RDMA/cxgb4: Add driver for Chelsio T4 RNIC

Add an RDMA/iWARP driver for Chelsio T4 Ethernet adapters.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

16 files changed, 9447 insertions, 0 deletions

diff --git a/drivers/infiniband/Kconfig b/drivers/infiniband/Kconfig
index 975adce5f40..330d2a42336 100644
--- a/drivers/infiniband/Kconfig
+++ b/drivers/infiniband/Kconfig
@@ -46,6 +46,7 @@ source "drivers/infiniband/hw/ipath/Kconfig"
 source "drivers/infiniband/hw/ehca/Kconfig"
 source "drivers/infiniband/hw/amso1100/Kconfig"
 source "drivers/infiniband/hw/cxgb3/Kconfig"
+source "drivers/infiniband/hw/cxgb4/Kconfig"
 source "drivers/infiniband/hw/mlx4/Kconfig"
 source "drivers/infiniband/hw/nes/Kconfig"
 
diff --git a/drivers/infiniband/Makefile b/drivers/infiniband/Makefile
index ed35e449624..0c4e589d746 100644
--- a/drivers/infiniband/Makefile
+++ b/drivers/infiniband/Makefile
@@ -4,6 +4,7 @@ obj-$(CONFIG_INFINIBAND_IPATH)		+= hw/ipath/
 obj-$(CONFIG_INFINIBAND_EHCA)		+= hw/ehca/
 obj-$(CONFIG_INFINIBAND_AMSO1100)	+= hw/amso1100/
 obj-$(CONFIG_INFINIBAND_CXGB3)		+= hw/cxgb3/
+obj-$(CONFIG_INFINIBAND_CXGB4)		+= hw/cxgb4/
 obj-$(CONFIG_MLX4_INFINIBAND)		+= hw/mlx4/
 obj-$(CONFIG_INFINIBAND_NES)		+= hw/nes/
 obj-$(CONFIG_INFINIBAND_IPOIB)		+= ulp/ipoib/
diff --git a/drivers/infiniband/hw/cxgb4/Kconfig b/drivers/infiniband/hw/cxgb4/Kconfig
new file mode 100644
index 00000000000..ccb85eaaad7
--- /dev/null
+++ b/drivers/infiniband/hw/cxgb4/Kconfig
@@ -0,0 +1,18 @@
+config INFINIBAND_CXGB4
+	tristate "Chelsio T4 RDMA Driver"
+	depends on CHELSIO_T4 && INET
+	select GENERIC_ALLOCATOR
+	---help---
+	  This is an iWARP/RDMA driver for the Chelsio T4 1GbE and
+	  10GbE adapters.
+
+	  For general information about Chelsio and our products, visit
+	  our website at <http://www.chelsio.com>.
+
+	  For customer support, please visit our customer support page at
+	  <http://www.chelsio.com/support.htm>.
+
+	  Please send feedback to <linux-bugs@chelsio.com>.
+
+	  To compile this driver as a module, choose M here: the module
+	  will be called iw_cxgb4.
diff --git a/drivers/infiniband/hw/cxgb4/Makefile b/drivers/infiniband/hw/cxgb4/Makefile
new file mode 100644
index 00000000000..e31a499f017
--- /dev/null
+++ b/drivers/infiniband/hw/cxgb4/Makefile
@@ -0,0 +1,5 @@
+EXTRA_CFLAGS += -Idrivers/net/cxgb4
+
+obj-$(CONFIG_INFINIBAND_CXGB4) += iw_cxgb4.o
+
+iw_cxgb4-y :=  device.o cm.o provider.o mem.o cq.o qp.o resource.o ev.o
diff --git a/drivers/infiniband/hw/cxgb4/cm.c b/drivers/infiniband/hw/cxgb4/cm.c
new file mode 100644
index 00000000000..07b068be0cf
--- /dev/null
+++ b/drivers/infiniband/hw/cxgb4/cm.c
@@ -0,0 +1,2330 @@
+/*
+ * Copyright (c) 2009-2010 Chelsio, 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/module.h>
+#include <linux/list.h>
+#include <linux/workqueue.h>
+#include <linux/skbuff.h>
+#include <linux/timer.h>
+#include <linux/notifier.h>
+#include <linux/inetdevice.h>
+#include <linux/ip.h>
+#include <linux/tcp.h>
+
+#include <net/neighbour.h>
+#include <net/netevent.h>
+#include <net/route.h>
+
+#include "iw_cxgb4.h"
+
+static char *states[] = {
+	"idle",
+	"listen",
+	"connecting",
+	"mpa_wait_req",
+	"mpa_req_sent",
+	"mpa_req_rcvd",
+	"mpa_rep_sent",
+	"fpdu_mode",
+	"aborting",
+	"closing",
+	"moribund",
+	"dead",
+	NULL,
+};
+
+static int enable_tcp_timestamps;
+module_param(enable_tcp_timestamps, int, 0644);
+MODULE_PARM_DESC(enable_tcp_timestamps, "Enable tcp timestamps (default=0)");
+
+static int enable_tcp_sack;
+module_param(enable_tcp_sack, int, 0644);
+MODULE_PARM_DESC(enable_tcp_sack, "Enable tcp SACK (default=0)");
+
+static int enable_tcp_window_scaling = 1;
+module_param(enable_tcp_window_scaling, int, 0644);
+MODULE_PARM_DESC(enable_tcp_window_scaling,
+		 "Enable tcp window scaling (default=1)");
+
+int c4iw_debug;
+module_param(c4iw_debug, int, 0644);
+MODULE_PARM_DESC(c4iw_debug, "Enable debug logging (default=0)");
+
+static int peer2peer;
+module_param(peer2peer, int, 0644);
+MODULE_PARM_DESC(peer2peer, "Support peer2peer ULPs (default=0)");
+
+static int p2p_type = FW_RI_INIT_P2PTYPE_READ_REQ;
+module_param(p2p_type, int, 0644);
+MODULE_PARM_DESC(p2p_type, "RDMAP opcode to use for the RTR message: "
+			   "1=RDMA_READ 0=RDMA_WRITE (default 1)");
+
+static int ep_timeout_secs = 60;
+module_param(ep_timeout_secs, int, 0644);
+MODULE_PARM_DESC(ep_timeout_secs, "CM Endpoint operation timeout "
+				   "in seconds (default=60)");
+
+static int mpa_rev = 1;
+module_param(mpa_rev, int, 0644);
+MODULE_PARM_DESC(mpa_rev, "MPA Revision, 0 supports amso1100, "
+		 "1 is spec compliant. (default=1)");
+
+static int markers_enabled;
+module_param(markers_enabled, int, 0644);
+MODULE_PARM_DESC(markers_enabled, "Enable MPA MARKERS (default(0)=disabled)");
+
+static int crc_enabled = 1;
+module_param(crc_enabled, int, 0644);
+MODULE_PARM_DESC(crc_enabled, "Enable MPA CRC (default(1)=enabled)");
+
+static int rcv_win = 256 * 1024;
+module_param(rcv_win, int, 0644);
+MODULE_PARM_DESC(rcv_win, "TCP receive window in bytes (default=256KB)");
+
+static int snd_win = 32 * 1024;
+module_param(snd_win, int, 0644);
+MODULE_PARM_DESC(snd_win, "TCP send window in bytes (default=32KB)");
+
+static void process_work(struct work_struct *work);
+static struct workqueue_struct *workq;
+static DECLARE_WORK(skb_work, process_work);
+
+static struct sk_buff_head rxq;
+static c4iw_handler_func work_handlers[NUM_CPL_CMDS];
+c4iw_handler_func c4iw_handlers[NUM_CPL_CMDS];
+
+static struct sk_buff *get_skb(struct sk_buff *skb, int len, gfp_t gfp);
+static void ep_timeout(unsigned long arg);
+static void connect_reply_upcall(struct c4iw_ep *ep, int status);
+
+static void start_ep_timer(struct c4iw_ep *ep)
+{
+	PDBG("%s ep %p\n", __func__, ep);
+	if (timer_pending(&ep->timer)) {
+		PDBG("%s stopped / restarted timer ep %p\n", __func__, ep);
+		del_timer_sync(&ep->timer);
+	} else
+		c4iw_get_ep(&ep->com);
+	ep->timer.expires = jiffies + ep_timeout_secs * HZ;
+	ep->timer.data = (unsigned long)ep;
+	ep->timer.function = ep_timeout;
+	add_timer(&ep->timer);
+}
+
+static void stop_ep_timer(struct c4iw_ep *ep)
+{
+	PDBG("%s ep %p\n", __func__, ep);
+	if (!timer_pending(&ep->timer)) {
+		printk(KERN_ERR "%s timer stopped when its not running! "
+		       "ep %p state %u\n", __func__, ep, ep->com.state);
+		WARN_ON(1);
+		return;
+	}
+	del_timer_sync(&ep->timer);
+	c4iw_put_ep(&ep->com);
+}
+
+static int c4iw_l2t_send(struct c4iw_rdev *rdev, struct sk_buff *skb,
+		  struct l2t_entry *l2e)
+{
+	int	error = 0;
+
+	if (c4iw_fatal_error(rdev)) {
+		kfree_skb(skb);
+		PDBG("%s - device in error state - dropping\n", __func__);
+		return -EIO;
+	}
+	error = cxgb4_l2t_send(rdev->lldi.ports[0], skb, l2e);
+	if (error < 0)
+		kfree_skb(skb);
+	return error;
+}
+
+int c4iw_ofld_send(struct c4iw_rdev *rdev, struct sk_buff *skb)
+{
+	int	error = 0;
+
+	if (c4iw_fatal_error(rdev)) {
+		kfree_skb(skb);
+		PDBG("%s - device in error state - dropping\n", __func__);
+		return -EIO;
+	}
+	error = cxgb4_ofld_send(rdev->lldi.ports[0], skb);
+	if (error < 0)
+		kfree_skb(skb);
+	return error;
+}
+
+static void release_tid(struct c4iw_rdev *rdev, u32 hwtid, struct sk_buff *skb)
+{
+	struct cpl_tid_release *req;
+
+	skb = get_skb(skb, sizeof *req, GFP_KERNEL);
+	if (!skb)
+		return;
+	req = (struct cpl_tid_release *) skb_put(skb, sizeof(*req));
+	INIT_TP_WR(req, hwtid);
+	OPCODE_TID(req) = cpu_to_be32(MK_OPCODE_TID(CPL_TID_RELEASE, hwtid));
+	set_wr_txq(skb, CPL_PRIORITY_SETUP, 0);
+	c4iw_ofld_send(rdev, skb);
+	return;
+}
+
+static void set_emss(struct c4iw_ep *ep, u16 opt)
+{
+	ep->emss = ep->com.dev->rdev.lldi.mtus[GET_TCPOPT_MSS(opt)] - 40;
+	ep->mss = ep->emss;
+	if (GET_TCPOPT_TSTAMP(opt))
+		ep->emss -= 12;
+	if (ep->emss < 128)
+		ep->emss = 128;
+	PDBG("%s mss_idx %u mss %u emss=%u\n", __func__, GET_TCPOPT_MSS(opt),
+	     ep->mss, ep->emss);
+}
+
+static enum c4iw_ep_state state_read(struct c4iw_ep_common *epc)
+{
+	unsigned long flags;
+	enum c4iw_ep_state state;
+
+	spin_lock_irqsave(&epc->lock, flags);
+	state = epc->state;
+	spin_unlock_irqrestore(&epc->lock, flags);
+	return state;
+}
+
+static void __state_set(struct c4iw_ep_common *epc, enum c4iw_ep_state new)
+{
+	epc->state = new;
+}
+
+static void state_set(struct c4iw_ep_common *epc, enum c4iw_ep_state new)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&epc->lock, flags);
+	PDBG("%s - %s -> %s\n", __func__, states[epc->state], states[new]);
+	__state_set(epc, new);
+	spin_unlock_irqrestore(&epc->lock, flags);
+	return;
+}
+
+static void *alloc_ep(int size, gfp_t gfp)
+{
+	struct c4iw_ep_common *epc;
+
+	epc = kzalloc(size, gfp);
+	if (epc) {
+		kref_init(&epc->kref);
+		spin_lock_init(&epc->lock);
+		init_waitqueue_head(&epc->waitq);
+	}
+	PDBG("%s alloc ep %p\n", __func__, epc);
+	return epc;
+}
+
+void _c4iw_free_ep(struct kref *kref)
+{
+	struct c4iw_ep *ep;
+
+	ep = container_of(kref, struct c4iw_ep, com.kref);
+	PDBG("%s ep %p state %s\n", __func__, ep, states[state_read(&ep->com)]);
+	if (test_bit(RELEASE_RESOURCES, &ep->com.flags)) {
+		cxgb4_remove_tid(ep->com.dev->rdev.lldi.tids, 0, ep->hwtid);
+		dst_release(ep->dst);
+		cxgb4_l2t_release(ep->l2t);
+	}
+	kfree(ep);
+}
+
+static void release_ep_resources(struct c4iw_ep *ep)
+{
+	set_bit(RELEASE_RESOURCES, &ep->com.flags);
+	c4iw_put_ep(&ep->com);
+}
+
+static void process_work(struct work_struct *work)
+{
+	struct sk_buff *skb = NULL;
+	struct c4iw_dev *dev;
+	struct cpl_act_establish *rpl = cplhdr(skb);
+	unsigned int opcode;
+	int ret;
+
+	while ((skb = skb_dequeue(&rxq))) {
+		rpl = cplhdr(skb);
+		dev = *((struct c4iw_dev **) (skb->cb + sizeof(void *)));
+		opcode = rpl->ot.opcode;
+
+		BUG_ON(!work_handlers[opcode]);
+		ret = work_handlers[opcode](dev, skb);
+		if (!ret)
+			kfree_skb(skb);
+	}
+}
+
+static int status2errno(int status)
+{
+	switch (status) {
+	case CPL_ERR_NONE:
+		return 0;
+	case CPL_ERR_CONN_RESET:
+		return -ECONNRESET;
+	case CPL_ERR_ARP_MISS:
+		return -EHOSTUNREACH;
+	case CPL_ERR_CONN_TIMEDOUT:
+		return -ETIMEDOUT;
+	case CPL_ERR_TCAM_FULL:
+		return -ENOMEM;
+	case CPL_ERR_CONN_EXIST:
+		return -EADDRINUSE;
+	default:
+		return -EIO;
+	}
+}
+
+/*
+ * Try and reuse skbs already allocated...
+ */
+static struct sk_buff *get_skb(struct sk_buff *skb, int len, gfp_t gfp)
+{
+	if (skb && !skb_is_nonlinear(skb) && !skb_cloned(skb)) {
+		skb_trim(skb, 0);
+		skb_get(skb);
+		skb_reset_transport_header(skb);
+	} else {
+		skb = alloc_skb(len, gfp);
+	}
+	return skb;
+}
+
+static struct rtable *find_route(struct c4iw_dev *dev, __be32 local_ip,
+				 __be32 peer_ip, __be16 local_port,
+				 __be16 peer_port, u8 tos)
+{
+	struct rtable *rt;
+	struct flowi fl = {
+		.oif = 0,
+		.nl_u = {
+			 .ip4_u = {
+				   .daddr = peer_ip,
+				   .saddr = local_ip,
+				   .tos = tos}
+			 },
+		.proto = IPPROTO_TCP,
+		.uli_u = {
+			  .ports = {
+				    .sport = local_port,
+				    .dport = peer_port}
+			  }
+	};
+
+	if (ip_route_output_flow(&init_net, &rt, &fl, NULL, 0))
+		return NULL;
+	return rt;
+}
+
+static void arp_failure_discard(void *handle, struct sk_buff *skb)
+{
+	PDBG("%s c4iw_dev %p\n", __func__, handle);
+	kfree_skb(skb);
+}
+
+/*
+ * Handle an ARP failure for an active open.
+ */
+static void act_open_req_arp_failure(void *handle, struct sk_buff *skb)
+{
+	printk(KERN_ERR MOD "ARP failure duing connect\n");
+	kfree_skb(skb);
+}
+
+/*
+ * Handle an ARP failure for a CPL_ABORT_REQ.  Change it into a no RST variant
+ * and send it along.
+ */
+static void abort_arp_failure(void *handle, struct sk_buff *skb)
+{
+	struct c4iw_rdev *rdev = handle;
+	struct cpl_abort_req *req = cplhdr(skb);
+
+	PDBG("%s rdev %p\n", __func__, rdev);
+	req->cmd = CPL_ABORT_NO_RST;
+	c4iw_ofld_send(rdev, skb);
+}
+
+static void send_flowc(struct c4iw_ep *ep, struct sk_buff *skb)
+{
+	unsigned int flowclen = 80;
+	struct fw_flowc_wr *flowc;
+	int i;
+
+	skb = get_skb(skb, flowclen, GFP_KERNEL);
+	flowc = (struct fw_flowc_wr *)__skb_put(skb, flowclen);
+
+	flowc->op_to_nparams = cpu_to_be32(FW_WR_OP(FW_FLOWC_WR) |
+					   FW_FLOWC_WR_NPARAMS(8));
+	flowc->flowid_len16 = cpu_to_be32(FW_WR_LEN16(DIV_ROUND_UP(flowclen,
+					  16)) | FW_WR_FLOWID(ep->hwtid));
+
+	flowc->mnemval[0].mnemonic = FW_FLOWC_MNEM_PFNVFN;
+	flowc->mnemval[0].val = cpu_to_be32(0);
+	flowc->mnemval[1].mnemonic = FW_FLOWC_MNEM_CH;
+	flowc->mnemval[1].val = cpu_to_be32(ep->tx_chan);
+	flowc->mnemval[2].mnemonic = FW_FLOWC_MNEM_PORT;
+	flowc->mnemval[2].val = cpu_to_be32(ep->tx_chan);
+	flowc->mnemval[3].mnemonic = FW_FLOWC_MNEM_IQID;
+	flowc->mnemval[3].val = cpu_to_be32(ep->rss_qid);
+	flowc->mnemval[4].mnemonic = FW_FLOWC_MNEM_SNDNXT;
+	flowc->mnemval[4].val = cpu_to_be32(ep->snd_seq);
+	flowc->mnemval[5].mnemonic = FW_FLOWC_MNEM_RCVNXT;
+	flowc->mnemval[5].val = cpu_to_be32(ep->rcv_seq);
+	flowc->mnemval[6].mnemonic = FW_FLOWC_MNEM_SNDBUF;
+	flowc->mnemval[6].val = cpu_to_be32(snd_win);
+	flowc->mnemval[7].mnemonic = FW_FLOWC_MNEM_MSS;
+	flowc->mnemval[7].val = cpu_to_be32(ep->emss);
+	/* Pad WR to 16 byte boundary */
+	flowc->mnemval[8].mnemonic = 0;
+	flowc->mnemval[8].val = 0;
+	for (i = 0; i < 9; i++) {
+		flowc->mnemval[i].r4[0] = 0;
+		flowc->mnemval[i].r4[1] = 0;
+		flowc->mnemval[i].r4[2] = 0;
+	}
+
+	set_wr_txq(skb, CPL_PRIORITY_DATA, ep->txq_idx);
+	c4iw_ofld_send(&ep->com.dev->rdev, skb);
+}
+
+static int send_halfclose(struct c4iw_ep *ep, gfp_t gfp)
+{
+	struct cpl_close_con_req *req;
+	struct sk_buff *skb;
+	int wrlen = roundup(sizeof *req, 16);
+
+	PDBG("%s ep %p tid %u\n", __func__, ep, ep->hwtid);
+	skb = get_skb(NULL, wrlen, gfp);
+	if (!skb) {
+		printk(KERN_ERR MOD "%s - failed to alloc skb\n", __func__);
+		return -ENOMEM;
+	}
+	set_wr_txq(skb, CPL_PRIORITY_DATA, ep->txq_idx);
+	t4_set_arp_err_handler(skb, NULL, arp_failure_discard);
+	req = (struct cpl_close_con_req *) skb_put(skb, wrlen);
+	memset(req, 0, wrlen);
+	INIT_TP_WR(req, ep->hwtid);
+	OPCODE_TID(req) = cpu_to_be32(MK_OPCODE_TID(CPL_CLOSE_CON_REQ,
+						    ep->hwtid));
+	return c4iw_l2t_send(&ep->com.dev->rdev, skb, ep->l2t);
+}
+
+static int send_abort(struct c4iw_ep *ep, struct sk_buff *skb, gfp_t gfp)
+{
+	struct cpl_abort_req *req;
+	int wrlen = roundup(sizeof *req, 16);
+
+	PDBG("%s ep %p tid %u\n", __func__, ep, ep->hwtid);
+	skb = get_skb(skb, wrlen, gfp);
+	if (!skb) {
+		printk(KERN_ERR MOD "%s - failed to alloc skb.\n",
+		       __func__);
+		return -ENOMEM;
+	}
+	set_wr_txq(skb, CPL_PRIORITY_DATA, ep->txq_idx);
+	t4_set_arp_err_handler(skb, &ep->com.dev->rdev, abort_arp_failure);
+	req = (struct cpl_abort_req *) skb_put(skb, wrlen);
+	memset(req, 0, wrlen);
+	INIT_TP_WR(req, ep->hwtid);
+	OPCODE_TID(req) = cpu_to_be32(MK_OPCODE_TID(CPL_ABORT_REQ, ep->hwtid));
+	req->cmd = CPL_ABORT_SEND_RST;
+	return c4iw_l2t_send(&ep->com.dev->rdev, skb, ep->l2t);
+}
+
+static int send_connect(struct c4iw_ep *ep)
+{
+	struct cpl_act_open_req *req;
+	struct sk_buff *skb;
+	u64 opt0;
+	u32 opt2;
+	unsigned int mtu_idx;
+	int wscale;
+	int wrlen = roundup(sizeof *req, 16);
+
+	PDBG("%s ep %p atid %u\n", __func__, ep, ep->atid);
+
+	skb = get_skb(NULL, wrlen, GFP_KERNEL);
+	if (!skb) {
+		printk(KERN_ERR MOD "%s - failed to alloc skb.\n",
+		       __func__);
+		return -ENOMEM;
+	}
+	set_wr_txq(skb, CPL_PRIORITY_SETUP, ep->txq_idx);
+
+	cxgb4_best_mtu(ep->com.dev->rdev.lldi.mtus, ep->mtu, &mtu_idx);
+	wscale = compute_wscale(rcv_win);
+	opt0 = KEEP_ALIVE(1) |
+	       WND_SCALE(wscale) |
+	       MSS_IDX(mtu_idx) |
+	       L2T_IDX(ep->l2t->idx) |
+	       TX_CHAN(ep->tx_chan) |
+	       SMAC_SEL(ep->smac_idx) |
+	       DSCP(ep->tos) |
+	       RCV_BUFSIZ(rcv_win>>10);
+	opt2 = RX_CHANNEL(0) |
+	       RSS_QUEUE_VALID | RSS_QUEUE(ep->rss_qid);
+	if (enable_tcp_timestamps)
+		opt2 |= TSTAMPS_EN(1);
+	if (enable_tcp_sack)
+		opt2 |= SACK_EN(1);
+	if (wscale && enable_tcp_window_scaling)
+		opt2 |= WND_SCALE_EN(1);
+	t4_set_arp_err_handler(skb, NULL, act_open_req_arp_failure);
+
+	req = (struct cpl_act_open_req *) skb_put(skb, wrlen);
+	INIT_TP_WR(req, 0);
+	OPCODE_TID(req) = cpu_to_be32(
+		MK_OPCODE_TID(CPL_ACT_OPEN_REQ, ((ep->rss_qid<<14)|ep->atid)));
+	req->local_port = ep->com.local_addr.sin_port;
+	req->peer_port = ep->com.remote_addr.sin_port;
+	req->local_ip = ep->com.local_addr.sin_addr.s_addr;
+	req->peer_ip = ep->com.remote_addr.sin_addr.s_addr;
+	req->opt0 = cpu_to_be64(opt0);
+	req->params = 0;
+	req->opt2 = cpu_to_be32(opt2);
+	return c4iw_l2t_send(&ep->com.dev->rdev, skb, ep->l2t);
+}
+
+static void send_mpa_req(struct c4iw_ep *ep, struct sk_buff *skb)
+{
+	int mpalen, wrlen;
+	struct fw_ofld_tx_data_wr *req;
+	struct mpa_message *mpa;
+
+	PDBG("%s ep %p tid %u pd_len %d\n", __func__, ep, ep->hwtid, ep->plen);
+
+	BUG_ON(skb_cloned(skb));
+
+	mpalen = sizeof(*mpa) + ep->plen;
+	wrlen = roundup(mpalen + sizeof *req, 16);
+	skb = get_skb(skb, wrlen, GFP_KERNEL);
+	if (!skb) {
+		connect_reply_upcall(ep, -ENOMEM);
+		return;
+	}
+	set_wr_txq(skb, CPL_PRIORITY_DATA, ep->txq_idx);
+
+	req = (struct fw_ofld_tx_data_wr *)skb_put(skb, wrlen);
+	memset(req, 0, wrlen);
+	req->op_to_immdlen = cpu_to_be32(
+		FW_WR_OP(FW_OFLD_TX_DATA_WR) |
+		FW_WR_COMPL(1) |
+		FW_WR_IMMDLEN(mpalen));
+	req->flowid_len16 = cpu_to_be32(
+		FW_WR_FLOWID(ep->hwtid) |
+		FW_WR_LEN16(wrlen >> 4));
+	req->plen = cpu_to_be32(mpalen);
+	req->tunnel_to_proxy = cpu_to_be32(
+		FW_OFLD_TX_DATA_WR_FLUSH(1) |
+		FW_OFLD_TX_DATA_WR_SHOVE(1));
+
+	mpa = (struct mpa_message *)(req + 1);
+	memcpy(mpa->key, MPA_KEY_REQ, sizeof(mpa->key));
+	mpa->flags = (crc_enabled ? MPA_CRC : 0) |
+		     (markers_enabled ? MPA_MARKERS : 0);
+	mpa->private_data_size = htons(ep->plen);
+	mpa->revision = mpa_rev;
+
+	if (ep->plen)
+		memcpy(mpa->private_data, ep->mpa_pkt + sizeof(*mpa), ep->plen);
+
+	/*
+	 * Reference the mpa skb.  This ensures the data area
+	 * will remain in memory until the hw acks the tx.
+	 * Function fw4_ack() will deref it.
+	 */
+	skb_get(skb);
+	t4_set_arp_err_handler(skb, NULL, arp_failure_discard);
+	BUG_ON(ep->mpa_skb);
+	ep->mpa_skb = skb;
+	c4iw_l2t_send(&ep->com.dev->rdev, skb, ep->l2t);
+	start_ep_timer(ep);
+	state_set(&ep->com, MPA_REQ_SENT);
+	ep->mpa_attr.initiator = 1;
+	return;
+}
+
+static int send_mpa_reject(struct c4iw_ep *ep, const void *pdata, u8 plen)
+{
+	int mpalen, wrlen;
+	struct fw_ofld_tx_data_wr *req;
+	struct mpa_message *mpa;
+	struct sk_buff *skb;
+
+	PDBG("%s ep %p tid %u pd_len %d\n", __func__, ep, ep->hwtid, ep->plen);
+
+	mpalen = sizeof(*mpa) + plen;
+	wrlen = roundup(mpalen + sizeof *req, 16);
+
+	skb = get_skb(NULL, wrlen, GFP_KERNEL);
+	if (!skb) {
+		printk(KERN_ERR MOD "%s - cannot alloc skb!\n", __func__);
+		return -ENOMEM;
+	}
+	set_wr_txq(skb, CPL_PRIORITY_DATA, ep->txq_idx);
+
+	req = (struct fw_ofld_tx_data_wr *)skb_put(skb, wrlen);
+	memset(req, 0, wrlen);
+	req->op_to_immdlen = cpu_to_be32(
+		FW_WR_OP(FW_OFLD_TX_DATA_WR) |
+		FW_WR_COMPL(1) |
+		FW_WR_IMMDLEN(mpalen));
+	req->flowid_len16 = cpu_to_be32(
+		FW_WR_FLOWID(ep->hwtid) |
+		FW_WR_LEN16(wrlen >> 4));
+	req->plen = cpu_to_be32(mpalen);
+	req->tunnel_to_proxy = cpu_to_be32(
+		FW_OFLD_TX_DATA_WR_FLUSH(1) |
+		FW_OFLD_TX_DATA_WR_SHOVE(1));
+
+	mpa = (struct mpa_message *)(req + 1);
+	memset(mpa, 0, sizeof(*mpa));
+	memcpy(mpa->key, MPA_KEY_REP, sizeof(mpa->key));
+	mpa->flags = MPA_REJECT;
+	mpa->revision = mpa_rev;
+	mpa->private_data_size = htons(plen);
+	if (plen)
+		memcpy(mpa->private_data, pdata, plen);
+
+	/*
+	 * Reference the mpa skb again.  This ensures the data area
+	 * will remain in memory until the hw acks the tx.
+	 * Function fw4_ack() will deref it.
+	 */
+	skb_get(skb);
+	set_wr_txq(skb, CPL_PRIORITY_DATA, ep->txq_idx);
+	t4_set_arp_err_handler(skb, NULL, arp_failure_discard);
+	BUG_ON(ep->mpa_skb);
+	ep->mpa_skb = skb;
+	return c4iw_l2t_send(&ep->com.dev->rdev, skb, ep->l2t);
+}
+
+static int send_mpa_reply(struct c4iw_ep *ep, const void *pdata, u8 plen)
+{
+	int mpalen, wrlen;
+	struct fw_ofld_tx_data_wr *req;
+	struct mpa_message *mpa;
+	struct sk_buff *skb;
+
+	PDBG("%s ep %p tid %u pd_len %d\n", __func__, ep, ep->hwtid, ep->plen);
+
+	mpalen = sizeof(*mpa) + plen;
+	wrlen = roundup(mpalen + sizeof *req, 16);
+
+	skb = get_skb(NULL, wrlen, GFP_KERNEL);
+	if (!skb) {
+		printk(KERN_ERR MOD "%s - cannot alloc skb!\n", __func__);
+		return -ENOMEM;
+	}
+	set_wr_txq(skb, CPL_PRIORITY_DATA, ep->txq_idx);
+
+	req = (struct fw_ofld_tx_data_wr *) skb_put(skb, wrlen);
+	memset(req, 0, wrlen);
+	req->op_to_immdlen = cpu_to_be32(
+		FW_WR_OP(FW_OFLD_TX_DATA_WR) |
+		FW_WR_COMPL(1) |
+		FW_WR_IMMDLEN(mpalen));
+	req->flowid_len16 = cpu_to_be32(
+		FW_WR_FLOWID(ep->hwtid) |
+		FW_WR_LEN16(wrlen >> 4));
+	req->plen = cpu_to_be32(mpalen);
+	req->tunnel_to_proxy = cpu_to_be32(
+		FW_OFLD_TX_DATA_WR_FLUSH(1) |
+		FW_OFLD_TX_DATA_WR_SHOVE(1));
+
+	mpa = (struct mpa_message *)(req + 1);
+	memset(mpa, 0, sizeof(*mpa));
+	memcpy(mpa->key, MPA_KEY_REP, sizeof(mpa->key));
+	mpa->flags = (ep->mpa_attr.crc_enabled ? MPA_CRC : 0) |
+		     (markers_enabled ? MPA_MARKERS : 0);
+	mpa->revision = mpa_rev;
+	mpa->private_data_size = htons(plen);
+	if (plen)
+		memcpy(mpa->private_data, pdata, plen);
+
+	/*
+	 * Reference the mpa skb.  This ensures the data area
+	 * will remain in memory until the hw acks the tx.
+	 * Function fw4_ack() will deref it.
+	 */
+	skb_get(skb);
+	t4_set_arp_err_handler(skb, NULL, arp_failure_discard);
+	ep->mpa_skb = skb;
+	state_set(&ep->com, MPA_REP_SENT);
+	return c4iw_l2t_send(&ep->com.dev->rdev, skb, ep->l2t);
+}
+
+static int act_establish(struct c4iw_dev *dev, struct sk_buff *skb)
+{
+	struct c4iw_ep *ep;
+	struct cpl_act_establish *req = cplhdr(skb);
+	unsigned int tid = GET_TID(req);
+	unsigned int atid = GET_TID_TID(ntohl(req->tos_atid));
+	struct tid_info *t = dev->rdev.lldi.tids;
+
+	ep = lookup_atid(t, atid);
+
+	PDBG("%s ep %p tid %u snd_isn %u rcv_isn %u\n", __func__, ep, tid,
+	     be32_to_cpu(req->snd_isn), be32_to_cpu(req->rcv_isn));
+
+	dst_confirm(ep->dst);
+
+	/* setup the hwtid for this connection */
+	ep->hwtid = tid;
+	cxgb4_insert_tid(t, ep, tid);
+
+	ep->snd_seq = be32_to_cpu(req->snd_isn);
+	ep->rcv_seq = be32_to_cpu(req->rcv_isn);
+
+	set_emss(ep, ntohs(req->tcp_opt));
+
+	/* dealloc the atid */
+	cxgb4_free_atid(t, atid);
+
+	/* start MPA negotiation */
+	send_flowc(ep, NULL);
+	send_mpa_req(ep, skb);
+
+	return 0;
+}
+
+static void close_complete_upcall(struct c4iw_ep *ep)
+{
+	struct iw_cm_event event;
+
+	PDBG("%s ep %p tid %u\n", __func__, ep, ep->hwtid);
+	memset(&event, 0, sizeof(event));
+	event.event = IW_CM_EVENT_CLOSE;
+	if (ep->com.cm_id) {
+		PDBG("close complete delivered ep %p cm_id %p tid %u\n",
+		     ep, ep->com.cm_id, ep->hwtid);
+		ep->com.cm_id->event_handler(ep->com.cm_id, &event);
+		ep->com.cm_id->rem_ref(ep->com.cm_id);
+		ep->com.cm_id = NULL;
+		ep->com.qp = NULL;
+	}
+}
+
+static int abort_connection(struct c4iw_ep *ep, struct sk_buff *skb, gfp_t gfp)
+{
+	PDBG("%s ep %p tid %u\n", __func__, ep, ep->hwtid);
+	close_complete_upcall(ep);
+	state_set(&ep->com, ABORTING);
+	return send_abort(ep, skb, gfp);
+}
+
+static void peer_close_upcall(struct c4iw_ep *ep)
+{
+	struct iw_cm_event event;
+
+	PDBG("%s ep %p tid %u\n", __func__, ep, ep->hwtid);
+	memset(&event, 0, sizeof(event));
+	event.event = IW_CM_EVENT_DISCONNECT;
+	if (ep->com.cm_id) {
+		PDBG("peer close delivered ep %p cm_id %p tid %u\n",
+		     ep, ep->com.cm_id, ep->hwtid);
+		ep->com.cm_id->event_handler(ep->com.cm_id, &event);
+	}
+}
+
+static void peer_abort_upcall(struct c4iw_ep *ep)
+{
+	struct iw_cm_event event;
+
+	PDBG("%s ep %p tid %u\n", __func__, ep, ep->hwtid);
+	memset(&event, 0, sizeof(event));
+	event.event = IW_CM_EVENT_CLOSE;
+	event.status = -ECONNRESET;
+	if (ep->com.cm_id) {
+		PDBG("abort delivered ep %p cm_id %p tid %u\n", ep,
+		     ep->com.cm_id, ep->hwtid);
+		ep->com.cm_id->event_handler(ep->com.cm_id, &event);
+		ep->com.cm_id->rem_ref(ep->com.cm_id);
+		ep->com.cm_id = NULL;
+		ep->com.qp = NULL;
+	}
+}
+
+static void connect_reply_upcall(struct c4iw_ep *ep, int status)
+{
+	struct iw_cm_event event;
+
+	PDBG("%s ep %p tid %u status %d\n", __func__, ep, ep->hwtid, status);
+	memset(&event, 0, sizeof(event));
+	event.event = IW_CM_EVENT_CONNECT_REPLY;
+	event.status = status;
+	event.local_addr = ep->com.local_addr;
+	event.remote_addr = ep->com.remote_addr;
+
+	if ((status == 0) || (status == -ECONNREFUSED)) {
+		event.private_data_len = ep->plen;
+		event.private_data = ep->mpa_pkt + sizeof(struct mpa_message);
+	}
+	if (ep->com.cm_id) {
+		PDBG("%s ep %p tid %u status %d\n", __func__, ep,
+		     ep->hwtid, status);
+		ep->com.cm_id->event_handler(ep->com.cm_id, &event);
+	}
+	if (status < 0) {
+		ep->com.cm_id->rem_ref(ep->com.cm_id);
+		ep->com.cm_id = NULL;
+		ep->com.qp = NULL;
+	}
+}
+
+static void connect_request_upcall(struct c4iw_ep *ep)
+{
+	struct iw_cm_event event;
+
+	PDBG("%s ep %p tid %u\n", __func__, ep, ep->hwtid);
+	memset(&event, 0, sizeof(event));
+	event.event = IW_CM_EVENT_CONNECT_REQUEST;
+	event.local_addr = ep->com.local_addr;
+	event.remote_addr = ep->com.remote_addr;
+	event.private_data_len = ep->plen;
+	event.private_data = ep->mpa_pkt + sizeof(struct mpa_message);
+	event.provider_data = ep;
+	if (state_read(&ep->parent_ep->com) != DEAD) {
+		c4iw_get_ep(&ep->com);
+		ep->parent_ep->com.cm_id->event_handler(
+						ep->parent_ep->com.cm_id,
+						&event);
+	}
+	c4iw_put_ep(&ep->parent_ep->com);
+	ep->parent_ep = NULL;
+}
+
+static void established_upcall(struct c4iw_ep *ep)
+{
+	struct iw_cm_event event;
+
+	PDBG("%s ep %p tid %u\n", __func__, ep, ep->hwtid);
+	memset(&event, 0, sizeof(event));
+	event.event = IW_CM_EVENT_ESTABLISHED;
+	if (ep->com.cm_id) {
+		PDBG("%s ep %p tid %u\n", __func__, ep, ep->hwtid);
+		ep->com.cm_id->event_handler(ep->com.cm_id, &event);
+	}
+}
+
+static int update_rx_credits(struct c4iw_ep *ep, u32 credits)
+{
+	struct cpl_rx_data_ack *req;
+	struct sk_buff *skb;
+	int wrlen = roundup(sizeof *req, 16);
+
+	PDBG("%s ep %p tid %u credits %u\n", __func__, ep, ep->hwtid, credits);
+	skb = get_skb(NULL, wrlen, GFP_KERNEL);
+	if (!skb) {
+		printk(KERN_ERR MOD "update_rx_credits - cannot alloc skb!\n");
+		return 0;
+	}
+
+	req = (struct cpl_rx_data_ack *) skb_put(skb, wrlen);
+	memset(req, 0, wrlen);
+	INIT_TP_WR(req, ep->hwtid);
+	OPCODE_TID(req) = cpu_to_be32(MK_OPCODE_TID(CPL_RX_DATA_ACK,
+						    ep->hwtid));
+	req->credit_dack = cpu_to_be32(credits);
+	set_wr_txq(skb, CPL_PRIORITY_ACK, ep->txq_idx);
+	c4iw_ofld_send(&ep->com.dev->rdev, skb);
+	return credits;
+}
+
+static void process_mpa_reply(struct c4iw_ep *ep, struct sk_buff *skb)
+{
+	struct mpa_message *mpa;
+	u16 plen;
+	struct c4iw_qp_attributes attrs;
+	enum c4iw_qp_attr_mask mask;
+	int err;
+
+	PDBG("%s ep %p tid %u\n", __func__, ep, ep->hwtid);
+
+	/*
+	 * Stop mpa timer.  If it expired, then the state has
+	 * changed and we bail since ep_timeout already aborted
+	 * the connection.
+	 */
+	stop_ep_timer(ep);
+	if (state_read(&ep->com) != MPA_REQ_SENT)
+		return;
+
+	/*
+	 * If we get more than the supported amount of private data
+	 * then we must fail this connection.
+	 */
+	if (ep->mpa_pkt_len + skb->len > sizeof(ep->mpa_pkt)) {
+		err = -EINVAL;
+		goto err;
+	}
+
+	/*
+	 * copy the new data into our accumulation buffer.
+	 */
+	skb_copy_from_linear_data(skb, &(ep->mpa_pkt[ep->mpa_pkt_len]),
+				  skb->len);
+	ep->mpa_pkt_len += skb->len;
+
+	/*
+	 * if we don't even have the mpa message, then bail.
+	 */
+	if (ep->mpa_pkt_len < sizeof(*mpa))
+		return;
+	mpa = (struct mpa_message *) ep->mpa_pkt;
+
+	/* Validate MPA header. */
+	if (mpa->revision != mpa_rev) {
+		err = -EPROTO;
+		goto err;
+	}
+	if (memcmp(mpa->key, MPA_KEY_REP, sizeof(mpa->key))) {
+		err = -EPROTO;
+		goto err;
+	}
+
+	plen = ntohs(mpa->private_data_size);
+
+	/*
+	 * Fail if there's too much private data.
+	 */