASTER_MUST
+};
+
+enum dev_state {
+	DEV_STATE_UNINIT,
+	DEV_STATE_INIT,
+	DEV_STATE_ERR
+};
+
+enum {
+	PAUSE_RX      = 1 << 0,
+	PAUSE_TX      = 1 << 1,
+	PAUSE_AUTONEG = 1 << 2
+};
+
+struct port_stats {
+	u64 tx_octets;            /* total # of octets in good frames */
+	u64 tx_frames;            /* all good frames */
+	u64 tx_bcast_frames;      /* all broadcast frames */
+	u64 tx_mcast_frames;      /* all multicast frames */
+	u64 tx_ucast_frames;      /* all unicast frames */
+	u64 tx_error_frames;      /* all error frames */
+
+	u64 tx_frames_64;         /* # of Tx frames in a particular range */
+	u64 tx_frames_65_127;
+	u64 tx_frames_128_255;
+	u64 tx_frames_256_511;
+	u64 tx_frames_512_1023;
+	u64 tx_frames_1024_1518;
+	u64 tx_frames_1519_max;
+
+	u64 tx_drop;              /* # of dropped Tx frames */
+	u64 tx_pause;             /* # of transmitted pause frames */
+	u64 tx_ppp0;              /* # of transmitted PPP prio 0 frames */
+	u64 tx_ppp1;              /* # of transmitted PPP prio 1 frames */
+	u64 tx_ppp2;              /* # of transmitted PPP prio 2 frames */
+	u64 tx_ppp3;              /* # of transmitted PPP prio 3 frames */
+	u64 tx_ppp4;              /* # of transmitted PPP prio 4 frames */
+	u64 tx_ppp5;              /* # of transmitted PPP prio 5 frames */
+	u64 tx_ppp6;              /* # of transmitted PPP prio 6 frames */
+	u64 tx_ppp7;              /* # of transmitted PPP prio 7 frames */
+
+	u64 rx_octets;            /* total # of octets in good frames */
+	u64 rx_frames;            /* all good frames */
+	u64 rx_bcast_frames;      /* all broadcast frames */
+	u64 rx_mcast_frames;      /* all multicast frames */
+	u64 rx_ucast_frames;      /* all unicast frames */
+	u64 rx_too_long;          /* # of frames exceeding MTU */
+	u64 rx_jabber;            /* # of jabber frames */
+	u64 rx_fcs_err;           /* # of received frames with bad FCS */
+	u64 rx_len_err;           /* # of received frames with length error */
+	u64 rx_symbol_err;        /* symbol errors */
+	u64 rx_runt;              /* # of short frames */
+
+	u64 rx_frames_64;         /* # of Rx frames in a particular range */
+	u64 rx_frames_65_127;
+	u64 rx_frames_128_255;
+	u64 rx_frames_256_511;
+	u64 rx_frames_512_1023;
+	u64 rx_frames_1024_1518;
+	u64 rx_frames_1519_max;
+
+	u64 rx_pause;             /* # of received pause frames */
+	u64 rx_ppp0;              /* # of received PPP prio 0 frames */
+	u64 rx_ppp1;              /* # of received PPP prio 1 frames */
+	u64 rx_ppp2;              /* # of received PPP prio 2 frames */
+	u64 rx_ppp3;              /* # of received PPP prio 3 frames */
+	u64 rx_ppp4;              /* # of received PPP prio 4 frames */
+	u64 rx_ppp5;              /* # of received PPP prio 5 frames */
+	u64 rx_ppp6;              /* # of received PPP prio 6 frames */
+	u64 rx_ppp7;              /* # of received PPP prio 7 frames */
+
+	u64 rx_ovflow0;           /* drops due to buffer-group 0 overflows */
+	u64 rx_ovflow1;           /* drops due to buffer-group 1 overflows */
+	u64 rx_ovflow2;           /* drops due to buffer-group 2 overflows */
+	u64 rx_ovflow3;           /* drops due to buffer-group 3 overflows */
+	u64 rx_trunc0;            /* buffer-group 0 truncated packets */
+	u64 rx_trunc1;            /* buffer-group 1 truncated packets */
+	u64 rx_trunc2;            /* buffer-group 2 truncated packets */
+	u64 rx_trunc3;            /* buffer-group 3 truncated packets */
+};
+
+struct lb_port_stats {
+	u64 octets;
+	u64 frames;
+	u64 bcast_frames;
+	u64 mcast_frames;
+	u64 ucast_frames;
+	u64 error_frames;
+
+	u64 frames_64;
+	u64 frames_65_127;
+	u64 frames_128_255;
+	u64 frames_256_511;
+	u64 frames_512_1023;
+	u64 frames_1024_1518;
+	u64 frames_1519_max;
+
+	u64 drop;
+
+	u64 ovflow0;
+	u64 ovflow1;
+	u64 ovflow2;
+	u64 ovflow3;
+	u64 trunc0;
+	u64 trunc1;
+	u64 trunc2;
+	u64 trunc3;
+};
+
+struct tp_tcp_stats {
+	u32 tcpOutRsts;
+	u64 tcpInSegs;
+	u64 tcpOutSegs;
+	u64 tcpRetransSegs;
+};
+
+struct tp_err_stats {
+	u32 macInErrs[4];
+	u32 hdrInErrs[4];
+	u32 tcpInErrs[4];
+	u32 tnlCongDrops[4];
+	u32 ofldChanDrops[4];
+	u32 tnlTxDrops[4];
+	u32 ofldVlanDrops[4];
+	u32 tcp6InErrs[4];
+	u32 ofldNoNeigh;
+	u32 ofldCongDefer;
+};
+
+struct tp_params {
+	unsigned int ntxchan;        /* # of Tx channels */
+	unsigned int tre;            /* log2 of core clocks per TP tick */
+};
+
+struct vpd_params {
+	unsigned int cclk;
+	u8 ec[EC_LEN + 1];
+	u8 sn[SERNUM_LEN + 1];
+	u8 id[ID_LEN + 1];
+};
+
+struct pci_params {
+	unsigned char speed;
+	unsigned char width;
+};
+
+struct adapter_params {
+	struct tp_params  tp;
+	struct vpd_params vpd;
+	struct pci_params pci;
+
+	unsigned int fw_vers;
+	unsigned int tp_vers;
+	u8 api_vers[7];
+
+	unsigned short mtus[NMTUS];
+	unsigned short a_wnd[NCCTRL_WIN];
+	unsigned short b_wnd[NCCTRL_WIN];
+
+	unsigned char nports;             /* # of ethernet ports */
+	unsigned char portvec;
+	unsigned char rev;                /* chip revision */
+	unsigned char offload;
+
+	unsigned int ofldq_wr_cred;
+};
+
+struct trace_params {
+	u32 data[TRACE_LEN / 4];
+	u32 mask[TRACE_LEN / 4];
+	unsigned short snap_len;
+	unsigned short min_len;
+	unsigned char skip_ofst;
+	unsigned char skip_len;
+	unsigned char invert;
+	unsigned char port;
+};
+
+struct link_config {
+	unsigned short supported;        /* link capabilities */
+	unsigned short advertising;      /* advertised capabilities */
+	unsigned short requested_speed;  /* speed user has requested */
+	unsigned short speed;            /* actual link speed */
+	unsigned char  requested_fc;     /* flow control user has requested */
+	unsigned char  fc;               /* actual link flow control */
+	unsigned char  autoneg;          /* autonegotiating? */
+	unsigned char  link_ok;          /* link up? */
+};
+
+#define FW_LEN16(fw_struct) FW_CMD_LEN16(sizeof(fw_struct) / 16)
+
+enum {
+	MAX_ETH_QSETS = 32,           /* # of Ethernet Tx/Rx queue sets */
+	MAX_OFLD_QSETS = 16,          /* # of offload Tx/Rx queue sets */
+	MAX_CTRL_QUEUES = NCHAN,      /* # of control Tx queues */
+	MAX_RDMA_QUEUES = NCHAN,      /* # of streaming RDMA Rx queues */
+};
+
+enum {
+	MAX_EGRQ = 128,         /* max # of egress queues, including FLs */
+	MAX_INGQ = 64           /* max # of interrupt-capable ingress queues */
+};
+
+struct adapter;
+struct vlan_group;
+struct sge_rspq;
+
+struct port_info {
+	struct adapter *adapter;
+	struct vlan_group *vlan_grp;
+	u16    viid;
+	s16    xact_addr_filt;        /* index of exact MAC address filter */
+	u16    rss_size;              /* size of VI's RSS table slice */
+	s8     mdio_addr;
+	u8     port_type;
+	u8     mod_type;
+	u8     port_id;
+	u8     tx_chan;
+	u8     lport;                 /* associated offload logical port */
+	u8     rx_offload;            /* CSO, etc */
+	u8     nqsets;                /* # of qsets */
+	u8     first_qset;            /* index of first qset */
+	struct link_config link_cfg;
+};
+
+/* port_info.rx_offload flags */
+enum {
+	RX_CSO = 1 << 0,
+};
+
+struct dentry;
+struct work_struct;
+struct proc_dir_entry;
+
+enum {                                 /* adapter flags */
+	FULL_INIT_DONE     = (1 << 0),
+	USING_MSI          = (1 << 1),
+	USING_MSIX         = (1 << 2),
+	QUEUES_BOUND       = (1 << 3),
+	FW_OK              = (1 << 4),
+};
+
+struct rx_sw_desc;
+
+struct sge_fl {                     /* SGE free-buffer queue state */
+	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 times buffer allocation failed */
+	unsigned long large_alloc_failed;
+	unsigned long starving;
+	/* RO fields */
+	unsigned int cntxt_id;      /* SGE context id 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;            /* bus address of HW ring start */
+};
+
+/* A 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 *q, const __be64 *rsp,
+			      const struct pkt_gl *gl);
+
+struct sge_rspq {                   /* state for an SGE response queue */
+	struct napi_struct napi;
+	const __be64 *cur_desc;     /* current descriptor in queue */
+	unsigned int cidx;          /* consumer index */
+	u8 gen;                     /* current generation bit */
+	u8 intr_params;             /* interrupt holdoff parameters */
+	u8 next_intr_params;        /* holdoff params for next interrupt */
+	u8 pktcnt_idx;              /* interrupt packet threshold */
+	u8 uld;                     /* ULD handling this queue */
+	u8 idx;                     /* queue index within its group */
+	int offset;                 /* offset into current Rx buffer */
+	u16 cntxt_id;               /* SGE context id for the response q */
+	u16 abs_id;                 /* absolute SGE id for the response q */
+	__be64 *desc;               /* address of HW response ring */
+	dma_addr_t phys_addr;       /* physical address of the ring */
+	unsigned int iqe_len;       /* entry size */
+	unsigned int size;          /* capacity of response queue */
+	struct adapter *adap;
+	struct net_device *netdev;  /* associated net device */
+	rspq_handler_t handler;
+};
+
+struct sge_eth_stats {              /* Ethernet queue statistics */
+	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 */
+};
+
+struct sge_eth_rxq {                /* SW Ethernet Rx queue */
+	struct sge_rspq rspq;
+	struct sge_fl fl;
+	struct sge_eth_stats stats;
+} ____cacheline_aligned_in_smp;
+
+struct sge_ofld_stats {             /* offload queue statistics */
+	unsigned long pkts;         /* # of packets */
+	unsigned long imm;          /* # of immediate-data packets */
+	unsigned long an;           /* # of asynchronous notifications */
+	unsigned long nomem;        /* # of responses deferred due to no mem */
+};
+
+struct sge_ofld_rxq {               /* SW offload Rx queue */
+	struct sge_rspq rspq;
+	struct sge_fl fl;
+	struct sge_ofld_stats stats;
+} ____cacheline_aligned_in_smp;
+
+struct tx_desc {
+	__be64 flit[8];
+};
+
+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 q has been stopped */
+	unsigned long restarts;     /* # of queue restarts */
+	unsigned int  cntxt_id;     /* SGE context id 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;    /* physical address of the ring */
+};
+
+struct sge_eth_txq {                /* state for an SGE Ethernet Tx queue */
+	struct sge_txq q;
+	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 */
+} ____cacheline_aligned_in_smp;
+
+struct sge_ofld_txq {               /* state for an SGE offload Tx queue */
+	struct sge_txq q;
+	struct adapter *adap;
+	struct sk_buff_head sendq;  /* list of backpressured packets */
+	struct tasklet_struct qresume_tsk; /* restarts the queue */
+	u8 full;                    /* the Tx ring is full */
+	unsigned long mapping_err;  /* # of I/O MMU packet mapping errors */
+} ____cacheline_aligned_in_smp;
+
+struct sge_ctrl_txq {               /* state for an SGE control Tx queue */
+	struct sge_txq q;
+	struct adapter *adap;
+	struct sk_buff_head sendq;  /* list of backpressured packets */
+	struct tasklet_struct qresume_tsk; /* restarts the queue */
+	u8 full;                    /* the Tx ring is full */
+} ____cacheline_aligned_in_smp;
+
+struct sge {
+	struct sge_eth_txq ethtxq[MAX_ETH_QSETS];
+	struct sge_ofld_txq ofldtxq[MAX_OFLD_QSETS];
+	struct sge_ctrl_txq ctrlq[MAX_CTRL_QUEUES];
+
+	struct sge_eth_rxq ethrxq[MAX_ETH_QSETS];
+	struct sge_ofld_rxq ofldrxq[MAX_OFLD_QSETS];
+	struct sge_ofld_rxq rdmarxq[MAX_RDMA_QUEUES];
+	struct sge_rspq fw_evtq ____cacheline_aligned_in_smp;
+
+	struct sge_rspq intrq ____cacheline_aligned_in_smp;
+	spinlock_t intrq_lock;
+
+	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 ofldqsets;              /* # of active offload queue sets */
+	u16 rdmaqs;                 /* # of available RDMA Rx queues */
+	u16 ofld_rxq[MAX_OFLD_QSETS];
+	u16 rdma_rxq[NCHAN];
+	u16 timer_val[SGE_NTIMERS];
+	u8 counter_val[SGE_NCOUNTERS];
+	unsigned int starve_thres;
+	u8 idma_state[2];
+	void *egr_map[MAX_EGRQ];    /* qid->queue egress queue map */
+	struct sge_rspq *ingr_map[MAX_INGQ]; /* qid->queue ingress queue map */
+	DECLARE_BITMAP(starving_fl, MAX_EGRQ);
+	DECLARE_BITMAP(txq_maperr, MAX_EGRQ);
+	struct timer_list rx_timer; /* refills starving FLs */
+	struct timer_list tx_timer; /* checks Tx queues */
+};
+
+#define for_each_ethrxq(sge, i) for (i = 0; i < (sge)->ethqsets; i++)
+#define for_each_ofldrxq(sge, i) for (i = 0; i < (sge)->ofldqsets; i++)
+#define for_each_rdmarxq(sge, i) for (i = 0; i < (sge)->rdmaqs; i++)
+
+struct l2t_data;
+
+struct adapter {
+	void __iomem *regs;
+	struct pci_dev *pdev;
+	struct device *pdev_dev;
+	unsigned long registered_device_map;
+	unsigned long open_device_map;
+	unsigned long flags;
+
+	const char *name;
+	int msg_enable;
+
+	struct adapter_params params;
+	struct cxgb4_virt_res vres;
+	unsigned int swintr;
+
+	unsigned int wol;
+
+	struct {
+		unsigned short vec;
+		char desc[14];
+	} msix_info[MAX_INGQ + 1];
+
+	struct sge sge;
+
+	struct net_device *port[MAX_NPORTS];
+	u8 chan_map[NCHAN];                   /* channel -> port map */
+
+	struct l2t_data *l2t;
+	void *uld_handle[CXGB4_ULD_MAX];
+	struct list_head list_node;
+
+	struct tid_info tids;
+	void **tid_release_head;
+	spinlock_t tid_release_lock;
+	struct work_struct tid_release_task;
+	bool tid_release_task_busy;
+
+	struct dentry *debugfs_root;
+	struct proc_dir_entry *proc_root;
+
+	spinlock_t stats_lock;
+};
+
+static inline u32 t4_read_reg(struct adapter *adap, u32 reg_addr)
+{
+	return readl(adap->regs + reg_addr);
+}
+
+static inline void t4_write_reg(struct adapter *adap, u32 reg_addr, u32 val)
+{
+	writel(val, adap->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
+
+static inline u64 t4_read_reg64(struct adapter *adap, u32 reg_addr)
+{
+	return readq(adap->regs + reg_addr);
+}
+
+static inline void t4_write_reg64(struct adapter *adap, u32 reg_addr, u64 val)
+{
+	writeq(val, adap->regs + reg_addr);
+}
+
+/**
+ * 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
+ * @idx: the port index
+ *
+ * Return the port_info structure for the port of the given index.
+ */
+static inline struct port_info *adap2pinfo(struct adapter *adap, int idx)
+{
+	return netdev_priv(adap->port[idx]);
+}
+
+/**
+ * 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;
+}
+
+void t4_os_portmod_changed(const struct adapter *adap, int port_id);
+void t4_os_link_changed(struct adapter *adap, int port_id, int link_stat);
+
+void *t4_alloc_mem(size_t size);
+void t4_free_mem(void *addr);
+
+void t4_free_sge_resources(struct adapter *adap);
+irq_handler_t t4_intr_handler(struct adapter *adap);
+netdev_tx_t t4_eth_xmit(struct sk_buff *skb, struct net_device *dev);
+int t4_ethrx_handler(struct sge_rspq *q, const __be64 *rsp,
+		     const struct pkt_gl *gl);
+int t4_mgmt_tx(struct adapter *adap, struct sk_buff *skb);
+int t4_ofld_send(struct adapter *adap, struct sk_buff *skb);
+int t4_sge_alloc_rxq(struct adapter *adap, struct sge_rspq *iq, bool fwevtq,
+		     struct net_device *dev, int intr_idx,
+		     struct sge_fl *fl, rspq_handler_t hnd);
+int t4_sge_alloc_eth_txq(struct adapter *adap, struct sge_eth_txq *txq,
+			 struct net_device *dev, struct netdev_queue *netdevq,
+			 unsigned int iqid);
+int t4_sge_alloc_ctrl_txq(struct adapter *adap, struct sge_ctrl_txq *txq,
+			  struct net_device *dev, unsigned int iqid,
+			  unsigned int cmplqid);
+int t4_sge_alloc_ofld_txq(struct adapter *adap, struct sge_ofld_txq *txq,
+			  struct net_device *dev, unsigned int iqid);
+irqreturn_t t4_sge_intr_msix(int irq, void *cookie);
+void t4_sge_init(struct adapter *adap);
+void t4_sge_start(struct adapter *adap);
+void t4_sge_stop(struct adapter *adap);
+
+#define for_each_port(adapter, iter) \
+	for (iter = 0; iter < (adapter)->params.nports; ++iter)
+
+static inline unsigned int core_ticks_per_usec(const struct adapter *adap)
+{
+	return adap->params.vpd.cclk / 1000;
+}
+
+static inline unsigned int us_to_core_ticks(const struct adapter *adap,
+					    unsigned int us)
+{
+	return (us * adap->params.vpd.cclk) / 1000;
+}
+
+void t4_set_reg_field(struct adapter *adap, unsigned int addr, u32 mask,
+		      u32 val);
+
+int t4_wr_mbox_meat(struct adapter *adap, int mbox, const void *cmd, int size,
+		    void *rpl, bool sleep_ok);
+
+static inline int t4_wr_mbox(struct adapter *adap, int mbox, const void *cmd,
+			     int size, void *rpl)
+{
+	return t4_wr_mbox_meat(adap, mbox, cmd, size, rpl, true);
+}
+
+static inline int t4_wr_mbox_ns(struct adapter *adap, int mbox, const void *cmd,
+				int size, void *rpl)
+{
+	return t4_wr_mbox_meat(adap, mbox, cmd, size, rpl, false);
+}
+
+void t4_intr_enable(struct adapter *adapter);
+void t4_intr_disable(struct adapter *adapter);
+void t4_intr_clear(struct adapter *adapter);
+int t4_slow_intr_handler(struct adapter *adapter);
+
+int t4_link_start(struct adapter *adap, unsigned int mbox, unsigned int port,
+		  struct link_config *lc);
+int t4_restart_aneg(struct adapter *adap, unsigned int mbox, unsigned int port);
+int t4_seeprom_wp(struct adapter *adapter, bool enable);
+int t4_read_flash(struct adapter *adapter, unsigned int addr,
+		  unsigned int nwords, u32 *data, int byte_oriented);
+int t4_load_fw(struct adapter *adapter, const u8 *fw_data, unsigned int size);
+int t4_check_fw_version(struct adapter *adapter);
+int t4_prep_adapter(struct adapter *adapter);
+int t4_port_init(struct adapter *adap, int mbox, int pf, int vf);
+void t4_fatal_err(struct adapter *adapter);
+void t4_set_vlan_accel(struct adapter *adapter, unsigned int ports, int on);
+int t4_set_trace_filter(struct adapter *adapter, const struct trace_params *tp,
+			int filter_index, int enable);
+void t4_get_trace_filter(struct adapter *adapter, struct trace_params *tp,
+			 int filter_index, int *enabled);
+int t4_config_rss_range(struct adapter *adapter, int mbox, unsigned int viid,
+			int start, int n, const u16 *rspq, unsigned int nrspq);
+int t4_config_glbl_rss(struct adapter *adapter, int mbox, unsigned int mode,
+		       unsigned int flags);
+int t4_read_rss(struct adapter *adapter, u16 *entries);
+int t4_mc_read(struct adapter *adap, u32 addr, __be32 *data, u64 *parity);
+int t4_edc_read(struct adapter *adap, int idx, u32 addr, __be32 *data,
+		u64 *parity);
+
+void t4_get_port_stats(struct adapter *adap, int idx, struct port_stats *p);
+void t4_get_lb_stats(struct adapter *adap, int idx, struct lb_port_stats *p);
+
+void t4_read_mtu_tbl(struct adapter *adap, u16 *mtus, u8 *mtu_log);
+void t4_tp_get_err_stats(struct adapter *adap, struct tp_err_stats *st);
+void t4_tp_get_tcp_stats(struct adapter *adap, struct tp_tcp_stats *v4,
+			 struct tp_tcp_stats *v6);
+void t4_load_mtus(struct adapter *adap, const unsigned short *mtus,
+		  const unsigned short *alpha, const unsigned short *beta);
+
+void t4_wol_magic_enable(struct adapter *adap, unsigned int port,
+			 const u8 *addr);
+int t4_wol_pat_enable(struct adapter *adap, unsigned int port, unsigned int map,
+		      u64 mask0, u64 mask1, unsigned int crc, bool enable);
+
+int t4_fw_hello(struct adapter *adap, unsigned int mbox, unsigned int evt_mbox,
+		enum dev_master master, enum dev_state *state);
+int t4_fw_bye(struct adapter *adap, unsigned int mbox);
+int t4_early_init(struct adapter *adap, unsigned int mbox);
+int t4_fw_reset(struct adapter *adap, unsigned int mbox, int reset);
+int t4_query_params(struct adapter *adap, unsigned int mbox, unsigned int pf,
+		    unsigned int vf, unsigned int nparams, const u32 *params,
+		    u32 *val);
+int t4_set_params(struct adapter *adap, unsigned int mbox, unsigned int pf,
+		  unsigned int vf, unsigned int nparams, const u32 *params,
+		  const u32 *val);
+int t4_cfg_pfvf(struct adapter *adap, unsigned int mbox, unsigned int pf,
+		unsigned int vf, unsigned int txq, unsigned int txq_eth_ctrl,
+		unsigned int rxqi, unsigned int rxq, unsigned int tc,
+		unsigned int vi, unsigned int cmask, unsigned int pmask,
+		unsigned int nexact, unsigned int rcaps, unsigned int wxcaps);
+int t4_alloc_vi(struct adapter *adap, unsigned int mbox, unsigned int port,
+		unsigned int pf, unsigned int vf, unsigned int nmac, u8 *mac,
+		unsigned int *rss_size);
+int t4_free_vi(struct adapter *adap, unsigned int mbox, unsigned int pf,
+	       unsigned int vf, unsigned int viid);
+int t4_set_rxmode(struct adapter *adap, unsigned int mbox, unsigned int viid,
+		int mtu, int promisc, int all_multi, int bcast, bool sleep_ok);
+int t4_alloc_mac_filt(struct adapter *adap, unsigned int mbox,
+		      unsigned int viid, bool free, unsigned int naddr,
+		      const u8 **addr, u16 *idx, u64 *hash, bool sleep_ok);
+int t4_change_mac(struct adapter *adap, unsigned int mbox, unsigned int viid,
+		  int idx, const u8 *addr, bool persist, bool add_smt);
+int t4_set_addr_hash(struct adapter *adap, unsigned int mbox, unsigned int viid,
+		     bool ucast, u64 vec, bool sleep_ok);
+int t4_enable_vi(struct adapter *adap, unsigned int mbox, unsigned int viid,
+		 bool rx_en, bool tx_en);
+int t4_identify_port(struct adapter *adap, unsigned int mbox, unsigned int viid,
+		     unsigned int nblinks);
+int t4_mdio_rd(struct adapter *adap, unsigned int mbox, unsigned int phy_addr,
+	       unsigned int mmd, unsigned int reg, u16 *valp);
+int t4_mdio_wr(struct adapter *adap, unsigned int mbox, unsigned int phy_addr,
+	       unsigned int mmd, unsigned int reg, u16 val);
+int t4_iq_start_stop(struct adapter *adap, unsigned int mbox, bool start,
+		     unsigned int pf, unsigned int vf, unsigned int iqid,
+		     unsigned int fl0id, unsigned int fl1id);
+int t4_iq_free(struct adapter *adap, unsigned int mbox, unsigned int pf,
+	       unsigned int vf, unsigned int iqtype, unsigned int iqid,
+	       unsigned int fl0id, unsigned int fl1id);
+int t4_eth_eq_free(struct adapter *adap, unsigned int mbox, unsigned int pf,
+		   unsigned int vf, unsigned int eqid);
+int t4_ctrl_eq_free(struct adapter *adap, unsigned int mbox, unsigned int pf,
+		    unsigned int vf, unsigned int eqid);
+int t4_ofld_eq_free(struct adapter *adap, unsigned int mbox, unsigned int pf,
+		    unsigned int vf, unsigned int eqid);
+int t4_handle_fw_rpl(struct adapter *adap, const __be64 *rpl);
+#endif /* __CXGB4_H__ */
drivers/net/cxgb4/cxgb4_uld.h new file +235 -0

235 lines not on this page · raw message

drivers/net/cxgb4/l2t.c new file +626 -0

626 lines not on this page · raw message

drivers/net/cxgb4/l2t.h new file +110 -0

110 lines not on this page · raw message

Signature · 1 line
1.5.4
#5 lore raw

PATCH 5/6 cxgb4: Add main driver file and driver Makefile

---
 drivers/net/cxgb4/Makefile     |    7 +
 drivers/net/cxgb4/cxgb4_main.c | 4240 ++++++++++++++++++++++++++++++++++++++++
 2 files changed, 4247 insertions(+), 0 deletions(-)
 create mode 100644 drivers/net/cxgb4/Makefile
 create mode 100644 drivers/net/cxgb4/cxgb4_main.c
drivers/net/cxgb4/Makefile new file +7 -0
new file mode 100644
index 0000000..4986674
@@ -0,0 +1,7 @@
+#
+# Chelsio T4 driver
+#
+
+obj-$(CONFIG_CHELSIO_T4) += cxgb4.o
+
+cxgb4-objs := cxgb4_main.o l2t.o t4_hw.o sge.o
drivers/net/cxgb4/cxgb4_main.c new file +4240 -0

2,993 lines not on this page · raw message

The rest of this diff is not shown. View the raw message.

Signature · 1 line
1.5.4

Part of the diff is not shown here. View the raw message.

#6 lore raw

PATCH 6/6 net: Hook up cxgb4 to Kconfig and Makefile

---
 drivers/net/Kconfig  |   25 +++++++++++++++++++++++++
 drivers/net/Makefile |    1 +
 2 files changed, 26 insertions(+), 0 deletions(-)
drivers/net/Kconfig +25 -0
index 4f1f974..944ffd1 100644
@@ -2582,6 +2582,31 @@ config CHELSIO_T3
 	  To compile this driver as a module, choose M here: the module
 	  will be called cxgb3.
 
+config CHELSIO_T4_DEPENDS
+	tristate
+	depends on PCI && INET
+	default y
+
+config CHELSIO_T4
+	tristate "Chelsio Communications T4 Ethernet support"
+	depends on CHELSIO_T4_DEPENDS
+	select FW_LOADER
+	select MDIO
+	help
+	  This driver supports Chelsio T4-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 <[email protected]>.
+
+	  To compile this driver as a module choose M here; the module
+	  will be called cxgb4.
+
 config EHEA
 	tristate "eHEA Ethernet support"
 	depends on IBMEBUS && INET && SPARSEMEM
drivers/net/Makefile +1 -0
index 4788862..a583b50 100644
@@ -19,6 +19,7 @@ obj-$(CONFIG_IXGB) += ixgb/
 obj-$(CONFIG_IP1000) += ipg.o
 obj-$(CONFIG_CHELSIO_T1) += chelsio/
 obj-$(CONFIG_CHELSIO_T3) += cxgb3/
+obj-$(CONFIG_CHELSIO_T4) += cxgb4/
 obj-$(CONFIG_EHEA) += ehea/
 obj-$(CONFIG_CAN) += can/
 obj-$(CONFIG_BONDING) += bonding/
Signature · 1 line
1.5.4
#7 lore raw

Re: [PATCH 5/6] cxgb4: Add main driver file and driver Makefile

In reply to Dimitris Michailidis
Quoted 8 lines · On Tue, 30 Mar 2010 10:52:21 -0800 Dimitris Michailidis <[email protected]> wrote:
drivers/net/cxgb4/cxgb4_main.c new file +8 -0
+static struct cxgb4_proc_entry proc_files[] = {
+#ifdef CONFIG_PROC_FS
+	{ "l2t", 0444, ADAP_NEED_L2T, 0, &t4_l2t_proc_fops },
+#endif
+	{ "lb_stats", 0444, 0, 0, &lb_stats_proc_fops },
+	{ "path_mtus", 0644, 0, 0, &mtutab_proc_fops },
+	{ "qstats", 0444, 0, 0, &sge_stats_proc_fops },
+	{ "rss", 0444, 0, 0, &rss_proc_fops },

drivers/net/cxgb4/cxgb4_main.c line 2647

Quoted 10 lines
drivers/net/cxgb4/cxgb4_main.c new file +10 -0
+	{ "tcp_stats", 0444, 0, 0, &tcp_stats_proc_fops },
+	{ "tids", 0444, ADAP_NEED_OFLD, 0, &tid_info_proc_fops },
+	{ "tp_err_stats", 0444, 0, 0, &tp_err_stats_proc_fops },
+	{ "trace0", 0644, 0, 0, &mps_trc_proc_fops },
+	{ "trace1", 0644, 0, 1, &mps_trc_proc_fops },
+	{ "trace2", 0644, 0, 2, &mps_trc_proc_fops },
+	{ "trace3", 0644, 0, 3, &mps_trc_proc_fops },
+	{ "uld", 0444, 0, 0, &uld_proc_fops },
+};
+

Do you really need this large number of /proc files. It creates another stable API to worry about. If it is just for debugging move it to debugfs, or better yet just drop it.

#8 lore raw

Re: [PATCH 5/6] cxgb4: Add main driver file and driver Makefile

Quoted 17 lines · Stephen Hemminger
Quoted 14 lines · On Tue, 30 Mar 2010 10:52:21 -0800 Dimitris Michailidis <[email protected]> wrote:
drivers/net/cxgb4/cxgb4_main.c new file +14 -0
+static struct cxgb4_proc_entry proc_files[] = {
+#ifdef CONFIG_PROC_FS
+	{ "l2t", 0444, ADAP_NEED_L2T, 0, &t4_l2t_proc_fops },
+#endif
+	{ "lb_stats", 0444, 0, 0, &lb_stats_proc_fops },
+	{ "path_mtus", 0644, 0, 0, &mtutab_proc_fops },
+	{ "qstats", 0444, 0, 0, &sge_stats_proc_fops },
+	{ "rss", 0444, 0, 0, &rss_proc_fops },
+	{ "tcp_stats", 0444, 0, 0, &tcp_stats_proc_fops },
+	{ "tids", 0444, ADAP_NEED_OFLD, 0, &tid_info_proc_fops },
+	{ "tp_err_stats", 0444, 0, 0, &tp_err_stats_proc_fops },
+	{ "trace0", 0644, 0, 0, &mps_trc_proc_fops },
+	{ "trace1", 0644, 0, 1, &mps_trc_proc_fops },
+	{ "trace2", 0644, 0, 2, &mps_trc_proc_fops },
Quoted 9 lines

drivers/net/cxgb4/cxgb4_main.c line 2647

Quoted 4 lines
drivers/net/cxgb4/cxgb4_main.c new file +4 -0
+	{ "trace3", 0644, 0, 3, &mps_trc_proc_fops },
+	{ "uld", 0444, 0, 0, &uld_proc_fops },
+};
+

Do you really need this large number of /proc files. It creates another stable API to worry about. If it is just for debugging move it to debugfs, or better yet just drop it.

The driver already separates its files between debugfs and proc. I put in debugfs files one wouldn't be looking at unless they were investigating some problem and the rest are in proc. Frankly I don't really care which fs they use, they just contain useful information and it would be nice to have them somewhere. Let me know what you want me to do here.

maintainer
#9 lore raw

Re: [PATCH 5/6] cxgb4: Add main driver file and driver Makefile

In reply to Stephen Hemminger

From: Stephen Hemminger <[email protected]>
Date: Tue, 30 Mar 2010 14:19:04 -0700

Quoted 17 lines
Quoted 14 lines · On Tue, 30 Mar 2010 10:52:21 -0800 Dimitris Michailidis <[email protected]> wrote:
drivers/net/cxgb4/cxgb4_main.c new file +14 -0
+static struct cxgb4_proc_entry proc_files[] = {
+#ifdef CONFIG_PROC_FS
+	{ "l2t", 0444, ADAP_NEED_L2T, 0, &t4_l2t_proc_fops },
+#endif
+	{ "lb_stats", 0444, 0, 0, &lb_stats_proc_fops },
+	{ "path_mtus", 0644, 0, 0, &mtutab_proc_fops },
+	{ "qstats", 0444, 0, 0, &sge_stats_proc_fops },
+	{ "rss", 0444, 0, 0, &rss_proc_fops },
+	{ "tcp_stats", 0444, 0, 0, &tcp_stats_proc_fops },
+	{ "tids", 0444, ADAP_NEED_OFLD, 0, &tid_info_proc_fops },
+	{ "tp_err_stats", 0444, 0, 0, &tp_err_stats_proc_fops },
+	{ "trace0", 0644, 0, 0, &mps_trc_proc_fops },
+	{ "trace1", 0644, 0, 1, &mps_trc_proc_fops },
+	{ "trace2", 0644, 0, 2, &mps_trc_proc_fops },
Quoted 8 lines

drivers/net/cxgb4/cxgb4_main.c line 2647

Quoted 4 lines
drivers/net/cxgb4/cxgb4_main.c new file +4 -0
+	{ "trace3", 0644, 0, 3, &mps_trc_proc_fops },
+	{ "uld", 0444, 0, 0, &uld_proc_fops },
+};
+

Do you really need this large number of /proc files. It creates another stable API to worry about. If it is just for debugging move it to debugfs, or better yet just drop it.

I also find this a bit too much.

We have all kinds of ways to export whatever statistics you want. In particular we have ethtool stats of which you can have as many as you wish.

If necessary, we could add a feature to define "views" of ethtool stats so that we can have domains of driver specific statistics if the problem is that you don't want all of these debugging stats to clutter up the "main" ethtool stats.

#10 lore raw

Re: [PATCH 5/6] cxgb4: Add main driver file and driver Makefile

Quoted 34 lines · David Miller

From: Stephen Hemminger <[email protected]>
Date: Tue, 30 Mar 2010 14:19:04 -0700

Quoted 18 lines
Quoted 14 lines · On Tue, 30 Mar 2010 10:52:21 -0800 Dimitris Michailidis <[email protected]> wrote:
drivers/net/cxgb4/cxgb4_main.c new file +14 -0
+static struct cxgb4_proc_entry proc_files[] = {
+#ifdef CONFIG_PROC_FS
+	{ "l2t", 0444, ADAP_NEED_L2T, 0, &t4_l2t_proc_fops },
+#endif
+	{ "lb_stats", 0444, 0, 0, &lb_stats_proc_fops },
+	{ "path_mtus", 0644, 0, 0, &mtutab_proc_fops },
+	{ "qstats", 0444, 0, 0, &sge_stats_proc_fops },
+	{ "rss", 0444, 0, 0, &rss_proc_fops },
+	{ "tcp_stats", 0444, 0, 0, &tcp_stats_proc_fops },
+	{ "tids", 0444, ADAP_NEED_OFLD, 0, &tid_info_proc_fops },
+	{ "tp_err_stats", 0444, 0, 0, &tp_err_stats_proc_fops },
+	{ "trace0", 0644, 0, 0, &mps_trc_proc_fops },
+	{ "trace1", 0644, 0, 1, &mps_trc_proc_fops },
+	{ "trace2", 0644, 0, 2, &mps_trc_proc_fops },
Quoted 9 lines

drivers/net/cxgb4/cxgb4_main.c line 2647

Quoted 5 lines
drivers/net/cxgb4/cxgb4_main.c new file +4 -0
+	{ "trace3", 0644, 0, 3, &mps_trc_proc_fops },
+	{ "uld", 0444, 0, 0, &uld_proc_fops },
+};
+

Do you really need this large number of /proc files. It creates another stable API to worry about. If it is just for debugging move it to debugfs, or better yet just drop it.

I also find this a bit too much.

Quoted 10 lines

We have all kinds of ways to export whatever statistics you want. In particular we have ethtool stats of which you can have as many as you wish.

If necessary, we could add a feature to define "views" of ethtool stats so that we can have domains of driver specific statistics if the problem is that you don't want all of these debugging stats to clutter up the "main" ethtool stats.

This sounds interesting. Dave, how would you envision the ethtool interface for this? A new modifier to --statistics? Like: 'ethtool --statistics ethX viewname' where viewname could maybe be passed to the driver to allow arbitrary views? Or something else?

Steve.

maintainer
#11 lore raw

Re: [PATCH 5/6] cxgb4: Add main driver file and driver Makefile

From: Steve Wise <[email protected]>
Date: Wed, 31 Mar 2010 14:31:13 -0500

Quoted 4 lines

This sounds interesting. Dave, how would you envision the ethtool interface for this? A new modifier to --statistics? Like: 'ethtool --statistics ethX viewname' where viewname could maybe be passed to the driver to allow arbitrary views? Or something else?

I really don't care much about the userland side, I'm only interested in how you implement this in the kernel :-)

#12 lore raw

Re: [PATCH 5/6] cxgb4: Add main driver file and driver Makefile

In reply to David Miller
Quoted 30 lines · David Miller

From: Stephen Hemminger <[email protected]>
Date: Tue, 30 Mar 2010 14:19:04 -0700

Quoted 16 lines
Quoted 13 lines · On Tue, 30 Mar 2010 10:52:21 -0800 Dimitris Michailidis <[email protected]> wrote:
drivers/net/cxgb4/cxgb4_main.c new file +13 -0
+static struct cxgb4_proc_entry proc_files[] = {
+#ifdef CONFIG_PROC_FS
+	{ "l2t", 0444, ADAP_NEED_L2T, 0, &t4_l2t_proc_fops },
+#endif
+	{ "lb_stats", 0444, 0, 0, &lb_stats_proc_fops },
+	{ "path_mtus", 0644, 0, 0, &mtutab_proc_fops },
+	{ "qstats", 0444, 0, 0, &sge_stats_proc_fops },
+	{ "rss", 0444, 0, 0, &rss_proc_fops },
+	{ "tcp_stats", 0444, 0, 0, &tcp_stats_proc_fops },
+	{ "tids", 0444, ADAP_NEED_OFLD, 0, &tid_info_proc_fops },
+	{ "tp_err_stats", 0444, 0, 0, &tp_err_stats_proc_fops },
+	{ "trace0", 0644, 0, 0, &mps_trc_proc_fops },
+	{ "trace1", 0644, 0, 1, &mps_trc_proc_fops },
Quoted 8 lines

drivers/net/cxgb4/cxgb4_main.c line 2647

Quoted 5 lines
drivers/net/cxgb4/cxgb4_main.c new file +5 -0
+	{ "trace2", 0644, 0, 2, &mps_trc_proc_fops },
+	{ "trace3", 0644, 0, 3, &mps_trc_proc_fops },
+	{ "uld", 0444, 0, 0, &uld_proc_fops },
+};
+

Do you really need this large number of /proc files. It creates another stable API to worry about. If it is just for debugging move it to debugfs, or better yet just drop it.

I also find this a bit too much.

Quoted 8 lines

We have all kinds of ways to export whatever statistics you want. In particular we have ethtool stats of which you can have as many as you wish.

If necessary, we could add a feature to define "views" of ethtool stats so that we can have domains of driver specific statistics if the problem is that you don't want all of these debugging stats to clutter up the "main" ethtool stats.

It wasn't a concern with having too many ethtool stats that led to the use of proc, rather it was that most of this information isn't associated with netdevs. Let me give an example. Like most virtualized NICs this device contains a switch. Some of the switch statistics are provided in the proc files. While they are statistics, logically they are not associated with any of the netdevs. If ethtool were to be extended so that the specified interface would access the underlying HW to get information possibly unrelated to the interface personally I'd be OK with that, I am just pointing out what kind of extension we are talking about.

Also some of the files here aren't statistics but report other functionality
of the device, usually also not tied to netdevs (e.g., the trace* files
are again related to the switch).

So, I propose getting rid of 3-4 of these files that are of lesser value and moving the rest to debugfs for now. If some alternative through ethtool or something becomes available I can get rid of anything that can be handled through a more general facility. Would that be acceptable?

maintainer
#13 lore raw

Re: [PATCH 5/6] cxgb4: Add main driver file and driver Makefile

From: Dimitris Michailidis <[email protected]>
Date: Wed, 31 Mar 2010 17:19:21 -0700

Quoted 5 lines

So, I propose getting rid of 3-4 of these files that are of lesser value and moving the rest to debugfs for now. If some alternative through ethtool or something becomes available I can get rid of anything that can be handled through a more general facility. Would that be acceptable?

You can use sysfs.

I have a similar issue on NIU wherein a top-level set of logic behaves like a parent virtual device and contains attributes I'd like to export.

I create a dummy platform device to represent this node and hang the device sysfs information in that device's sysfs directory. It shows up as /sys/devices/platform/niu.%d and the network ports underneath this parent appear as symlinks named "port%d" (part of this is setup in drivers/net/niu.c:niu_get_parent() and niu_new_parent())

You could do something similar to represent the fact that these statistics are virtualized and belong to some centralized virtual entity that represents one or more actual device ports.

#14 lore raw

Re: [PATCH 5/6] cxgb4: Add main driver file and driver Makefile

Quoted 7 lines · David Miller
Quoted 5 lines

So, I propose getting rid of 3-4 of these files that are of lesser value and moving the rest to debugfs for now. If some alternative through ethtool or something becomes available I can get rid of anything that can be handled through a more general facility. Would that be acceptable?

You can use sysfs.

sysfs is a possibility but I thought Stephen's initial concern was that I was adding too many of these proc files and that they were creating a potential API. sysfs will result in a lot more files with its value-per-file model and I think sysfs and proc are similar in "APIness". So it's not clear to me how going to sysfs would address Stephen's point. The remove-a-few plus move-to-debugfs proposal was in order to end up with fewer files in a non-API filesystem.

As these builtin switches become more common I expect an official way to represent and access them will emerge but maybe it's not a good idea to introduce a sysfs model for them as part of this driver submission.

maintainer
#15 lore raw

Re: [PATCH 5/6] cxgb4: Add main driver file and driver Makefile

From: Dimitris Michailidis <[email protected]>
Date: Wed, 31 Mar 2010 18:34:31 -0700

Quoted 7 lines
Quoted 5 lines · David Miller
Quoted 5 lines

So, I propose getting rid of 3-4 of these files that are of lesser value and moving the rest to debugfs for now. If some alternative through ethtool or something becomes available I can get rid of anything that can be handled through a more general facility. Would that be acceptable?

Quoted 5 lines
Quoted 1 line

You can use sysfs.

sysfs is a possibility but I thought Stephen's initial concern was that I was adding too many of these proc files and that they were creating a potential API.

Yes, since procfs is essentially deprecated these days.

Quoted 5 lines

sysfs will result in a lot more files with its value-per-file model and I think sysfs and proc are similar in "APIness". So it's not clear to me how going to sysfs would address Stephen's point. The remove-a-few plus move-to-debugfs proposal was in order to end up with fewer files in a non-API filesystem.

It's in fact easier to retain API by using sysfs. Instead of having to worry about the format of a procfs file listing entries one by one per line, under sysfs you just add a new file to export new values.

Quoted 4 lines

As these builtin switches become more common I expect an official way to represent and access them will emerge but maybe it's not a good idea to introduce a sysfs model for them as part of this driver submission.

What nodes you create under your own device object in sysfs is your domain and your business. Since it's one value per file there is no real complexity in making sure tools can display the values properly.

#16 lore raw

Re: [PATCH 1/6] cxgb4: Add register, message, and FW definitions

In reply to Dimitris Michailidis

drivers/net/cxgb4/t4_msg.h line 157

Quoted 5 lines · On Tue, 30 Mar 2010 10:52:17 -0800 Dimitris Michailidis <[email protected]> wrote:
drivers/net/cxgb4/t4_msg.h new file +5 -0
+union opcode_tid {
+	__be32 opcode_tid;
+	__u8 opcode;
+};
+

The intent of the __types like __u8 etc are that these types should only be used by structures that are part in some way of a kernel to userspace API. Any structures which are only used in the kernel should just use the regular type (ie u8).

I say intent because many places do not follow this practice.

#17 lore raw

Re: [PATCH 1/6] cxgb4: Add register, message, and FW definitions

Quoted 14 lines · Stephen Hemminger

drivers/net/cxgb4/t4_msg.h line 157

Quoted 5 lines · On Tue, 30 Mar 2010 10:52:17 -0800 Dimitris Michailidis <[email protected]> wrote:
drivers/net/cxgb4/t4_msg.h new file +5 -0
+union opcode_tid {
+	__be32 opcode_tid;
+	__u8 opcode;
+};
+

The intent of the __types like __u8 etc are that these types should only be used by structures that are part in some way of a kernel to userspace API. Any structures which are only used in the kernel should just use the regular type (ie u8).

Quoted 2 lines

I say intent because many places do not follow this practice.

These headers aren't meant to be seen by user space and I can certainly replace __u8 with u8. I'll update the patch.