Primary 4.7 merge window changes

- Updates to the new Intel X722 iWARP driver
 - Updates to the hfi1 driver
 - Fixes for the iw_cxgb4 driver
 - Misc core fixes
 - Generic RDMA READ/WRITE API addition
 - SRP updates
 - Misc ipoib updates
 - Minor mlx5 updates
 -----BEGIN PGP SIGNATURE-----
 Version: GnuPG v1
 
 iQIcBAABAgAGBQJXPIAnAAoJELgmozMOVy/dEYUP/0A6NH5ptzUrwPuLrWoz8h+e
 KBfJE7H09mfKBx0Rq8YmnU+pz4lk8vrMLLaqpbGN57mwO0a1lK9bgc3E6KUhQPhc
 dpGEX/NG1+aILomD7M4l1yAKkG17kxFLD75cLCeaxhO76jBRWsunukqk5mT/u0EG
 fUYZs1fRb9t2LDtTNhPfSFR1+dgP5S17xLhpl9ttn87hTmIuiGWR6ig2nTC7azZD
 G0d7RVjohfY2sDD28YgQiUEJ+q+1ymp3XTaCZhPCVl9VCRPweEdtLKcbNWZIvClx
 ewuTCgADXg8tAL/6zu6bEKqZlC17UrmVJee3csKQLT09PJkSEICeFR/ld6ONUKAF
 nDhi3ySa75Xxg9VDcCnuRkXKK+/zi7oDelZuh9mvMG0JJqPK9rTZDD29j2kBf7C0
 bdx4R5cI4KJWQ/GlCyi/nLiuYkmAiCugzcGnRho4ub+EJ0yX1w6n8KVYr37kFsFu
 q6MCnEfArEgDpbq1wo0+9MWtqBYrnOI/XtG81Zd+6X2MW975qU85wUdUSjg6OOb1
 v1osyAmFDy9A0Y80yY+l1HHrSVIvI0IAWZDfxsbCLQY8O03ZNcvxE2RsrzWd5CKL
 iZsX24tjV0WR9+lORHLfAKB3DL9CcfHv/tHo7q+5iAHmIuWZGrEN22ELkwS/4X7x
 d/V0XDjzs6lgQeTJ7R4B
 =e+Zv
 -----END PGP SIGNATURE-----

Merge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma

Pull rdma updates from Doug Ledford:
 "Primary 4.7 merge window changes

   - Updates to the new Intel X722 iWARP driver
   - Updates to the hfi1 driver
   - Fixes for the iw_cxgb4 driver
   - Misc core fixes
   - Generic RDMA READ/WRITE API addition
   - SRP updates
   - Misc ipoib updates
   - Minor mlx5 updates"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma: (148 commits)
  IB/mlx5: Fire the CQ completion handler from tasklet
  net/mlx5_core: Use tasklet for user-space CQ completion events
  IB/core: Do not require CAP_NET_ADMIN for packet sniffing
  IB/mlx4: Fix unaligned access in send_reply_to_slave
  IB/mlx5: Report Scatter FCS device capability when supported
  IB/mlx5: Add Scatter FCS support for Raw Packet QP
  IB/core: Add Scatter FCS create flag
  IB/core: Add Raw Scatter FCS device capability
  IB/core: Add extended device capability flags
  i40iw: pass hw_stats by reference rather than by value
  i40iw: Remove unnecessary synchronize_irq() before free_irq()
  i40iw: constify i40iw_vf_cqp_ops structure
  IB/mlx5: Add UARs write-combining and non-cached mapping
  IB/mlx5: Allow mapping the free running counter on PROT_EXEC
  IB/mlx4: Use list_for_each_entry_safe
  IB/SA: Use correct free function
  IB/core: Fix a potential array overrun in CMA and SA agent
  IB/core: Remove unnecessary check in ibnl_rcv_msg
  IB/IWPM: Fix a potential skb leak
  RDMA/nes: replace custom print_hex_dump()
  ...

drivers/infiniband/core/Makefile

@@ -8,9 +8,9 @@ obj-$(CONFIG_INFINIBAND_USER_MAD) +=	ib_umad.o
 obj-$(CONFIG_INFINIBAND_USER_ACCESS) +=	ib_uverbs.o ib_ucm.o \
 					$(user_access-y)
 
-ib_core-y :=			packer.o ud_header.o verbs.o cq.o sysfs.o \
+ib_core-y :=			packer.o ud_header.o verbs.o cq.o rw.o sysfs.o \
 				device.o fmr_pool.o cache.o netlink.o \
-				roce_gid_mgmt.o
+				roce_gid_mgmt.o mr_pool.o
 ib_core-$(CONFIG_INFINIBAND_USER_MEM) += umem.o
 ib_core-$(CONFIG_INFINIBAND_ON_DEMAND_PAGING) += umem_odp.o umem_rbtree.o
 

drivers/infiniband/core/cma.c

@@ -800,6 +800,7 @@ int rdma_create_qp(struct rdma_cm_id *id, struct ib_pd *pd,
 	if (id->device != pd->device)
 		return -EINVAL;
 
+	qp_init_attr->port_num = id->port_num;
 	qp = ib_create_qp(pd, qp_init_attr);
 	if (IS_ERR(qp))
 		return PTR_ERR(qp);
@@ -4294,7 +4295,8 @@ static int __init cma_init(void)
 	if (ret)
 		goto err;
 
-	if (ibnl_add_client(RDMA_NL_RDMA_CM, RDMA_NL_RDMA_CM_NUM_OPS, cma_cb_table))
+	if (ibnl_add_client(RDMA_NL_RDMA_CM, ARRAY_SIZE(cma_cb_table),
+			    cma_cb_table))
 		pr_warn("RDMA CMA: failed to add netlink callback\n");
 	cma_configfs_init();
 

drivers/infiniband/core/iwcm.c

@@ -459,7 +459,7 @@ static void iw_cm_check_wildcard(struct sockaddr_storage *pm_addr,
 	if (pm_addr->ss_family == AF_INET) {
 		struct sockaddr_in *pm4_addr = (struct sockaddr_in *)pm_addr;
 
-		if (pm4_addr->sin_addr.s_addr == INADDR_ANY) {
+		if (pm4_addr->sin_addr.s_addr == htonl(INADDR_ANY)) {
 			struct sockaddr_in *cm4_addr =
 				(struct sockaddr_in *)cm_addr;
 			struct sockaddr_in *cm4_outaddr =
@@ -1175,7 +1175,7 @@ static int __init iw_cm_init(void)
 	if (ret)
 		pr_err("iw_cm: couldn't init iwpm\n");
 
-	ret = ibnl_add_client(RDMA_NL_IWCM, RDMA_NL_IWPM_NUM_OPS,
+	ret = ibnl_add_client(RDMA_NL_IWCM, ARRAY_SIZE(iwcm_nl_cb_table),
 			      iwcm_nl_cb_table);
 	if (ret)
 		pr_err("iw_cm: couldn't register netlink callbacks\n");

drivers/infiniband/core/iwpm_util.c

@@ -634,6 +634,7 @@ static int send_nlmsg_done(struct sk_buff *skb, u8 nl_client, int iwpm_pid)
 	if (!(ibnl_put_msg(skb, &nlh, 0, 0, nl_client,
 			   RDMA_NL_IWPM_MAPINFO, NLM_F_MULTI))) {
 		pr_warn("%s Unable to put NLMSG_DONE\n", __func__);
+		dev_kfree_skb(skb);
 		return -ENOMEM;
 	}
 	nlh->nlmsg_type = NLMSG_DONE;

drivers/infiniband/core/mr_pool.c

@@ -0,0 +1,86 @@
+/*
+ * Copyright (c) 2016 HGST, a Western Digital Company.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ */
+#include <rdma/ib_verbs.h>
+#include <rdma/mr_pool.h>
+
+struct ib_mr *ib_mr_pool_get(struct ib_qp *qp, struct list_head *list)
+{
+	struct ib_mr *mr;
+	unsigned long flags;
+
+	spin_lock_irqsave(&qp->mr_lock, flags);
+	mr = list_first_entry_or_null(list, struct ib_mr, qp_entry);
+	if (mr) {
+		list_del(&mr->qp_entry);
+		qp->mrs_used++;
+	}
+	spin_unlock_irqrestore(&qp->mr_lock, flags);
+
+	return mr;
+}
+EXPORT_SYMBOL(ib_mr_pool_get);
+
+void ib_mr_pool_put(struct ib_qp *qp, struct list_head *list, struct ib_mr *mr)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&qp->mr_lock, flags);
+	list_add(&mr->qp_entry, list);
+	qp->mrs_used--;
+	spin_unlock_irqrestore(&qp->mr_lock, flags);
+}
+EXPORT_SYMBOL(ib_mr_pool_put);
+
+int ib_mr_pool_init(struct ib_qp *qp, struct list_head *list, int nr,
+		enum ib_mr_type type, u32 max_num_sg)
+{
+	struct ib_mr *mr;
+	unsigned long flags;
+	int ret, i;
+
+	for (i = 0; i < nr; i++) {
+		mr = ib_alloc_mr(qp->pd, type, max_num_sg);
+		if (IS_ERR(mr)) {
+			ret = PTR_ERR(mr);
+			goto out;
+		}
+
+		spin_lock_irqsave(&qp->mr_lock, flags);
+		list_add_tail(&mr->qp_entry, list);
+		spin_unlock_irqrestore(&qp->mr_lock, flags);
+	}
+
+	return 0;
+out:
+	ib_mr_pool_destroy(qp, list);
+	return ret;
+}
+EXPORT_SYMBOL(ib_mr_pool_init);
+
+void ib_mr_pool_destroy(struct ib_qp *qp, struct list_head *list)
+{
+	struct ib_mr *mr;
+	unsigned long flags;
+
+	spin_lock_irqsave(&qp->mr_lock, flags);
+	while (!list_empty(list)) {
+		mr = list_first_entry(list, struct ib_mr, qp_entry);
+		list_del(&mr->qp_entry);
+
+		spin_unlock_irqrestore(&qp->mr_lock, flags);
+		ib_dereg_mr(mr);
+		spin_lock_irqsave(&qp->mr_lock, flags);
+	}
+	spin_unlock_irqrestore(&qp->mr_lock, flags);
+}
+EXPORT_SYMBOL(ib_mr_pool_destroy);

drivers/infiniband/core/netlink.c

@@ -151,12 +151,11 @@ static int ibnl_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
 	struct ibnl_client *client;
 	int type = nlh->nlmsg_type;
 	int index = RDMA_NL_GET_CLIENT(type);
-	int op = RDMA_NL_GET_OP(type);
+	unsigned int op = RDMA_NL_GET_OP(type);
 
 	list_for_each_entry(client, &client_list, list) {
 		if (client->index == index) {
-			if (op < 0 || op >= client->nops ||
-			    !client->cb_table[op].dump)
+			if (op >= client->nops || !client->cb_table[op].dump)
 				return -EINVAL;
 
 			/*

drivers/infiniband/core/rw.c

@@ -0,0 +1,727 @@
+/*
+ * Copyright (c) 2016 HGST, a Western Digital Company.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ */
+#include <linux/moduleparam.h>
+#include <linux/slab.h>
+#include <rdma/mr_pool.h>
+#include <rdma/rw.h>
+
+enum {
+	RDMA_RW_SINGLE_WR,
+	RDMA_RW_MULTI_WR,
+	RDMA_RW_MR,
+	RDMA_RW_SIG_MR,
+};
+
+static bool rdma_rw_force_mr;
+module_param_named(force_mr, rdma_rw_force_mr, bool, 0);
+MODULE_PARM_DESC(force_mr, "Force usage of MRs for RDMA READ/WRITE operations");
+
+/*
+ * Check if the device might use memory registration.  This is currently only
+ * true for iWarp devices. In the future we can hopefully fine tune this based
+ * on HCA driver input.
+ */
+static inline bool rdma_rw_can_use_mr(struct ib_device *dev, u8 port_num)
+{
+	if (rdma_protocol_iwarp(dev, port_num))
+		return true;
+	if (unlikely(rdma_rw_force_mr))
+		return true;
+	return false;
+}
+
+/*
+ * Check if the device will use memory registration for this RW operation.
+ * We currently always use memory registrations for iWarp RDMA READs, and
+ * have a debug option to force usage of MRs.
+ *
+ * XXX: In the future we can hopefully fine tune this based on HCA driver
+ * input.
+ */
+static inline bool rdma_rw_io_needs_mr(struct ib_device *dev, u8 port_num,
+		enum dma_data_direction dir, int dma_nents)
+{
+	if (rdma_protocol_iwarp(dev, port_num) && dir == DMA_FROM_DEVICE)
+		return true;
+	if (unlikely(rdma_rw_force_mr))
+		return true;
+	return false;
+}
+
+static inline u32 rdma_rw_max_sge(struct ib_device *dev,
+		enum dma_data_direction dir)
+{
+	return dir == DMA_TO_DEVICE ?
+		dev->attrs.max_sge : dev->attrs.max_sge_rd;
+}
+
+static inline u32 rdma_rw_fr_page_list_len(struct ib_device *dev)
+{
+	/* arbitrary limit to avoid allocating gigantic resources */
+	return min_t(u32, dev->attrs.max_fast_reg_page_list_len, 256);
+}
+
+static int rdma_rw_init_one_mr(struct ib_qp *qp, u8 port_num,
+		struct rdma_rw_reg_ctx *reg, struct scatterlist *sg,
+		u32 sg_cnt, u32 offset)
+{
+	u32 pages_per_mr = rdma_rw_fr_page_list_len(qp->pd->device);
+	u32 nents = min(sg_cnt, pages_per_mr);
+	int count = 0, ret;
+
+	reg->mr = ib_mr_pool_get(qp, &qp->rdma_mrs);
+	if (!reg->mr)
+		return -EAGAIN;
+
+	if (reg->mr->need_inval) {
+		reg->inv_wr.opcode = IB_WR_LOCAL_INV;
+		reg->inv_wr.ex.invalidate_rkey = reg->mr->lkey;
+		reg->inv_wr.next = &reg->reg_wr.wr;
+		count++;
+	} else {
+		reg->inv_wr.next = NULL;
+	}
+
+	ret = ib_map_mr_sg(reg->mr, sg, nents, &offset, PAGE_SIZE);
+	if (ret < nents) {
+		ib_mr_pool_put(qp, &qp->rdma_mrs, reg->mr);
+		return -EINVAL;
+	}
+
+	reg->reg_wr.wr.opcode = IB_WR_REG_MR;
+	reg->reg_wr.mr = reg->mr;
+	reg->reg_wr.access = IB_ACCESS_LOCAL_WRITE;
+	if (rdma_protocol_iwarp(qp->device, port_num))
+		reg->reg_wr.access |= IB_ACCESS_REMOTE_WRITE;
+	count++;
+
+	reg->sge.addr = reg->mr->iova;
+	reg->sge.length = reg->mr->length;
+	return count;
+}
+
+static int rdma_rw_init_mr_wrs(struct rdma_rw_ctx *ctx, struct ib_qp *qp,
+		u8 port_num, struct scatterlist *sg, u32 sg_cnt, u32 offset,
+		u64 remote_addr, u32 rkey, enum dma_data_direction dir)
+{
+	u32 pages_per_mr = rdma_rw_fr_page_list_len(qp->pd->device);
+	int i, j, ret = 0, count = 0;
+
+	ctx->nr_ops = (sg_cnt + pages_per_mr - 1) / pages_per_mr;
+	ctx->reg = kcalloc(ctx->nr_ops, sizeof(*ctx->reg), GFP_KERNEL);
+	if (!ctx->reg) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	for (i = 0; i < ctx->nr_ops; i++) {
+		struct rdma_rw_reg_ctx *prev = i ? &ctx->reg[i - 1] : NULL;
+		struct rdma_rw_reg_ctx *reg = &ctx->reg[i];
+		u32 nents = min(sg_cnt, pages_per_mr);
+
+		ret = rdma_rw_init_one_mr(qp, port_num, reg, sg, sg_cnt,
+				offset);
+		if (ret < 0)
+			goto out_free;
+		count += ret;
+
+		if (prev) {
+			if (reg->mr->need_inval)
+				prev->wr.wr.next = &reg->inv_wr;
+			else
+				prev->wr.wr.next = &reg->reg_wr.wr;
+		}
+
+		reg->reg_wr.wr.next = &reg->wr.wr;
+
+		reg->wr.wr.sg_list = &reg->sge;
+		reg->wr.wr.num_sge = 1;
+		reg->wr.remote_addr = remote_addr;
+		reg->wr.rkey = rkey;
+		if (dir == DMA_TO_DEVICE) {
+			reg->wr.wr.opcode = IB_WR_RDMA_WRITE;
+		} else if (!rdma_cap_read_inv(qp->device, port_num)) {
+			reg->wr.wr.opcode = IB_WR_RDMA_READ;
+		} else {
+			reg->wr.wr.opcode = IB_WR_RDMA_READ_WITH_INV;
+			reg->wr.wr.ex.invalidate_rkey = reg->mr->lkey;
+		}
+		count++;
+
+		remote_addr += reg->sge.length;
+		sg_cnt -= nents;
+		for (j = 0; j < nents; j++)
+			sg = sg_next(sg);
+		offset = 0;
+	}
+
+	ctx->type = RDMA_RW_MR;
+	return count;
+
+out_free:
+	while (--i >= 0)
+		ib_mr_pool_put(qp, &qp->rdma_mrs, ctx->reg[i].mr);
+	kfree(ctx->reg);
+out:
+	return ret;
+}
+
+static int rdma_rw_init_map_wrs(struct rdma_rw_ctx *ctx, struct ib_qp *qp,
+		struct scatterlist *sg, u32 sg_cnt, u32 offset,
+		u64 remote_addr, u32 rkey, enum dma_data_direction dir)
+{
+	struct ib_device *dev = qp->pd->device;
+	u32 max_sge = rdma_rw_max_sge(dev, dir);
+	struct ib_sge *sge;
+	u32 total_len = 0, i, j;
+
+	ctx->nr_ops = DIV_ROUND_UP(sg_cnt, max_sge);
+
+	ctx->map.sges = sge = kcalloc(sg_cnt, sizeof(*sge), GFP_KERNEL);
+	if (!ctx->map.sges)
+		goto out;
+
+	ctx->map.wrs = kcalloc(ctx->nr_ops, sizeof(*ctx->map.wrs), GFP_KERNEL);
+	if (!ctx->map.wrs)
+		goto out_free_sges;
+
+	for (i = 0; i < ctx->nr_ops; i++) {
+		struct ib_rdma_wr *rdma_wr = &ctx->map.wrs[i];
+		u32 nr_sge = min(sg_cnt, max_sge);
+
+		if (dir == DMA_TO_DEVICE)
+			rdma_wr->wr.opcode = IB_WR_RDMA_WRITE;
+		else
+			rdma_wr->wr.opcode = IB_WR_RDMA_READ;
+		rdma_wr->remote_addr = remote_addr + total_len;
+		rdma_wr->rkey = rkey;
+		rdma_wr->wr.sg_list = sge;
+
+		for (j = 0; j < nr_sge; j++, sg = sg_next(sg)) {
+			rdma_wr->wr.num_sge++;
+
+			sge->addr = ib_sg_dma_address(dev, sg) + offset;
+			sge->length = ib_sg_dma_len(dev, sg) - offset;
+			sge->lkey = qp->pd->local_dma_lkey;
+
+			total_len += sge->length;
+			sge++;
+			sg_cnt--;
+			offset = 0;
+		}
+
+		if (i + 1 < ctx->nr_ops)
+			rdma_wr->wr.next = &ctx->map.wrs[i + 1].wr;
+	}
+
+	ctx->type = RDMA_RW_MULTI_WR;
+	return ctx->nr_ops;
+
+out_free_sges:
+	kfree(ctx->map.sges);
+out:
+	return -ENOMEM;
+}
+
+static int rdma_rw_init_single_wr(struct rdma_rw_ctx *ctx, struct ib_qp *qp,
+		struct scatterlist *sg, u32 offset, u64 remote_addr, u32 rkey,
+		enum dma_data_direction dir)
+{
+	struct ib_device *dev = qp->pd->device;
+	struct ib_rdma_wr *rdma_wr = &ctx->single.wr;
+
+	ctx->nr_ops = 1;
+
+	ctx->single.sge.lkey = qp->pd->local_dma_lkey;
+	ctx->single.sge.addr = ib_sg_dma_address(dev, sg) + offset;
+	ctx->single.sge.length = ib_sg_dma_len(dev, sg) - offset;
+
+	memset(rdma_wr, 0, sizeof(*rdma_wr));
+	if (dir == DMA_TO_DEVICE)
+		rdma_wr->wr.opcode = IB_WR_RDMA_WRITE;
+	else
+		rdma_wr->wr.opcode = IB_WR_RDMA_READ;
+	rdma_wr->wr.sg_list = &ctx->single.sge;
+	rdma_wr->wr.num_sge = 1;
+	rdma_wr->remote_addr = remote_addr;
+	rdma_wr->rkey = rkey;
+
+	ctx->type = RDMA_RW_SINGLE_WR;
+	return 1;
+}
+
+/**
+ * rdma_rw_ctx_init - initialize a RDMA READ/WRITE context
+ * @ctx:	context to initialize
+ * @qp:		queue pair to operate on
+ * @port_num:	port num to which the connection is bound
+ * @sg:		scatterlist to READ/WRITE from/to
+ * @sg_cnt:	number of entries in @sg
+ * @sg_offset:	current byte offset into @sg
+ * @remote_addr:remote address to read/write (relative to @rkey)
+ * @rkey:	remote key to operate on
+ * @dir:	%DMA_TO_DEVICE for RDMA WRITE, %DMA_FROM_DEVICE for RDMA READ
+ *
+ * Returns the number of WQEs that will be needed on the workqueue if
+ * successful, or a negative error code.
+ */
+int rdma_rw_ctx_init(struct rdma_rw_ctx *ctx, struct ib_qp *qp, u8 port_num,
+		struct scatterlist *sg, u32 sg_cnt, u32 sg_offset,
+		u64 remote_addr, u32 rkey, enum dma_data_direction dir)
+{
+	struct ib_device *dev = qp->pd->device;
+	int ret;
+
+	ret = ib_dma_map_sg(dev, sg, sg_cnt, dir);
+	if (!ret)
+		return -ENOMEM;
+	sg_cnt = ret;
+
+	/*
+	 * Skip to the S/G entry that sg_offset falls into:
+	 */
+	for (;;) {
+		u32 len = ib_sg_dma_len(dev, sg);
+
+		if (sg_offset < len)
+			break;
+
+		sg = sg_next(sg);
+		sg_offset -= len;
+		sg_cnt--;
+	}
+
+	ret = -EIO;
+	if (WARN_ON_ONCE(sg_cnt == 0))
+		goto out_unmap_sg;
+
+	if (rdma_rw_io_needs_mr(qp->device, port_num, dir, sg_cnt)) {
+		ret = rdma_rw_init_mr_wrs(ctx, qp, port_num, sg, sg_cnt,
+				sg_offset, remote_addr, rkey, dir);
+	} else if (sg_cnt > 1) {
+		ret = rdma_rw_init_map_wrs(ctx, qp, sg, sg_cnt, sg_offset,
+				remote_addr, rkey, dir);
+	} else {
+		ret = rdma_rw_init_single_wr(ctx, qp, sg, sg_offset,
+				remote_addr, rkey, dir);
+	}
+
+	if (ret < 0)
+		goto out_unmap_sg;
+	return ret;
+
+out_unmap_sg:
+	ib_dma_unmap_sg(dev, sg, sg_cnt, dir);
+	return ret;
+}
+EXPORT_SYMBOL(rdma_rw_ctx_init);
+
+/**
+ * rdma_rw_ctx_signature init - initialize a RW context with signature offload
+ * @ctx:	context to initialize
+ * @qp:		queue pair to operate on
+ * @port_num:	port num to which the connection is bound
+ * @sg:		scatterlist to READ/WRITE from/to
+ * @sg_cnt:	number of entries in @sg
+ * @prot_sg:	scatterlist to READ/WRITE protection information from/to
+ * @prot_sg_cnt: number of entries in @prot_sg
+ * @sig_attrs:	signature offloading algorithms
+ * @remote_addr:remote address to read/write (relative to @rkey)
+ * @rkey:	remote key to operate on
+ * @dir:	%DMA_TO_DEVICE for RDMA WRITE, %DMA_FROM_DEVICE for RDMA READ
+ *
+ * Returns the number of WQEs that will be needed on the workqueue if
+ * successful, or a negative error code.
+ */
+int rdma_rw_ctx_signature_init(struct rdma_rw_ctx *ctx, struct ib_qp *qp,
+		u8 port_num, struct scatterlist *sg, u32 sg_cnt,
+		struct scatterlist *prot_sg, u32 prot_sg_cnt,
+		struct ib_sig_attrs *sig_attrs,
+		u64 remote_addr, u32 rkey, enum dma_data_direction dir)
+{
+	struct ib_device *dev = qp->pd->device;
+	u32 pages_per_mr = rdma_rw_fr_page_list_len(qp->pd->device);
+	struct ib_rdma_wr *rdma_wr;
+	struct ib_send_wr *prev_wr = NULL;
+	int count = 0, ret;
+
+	if (sg_cnt > pages_per_mr || prot_sg_cnt > pages_per_mr) {
+		pr_err("SG count too large\n");
+		return -EINVAL;
+	}
+
+	ret = ib_dma_map_sg(dev, sg, sg_cnt, dir);
+	if (!ret)
+		return -ENOMEM;
+	sg_cnt = ret;
+
+	ret = ib_dma_map_sg(dev, prot_sg, prot_sg_cnt, dir);
+	if (!ret) {
+		ret = -ENOMEM;
+		goto out_unmap_sg;
+	}
+	prot_sg_cnt = ret;
+
+	ctx->type = RDMA_RW_SIG_MR;
+	ctx->nr_ops = 1;
+	ctx->sig = kcalloc(1, sizeof(*ctx->sig), GFP_KERNEL);
+	if (!ctx->sig) {
+		ret = -ENOMEM;
+		goto out_unmap_prot_sg;
+	}
+
+	ret = rdma_rw_init_one_mr(qp, port_num, &ctx->sig->data, sg, sg_cnt, 0);
+	if (ret < 0)
+		goto out_free_ctx;
+	count += ret;
+	prev_wr = &ctx->sig->data.reg_wr.wr;
+
+	if (prot_sg_cnt) {
+		ret = rdma_rw_init_one_mr(qp, port_num, &ctx->sig->prot,
+				prot_sg, prot_sg_cnt, 0);
+		if (ret < 0)
+			goto out_destroy_data_mr;
+		count += ret;
+
+		if (ctx->sig->prot.inv_wr.next)
+			prev_wr->next = &ctx->sig->prot.inv_wr;
+		else
+			prev_wr->next = &ctx->sig->prot.reg_wr.wr;
+		prev_wr = &ctx->sig->prot.reg_wr.wr;
+	} else {
+		ctx->sig->prot.mr = NULL;
+	}
+
+	ctx->sig->sig_mr = ib_mr_pool_get(qp, &qp->sig_mrs);
+	if (!ctx->sig->sig_mr) {
+		ret = -EAGAIN;
+		goto out_destroy_prot_mr;
+	}
+
+	if (ctx->sig->sig_mr->need_inval) {
+		memset(&ctx->sig->sig_inv_wr, 0, sizeof(ctx->sig->sig_inv_wr));
+
+		ctx->sig->sig_inv_wr.opcode = IB_WR_LOCAL_INV;
+		ctx->sig->sig_inv_wr.ex.invalidate_rkey = ctx->sig->sig_mr->rkey;
+
+		prev_wr->next = &ctx->sig->sig_inv_wr;
+		prev_wr = &ctx->sig->sig_inv_wr;
+	}
+
+	ctx->sig->sig_wr.wr.opcode = IB_WR_REG_SIG_MR;
+	ctx->sig->sig_wr.wr.wr_cqe = NULL;
+	ctx->sig->sig_wr.wr.sg_list = &ctx->sig->data.sge;
+	ctx->sig->sig_wr.wr.num_sge = 1;
+	ctx->sig->sig_wr.access_flags = IB_ACCESS_LOCAL_WRITE;
+	ctx->sig->sig_wr.sig_attrs = sig_attrs;
+	ctx->sig->sig_wr.sig_mr = ctx->sig->sig_mr;
+	if (prot_sg_cnt)
+		ctx->sig->sig_wr.prot = &ctx->sig->prot.sge;
+	prev_wr->next = &ctx->sig->sig_wr.wr;
+	prev_wr = &ctx->sig->sig_wr.wr;
+	count++;
+
+	ctx->sig->sig_sge.addr = 0;
+	ctx->sig->sig_sge.length = ctx->sig->data.sge.length;
+	if (sig_attrs->wire.sig_type != IB_SIG_TYPE_NONE)
+		ctx->sig->sig_sge.length += ctx->sig->prot.sge.length;
+
+	rdma_wr = &ctx->sig->data.wr;
+	rdma_wr->wr.sg_list = &ctx->sig->sig_sge;
+	rdma_wr->wr.num_sge = 1;
+	rdma_wr->remote_addr = remote_addr;
+	rdma_wr->rkey = rkey;
+	if (dir == DMA_TO_DEVICE)
+		rdma_wr->wr.opcode = IB_WR_RDMA_WRITE;
+	else
+		rdma_wr->wr.opcode = IB_WR_RDMA_READ;
+	prev_wr->next = &rdma_wr->wr;
+	prev_wr = &rdma_wr->wr;
+	count++;
+
+	return count;
+
+out_destroy_prot_mr:
+	if (prot_sg_cnt)
+		ib_mr_pool_put(qp, &qp->rdma_mrs, ctx->sig->prot.mr);
+out_destroy_data_mr:
+	ib_mr_pool_put(qp, &qp->rdma_mrs, ctx->sig->data.mr);
+out_free_ctx:
+	kfree(ctx->sig);
+out_unmap_prot_sg:
+	ib_dma_unmap_sg(dev, prot_sg, prot_sg_cnt, dir);
+out_unmap_sg:
+	ib_dma_unmap_sg(dev, sg, sg_cnt, dir);
+	return ret;
+}
+EXPORT_SYMBOL(rdma_rw_ctx_signature_init);
+
+/*
+ * Now that we are going to post the WRs we can update the lkey and need_inval
+ * state on the MRs.  If we were doing this at init time, we would get double
+ * or missing invalidations if a context was initialized but not actually
+ * posted.
+ */
+static void rdma_rw_update_lkey(struct rdma_rw_reg_ctx *reg, bool need_inval)
+{
+	reg->mr->need_inval = need_inval;
+	ib_update_fast_reg_key(reg->mr, ib_inc_rkey(reg->mr->lkey));
+	reg->reg_wr.key = reg->mr->lkey;
+	reg->sge.lkey = reg->mr->lkey;
+}
+
+/**
+ * rdma_rw_ctx_wrs - return chain of WRs for a RDMA READ or WRITE operation
+ * @ctx:	context to operate on
+ * @qp:		queue pair to operate on
+ * @port_num:	port num to which the connection is bound
+ * @cqe:	completion queue entry for the last WR
+ * @chain_wr:	WR to append to the posted chain
+ *
+ * Return the WR chain for the set of RDMA READ/WRITE operations described by
+ * @ctx, as well as any memory registration operations needed.  If @chain_wr
+ * is non-NULL the WR it points to will be appended to the chain of WRs posted.
+ * If @chain_wr is not set @cqe must be set so that the caller gets a
+ * completion notification.
+ */
+struct ib_send_wr *rdma_rw_ctx_wrs(struct rdma_rw_ctx *ctx, struct ib_qp *qp,
+		u8 port_num, struct ib_cqe *cqe, struct ib_send_wr *chain_wr)
+{
+	struct ib_send_wr *first_wr, *last_wr;
+	int i;
+
+	switch (ctx->type) {
+	case RDMA_RW_SIG_MR:
+		rdma_rw_update_lkey(&ctx->sig->data, true);
+		if (ctx->sig->prot.mr)
+			rdma_rw_update_lkey(&ctx->sig->prot, true);
+	
+		ctx->sig->sig_mr->need_inval = true;
+		ib_update_fast_reg_key(ctx->sig->sig_mr,
+			ib_inc_rkey(ctx->sig->sig_mr->lkey));
+		ctx->sig->sig_sge.lkey = ctx->sig->sig_mr->lkey;
+
+		if (ctx->sig->data.inv_wr.next)
+			first_wr = &ctx->sig->data.inv_wr;
+		else
+			first_wr = &ctx->sig->data.reg_wr.wr;
+		last_wr = &ctx->sig->data.wr.wr;
+		break;
+	case RDMA_RW_MR:
+		for (i = 0; i < ctx->nr_ops; i++) {
+			rdma_rw_update_lkey(&ctx->reg[i],
+				ctx->reg[i].wr.wr.opcode !=
+					IB_WR_RDMA_READ_WITH_INV);
+		}
+
+		if (ctx->reg[0].inv_wr.next)
+			first_wr = &ctx->reg[0].inv_wr;
+		else
+			first_wr = &ctx->reg[0].reg_wr.wr;
+		last_wr = &ctx->reg[ctx->nr_ops - 1].wr.wr;
+		break;
+	case RDMA_RW_MULTI_WR:
+		first_wr = &ctx->map.wrs[0].wr;
+		last_wr = &ctx->map.wrs[ctx->nr_ops - 1].wr;
+		break;
+	case RDMA_RW_SINGLE_WR:
+		first_wr = &ctx->single.wr.wr;
+		last_wr = &ctx->single.wr.wr;
+		break;
+	default:
+		BUG();
+	}
+
+	if (chain_wr) {
+		last_wr->next = chain_wr;
+	} else {
+		last_wr->wr_cqe = cqe;
+		last_wr->send_flags |= IB_SEND_SIGNALED;
+	}
+
+	return first_wr;
+}
+EXPORT_SYMBOL(rdma_rw_ctx_wrs);
+
+/**
+ * rdma_rw_ctx_post - post a RDMA READ or RDMA WRITE operation
+ * @ctx:	context to operate on
+ * @qp:		queue pair to operate on
+ * @port_num:	port num to which the connection is bound
+ * @cqe:	completion queue entry for the last WR
+ * @chain_wr:	WR to append to the posted chain
+ *
+ * Post the set of RDMA READ/WRITE operations described by @ctx, as well as
+ * any memory registration operations needed.  If @chain_wr is non-NULL the
+ * WR it points to will be appended to the chain of WRs posted.  If @chain_wr
+ * is not set @cqe must be set so that the caller gets a completion
+ * notification.
+ */
+int rdma_rw_ctx_post(struct rdma_rw_ctx *ctx, struct ib_qp *qp, u8 port_num,
+		struct ib_cqe *cqe, struct ib_send_wr *chain_wr)
+{
+	struct ib_send_wr *first_wr, *bad_wr;
+
+	first_wr = rdma_rw_ctx_wrs(ctx, qp, port_num, cqe, chain_wr);
+	return ib_post_send(qp, first_wr, &bad_wr);
+}
+EXPORT_SYMBOL(rdma_rw_ctx_post);
+
+/**
+ * rdma_rw_ctx_destroy - release all resources allocated by rdma_rw_ctx_init
+ * @ctx:	context to release
+ * @qp:		queue pair to operate on
+ * @port_num:	port num to which the connection is bound
+ * @sg:		scatterlist that was used for the READ/WRITE
+ * @sg_cnt:	number of entries in @sg
+ * @dir:	%DMA_TO_DEVICE for RDMA WRITE, %DMA_FROM_DEVICE for RDMA READ
+ */
+void rdma_rw_ctx_destroy(struct rdma_rw_ctx *ctx, struct ib_qp *qp, u8 port_num,
+		struct scatterlist *sg, u32 sg_cnt, enum dma_data_direction dir)
+{
+	int i;
+
+	switch (ctx->type) {
+	case RDMA_RW_MR:
+		for (i = 0; i < ctx->nr_ops; i++)
+			ib_mr_pool_put(qp, &qp->rdma_mrs, ctx->reg[i].mr);
+		kfree(ctx->reg);
+		break;
+	case RDMA_RW_MULTI_WR:
+		kfree(ctx->map.wrs);
+		kfree(ctx->map.sges);
+		break;
+	case RDMA_RW_SINGLE_WR:
+		break;
+	default:
+		BUG();
+		break;
+	}
+
+	ib_dma_unmap_sg(qp->pd->device, sg, sg_cnt, dir);
+}
+EXPORT_SYMBOL(rdma_rw_ctx_destroy);
+
+/**
+ * rdma_rw_ctx_destroy_signature - release all resources allocated by
+ *	rdma_rw_ctx_init_signature
+ * @ctx:	context to release
+ * @qp:		queue pair to operate on
+ * @port_num:	port num to which the connection is bound
+ * @sg:		scatterlist that was used for the READ/WRITE
+ * @sg_cnt:	number of entries in @sg
+ * @prot_sg:	scatterlist that was used for the READ/WRITE of the PI
+ * @prot_sg_cnt: number of entries in @prot_sg
+ * @dir:	%DMA_TO_DEVICE for RDMA WRITE, %DMA_FROM_DEVICE for RDMA READ
+ */
+void rdma_rw_ctx_destroy_signature(struct rdma_rw_ctx *ctx, struct ib_qp *qp,
+		u8 port_num, struct scatterlist *sg, u32 sg_cnt,
+		struct scatterlist *prot_sg, u32 prot_sg_cnt,
+		enum dma_data_direction dir)
+{
+	if (WARN_ON_ONCE(ctx->type != RDMA_RW_SIG_MR))
+		return;
+
+	ib_mr_pool_put(qp, &qp->rdma_mrs, ctx->sig->data.mr);
+	ib_dma_unmap_sg(qp->pd->device, sg, sg_cnt, dir);
+
+	if (ctx->sig->prot.mr) {
+		ib_mr_pool_put(qp, &qp->rdma_mrs, ctx->sig->prot.mr);
+		ib_dma_unmap_sg(qp->pd->device, prot_sg, prot_sg_cnt, dir);
+	}
+
+	ib_mr_pool_put(qp, &qp->sig_mrs, ctx->sig->sig_mr);
+	kfree(ctx->sig);
+}
+EXPORT_SYMBOL(rdma_rw_ctx_destroy_signature);
+
+void rdma_rw_init_qp(struct ib_device *dev, struct ib_qp_init_attr *attr)
+{
+	u32 factor;
+
+	WARN_ON_ONCE(attr->port_num == 0);
+
+	/*
+	 * Each context needs at least one RDMA READ or WRITE WR.
+	 *
+	 * For some hardware we might need more, eventually we should ask the
+	 * HCA driver for a multiplier here.
+	 */
+	factor = 1;
+
+	/*
+	 * If the devices needs MRs to perform RDMA READ or WRITE operations,
+	 * we'll need two additional MRs for the registrations and the
+	 * invalidation.
+	 */
+	if (attr->create_flags & IB_QP_CREATE_SIGNATURE_EN)
+		factor += 6;	/* (inv + reg) * (data + prot + sig) */
+	else if (rdma_rw_can_use_mr(dev, attr->port_num))
+		factor += 2;	/* inv + reg */
+
+	attr->cap.max_send_wr += factor * attr->cap.max_rdma_ctxs;
+
+	/*
+	 * But maybe we were just too high in the sky and the device doesn't
+	 * even support all we need, and we'll have to live with what we get..
+	 */
+	attr->cap.max_send_wr =
+		min_t(u32, attr->cap.max_send_wr, dev->attrs.max_qp_wr);
+}
+
+int rdma_rw_init_mrs(struct ib_qp *qp, struct ib_qp_init_attr *attr)
+{
+	struct ib_device *dev = qp->pd->device;
+	u32 nr_mrs = 0, nr_sig_mrs = 0;
+	int ret = 0;
+
+	if (attr->create_flags & IB_QP_CREATE_SIGNATURE_EN) {
+		nr_sig_mrs = attr->cap.max_rdma_ctxs;
+		nr_mrs = attr->cap.max_rdma_ctxs * 2;
+	} else if (rdma_rw_can_use_mr(dev, attr->port_num)) {
+		nr_mrs = attr->cap.max_rdma_ctxs;
+	}
+
+	if (nr_mrs) {
+		ret = ib_mr_pool_init(qp, &qp->rdma_mrs, nr_mrs,
+				IB_MR_TYPE_MEM_REG,
+				rdma_rw_fr_page_list_len(dev));
+		if (ret) {
+			pr_err("%s: failed to allocated %d MRs\n",
+				__func__, nr_mrs);
+			return ret;
+		}
+	}
+
+	if (nr_sig_mrs) {
+		ret = ib_mr_pool_init(qp, &qp->sig_mrs, nr_sig_mrs,
+				IB_MR_TYPE_SIGNATURE, 2);
+		if (ret) {
+			pr_err("%s: failed to allocated %d SIG MRs\n",
+				__func__, nr_mrs);
+			goto out_free_rdma_mrs;
+		}
+	}
+
+	return 0;
+
+out_free_rdma_mrs:
+	ib_mr_pool_destroy(qp, &qp->rdma_mrs);
+	return ret;
+}
+
+void rdma_rw_cleanup_mrs(struct ib_qp *qp)
+{
+	ib_mr_pool_destroy(qp, &qp->sig_mrs);
+	ib_mr_pool_destroy(qp, &qp->rdma_mrs);
+}

drivers/infiniband/core/sa_query.c

@@ -536,7 +536,7 @@ static int ib_nl_send_msg(struct ib_sa_query *query, gfp_t gfp_mask)
 	data = ibnl_put_msg(skb, &nlh, query->seq, 0, RDMA_NL_LS,
 			    RDMA_NL_LS_OP_RESOLVE, NLM_F_REQUEST);
 	if (!data) {
-		kfree_skb(skb);
+		nlmsg_free(skb);
 		return -EMSGSIZE;
 	}
 
@@ -1820,7 +1820,7 @@ static int __init ib_sa_init(void)
 		goto err3;
 	}
 
-	if (ibnl_add_client(RDMA_NL_LS, RDMA_NL_LS_NUM_OPS,
+	if (ibnl_add_client(RDMA_NL_LS, ARRAY_SIZE(ib_sa_cb_table),
 			    ib_sa_cb_table)) {
 		pr_err("Failed to add netlink callback\n");
 		ret = -EINVAL;

drivers/infiniband/core/uverbs_cmd.c

@@ -1833,7 +1833,8 @@ static int create_qp(struct ib_uverbs_file *file,
 	if (attr.create_flags & ~(IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK |
 				IB_QP_CREATE_CROSS_CHANNEL |
 				IB_QP_CREATE_MANAGED_SEND |
-				IB_QP_CREATE_MANAGED_RECV)) {
+				IB_QP_CREATE_MANAGED_RECV |
+				IB_QP_CREATE_SCATTER_FCS)) {
 		ret = -EINVAL;
 		goto err_put;
 	}
@@ -3088,8 +3089,7 @@ int ib_uverbs_ex_create_flow(struct ib_uverbs_file *file,
 	if (cmd.comp_mask)
 		return -EINVAL;
 
-	if ((cmd.flow_attr.type == IB_FLOW_ATTR_SNIFFER &&
-	     !capable(CAP_NET_ADMIN)) || !capable(CAP_NET_RAW))
+	if (!capable(CAP_NET_RAW))
 		return -EPERM;
 
 	if (cmd.flow_attr.flags >= IB_FLOW_ATTR_FLAGS_RESERVED)
@@ -3655,6 +3655,11 @@ int ib_uverbs_ex_query_device(struct ib_uverbs_file *file,
 	resp.hca_core_clock = attr.hca_core_clock;
 	resp.response_length += sizeof(resp.hca_core_clock);
 
+	if (ucore->outlen < resp.response_length + sizeof(resp.device_cap_flags_ex))
+		goto end;
+
+	resp.device_cap_flags_ex = attr.device_cap_flags;
+	resp.response_length += sizeof(resp.device_cap_flags_ex);
 end:
 	err = ib_copy_to_udata(ucore, &resp, resp.response_length);
 	return err;

drivers/infiniband/core/verbs.c

@@ -48,6 +48,7 @@
 #include <rdma/ib_verbs.h>
 #include <rdma/ib_cache.h>
 #include <rdma/ib_addr.h>
+#include <rdma/rw.h>
 
 #include "core_priv.h"
 
@@ -723,59 +724,89 @@ struct ib_qp *ib_open_qp(struct ib_xrcd *xrcd,
 }
 EXPORT_SYMBOL(ib_open_qp);
 
+static struct ib_qp *ib_create_xrc_qp(struct ib_qp *qp,
+		struct ib_qp_init_attr *qp_init_attr)
+{
+	struct ib_qp *real_qp = qp;
+
+	qp->event_handler = __ib_shared_qp_event_handler;
+	qp->qp_context = qp;
+	qp->pd = NULL;
+	qp->send_cq = qp->recv_cq = NULL;
+	qp->srq = NULL;
+	qp->xrcd = qp_init_attr->xrcd;
+	atomic_inc(&qp_init_attr->xrcd->usecnt);
+	INIT_LIST_HEAD(&qp->open_list);
+
+	qp = __ib_open_qp(real_qp, qp_init_attr->event_handler,
+			  qp_init_attr->qp_context);
+	if (!IS_ERR(qp))
+		__ib_insert_xrcd_qp(qp_init_attr->xrcd, real_qp);
+	else
+		real_qp->device->destroy_qp(real_qp);
+	return qp;
+}
+
 struct ib_qp *ib_create_qp(struct ib_pd *pd,
 			   struct ib_qp_init_attr *qp_init_attr)
 {
-	struct ib_qp *qp, *real_qp;
-	struct ib_device *device;
+	struct ib_device *device = pd ? pd->device : qp_init_attr->xrcd->device;
+	struct ib_qp *qp;
+	int ret;
+
+	/*
+	 * If the callers is using the RDMA API calculate the resources
+	 * needed for the RDMA READ/WRITE operations.
+	 *
+	 * Note that these callers need to pass in a port number.
+	 */
+	if (qp_init_attr->cap.max_rdma_ctxs)
+		rdma_rw_init_qp(device, qp_init_attr);
 
-	device = pd ? pd->device : qp_init_attr->xrcd->device;
 	qp = device->create_qp(pd, qp_init_attr, NULL);
+	if (IS_ERR(qp))
+		return qp;
 
-	if (!IS_ERR(qp)) {
-		qp->device     = device;
-		qp->real_qp    = qp;
-		qp->uobject    = NULL;
-		qp->qp_type    = qp_init_attr->qp_type;
+	qp->device     = device;
+	qp->real_qp    = qp;
+	qp->uobject    = NULL;
+	qp->qp_type    = qp_init_attr->qp_type;
 
-		atomic_set(&qp->usecnt, 0);
-		if (qp_init_attr->qp_type == IB_QPT_XRC_TGT) {
-			qp->event_handler = __ib_shared_qp_event_handler;
-			qp->qp_context = qp;
-			qp->pd = NULL;
-			qp->send_cq = qp->recv_cq = NULL;
-			qp->srq = NULL;
-			qp->xrcd = qp_init_attr->xrcd;
-			atomic_inc(&qp_init_attr->xrcd->usecnt);
-			INIT_LIST_HEAD(&qp->open_list);
+	atomic_set(&qp->usecnt, 0);
+	qp->mrs_used = 0;
+	spin_lock_init(&qp->mr_lock);
+	INIT_LIST_HEAD(&qp->rdma_mrs);
+	INIT_LIST_HEAD(&qp->sig_mrs);
 
-			real_qp = qp;
-			qp = __ib_open_qp(real_qp, qp_init_attr->event_handler,
-					  qp_init_attr->qp_context);
-			if (!IS_ERR(qp))
-				__ib_insert_xrcd_qp(qp_init_attr->xrcd, real_qp);
-			else
-				real_qp->device->destroy_qp(real_qp);
-		} else {
-			qp->event_handler = qp_init_attr->event_handler;
-			qp->qp_context = qp_init_attr->qp_context;
-			if (qp_init_attr->qp_type == IB_QPT_XRC_INI) {
-				qp->recv_cq = NULL;
-				qp->srq = NULL;
-			} else {
-				qp->recv_cq = qp_init_attr->recv_cq;
-				atomic_inc(&qp_init_attr->recv_cq->usecnt);
-				qp->srq = qp_init_attr->srq;
-				if (qp->srq)
-					atomic_inc(&qp_init_attr->srq->usecnt);
-			}
+	if (qp_init_attr->qp_type == IB_QPT_XRC_TGT)
+		return ib_create_xrc_qp(qp, qp_init_attr);
 
-			qp->pd	    = pd;
-			qp->send_cq = qp_init_attr->send_cq;
-			qp->xrcd    = NULL;
+	qp->event_handler = qp_init_attr->event_handler;
+	qp->qp_context = qp_init_attr->qp_context;
+	if (qp_init_attr->qp_type == IB_QPT_XRC_INI) {
+		qp->recv_cq = NULL;
+		qp->srq = NULL;
+	} else {
+		qp->recv_cq = qp_init_attr->recv_cq;
+		atomic_inc(&qp_init_attr->recv_cq->usecnt);
+		qp->srq = qp_init_attr->srq;
+		if (qp->srq)
+			atomic_inc(&qp_init_attr->srq->usecnt);
+	}
 
-			atomic_inc(&pd->usecnt);
-			atomic_inc(&qp_init_attr->send_cq->usecnt);
+	qp->pd	    = pd;
+	qp->send_cq = qp_init_attr->send_cq;
+	qp->xrcd    = NULL;
+
+	atomic_inc(&pd->usecnt);
+	atomic_inc(&qp_init_attr->send_cq->usecnt);
+
+	if (qp_init_attr->cap.max_rdma_ctxs) {
+		ret = rdma_rw_init_mrs(qp, qp_init_attr);
+		if (ret) {
+			pr_err("failed to init MR pool ret= %d\n", ret);
+			ib_destroy_qp(qp);
+			qp = ERR_PTR(ret);
 		}
 	}
 
@@ -1250,6 +1281,8 @@ int ib_destroy_qp(struct ib_qp *qp)
 	struct ib_srq *srq;
 	int ret;
 
+	WARN_ON_ONCE(qp->mrs_used > 0);
+
 	if (atomic_read(&qp->usecnt))
 		return -EBUSY;
 
@@ -1261,6 +1294,9 @@ int ib_destroy_qp(struct ib_qp *qp)
 	rcq  = qp->recv_cq;
 	srq  = qp->srq;
 
+	if (!qp->uobject)
+		rdma_rw_cleanup_mrs(qp);
+
 	ret = qp->device->destroy_qp(qp);
 	if (!ret) {
 		if (pd)
@@ -1343,6 +1379,7 @@ struct ib_mr *ib_get_dma_mr(struct ib_pd *pd, int mr_access_flags)
 		mr->pd      = pd;
 		mr->uobject = NULL;
 		atomic_inc(&pd->usecnt);
+		mr->need_inval = false;
 	}
 
 	return mr;
@@ -1389,6 +1426,7 @@ struct ib_mr *ib_alloc_mr(struct ib_pd *pd,
 		mr->pd      = pd;
 		mr->uobject = NULL;
 		atomic_inc(&pd->usecnt);
+		mr->need_inval = false;
 	}
 
 	return mr;
@@ -1597,6 +1635,7 @@ EXPORT_SYMBOL(ib_set_vf_guid);
  * @mr:            memory region
  * @sg:            dma mapped scatterlist
  * @sg_nents:      number of entries in sg
+ * @sg_offset:     offset in bytes into sg
  * @page_size:     page vector desired page size
  *
  * Constraints:
@@ -1615,17 +1654,15 @@ EXPORT_SYMBOL(ib_set_vf_guid);
  * After this completes successfully, the  memory region
  * is ready for registration.
  */
-int ib_map_mr_sg(struct ib_mr *mr,
-		 struct scatterlist *sg,
-		 int sg_nents,
-		 unsigned int page_size)
+int ib_map_mr_sg(struct ib_mr *mr, struct scatterlist *sg, int sg_nents,
+		 unsigned int *sg_offset, unsigned int page_size)
 {
 	if (unlikely(!mr->device->map_mr_sg))
 		return -ENOSYS;
 
 	mr->page_size = page_size;
 
-	return mr->device->map_mr_sg(mr, sg, sg_nents);
+	return mr->device->map_mr_sg(mr, sg, sg_nents, sg_offset);
 }
 EXPORT_SYMBOL(ib_map_mr_sg);
 
@@ -1635,6 +1672,10 @@ EXPORT_SYMBOL(ib_map_mr_sg);
  * @mr:            memory region
  * @sgl:           dma mapped scatterlist
  * @sg_nents:      number of entries in sg
+ * @sg_offset_p:   IN:  start offset in bytes into sg
+ *                 OUT: offset in bytes for element n of the sg of the first
+ *                      byte that has not been processed where n is the return
+ *                      value of this function.
  * @set_page:      driver page assignment function pointer
  *
  * Core service helper for drivers to convert the largest
@@ -1645,23 +1686,26 @@ EXPORT_SYMBOL(ib_map_mr_sg);
  * Returns the number of sg elements that were assigned to
  * a page vector.
  */
-int ib_sg_to_pages(struct ib_mr *mr,
-		   struct scatterlist *sgl,
-		   int sg_nents,
-		   int (*set_page)(struct ib_mr *, u64))
+int ib_sg_to_pages(struct ib_mr *mr, struct scatterlist *sgl, int sg_nents,
+		unsigned int *sg_offset_p, int (*set_page)(struct ib_mr *, u64))
 {
 	struct scatterlist *sg;
 	u64 last_end_dma_addr = 0;
+	unsigned int sg_offset = sg_offset_p ? *sg_offset_p : 0;
 	unsigned int last_page_off = 0;
 	u64 page_mask = ~((u64)mr->page_size - 1);
 	int i, ret;
 
-	mr->iova = sg_dma_address(&sgl[0]);
+	if (unlikely(sg_nents <= 0 || sg_offset > sg_dma_len(&sgl[0])))
+		return -EINVAL;
+
+	mr->iova = sg_dma_address(&sgl[0]) + sg_offset;
 	mr->length = 0;
 
 	for_each_sg(sgl, sg, sg_nents, i) {
-		u64 dma_addr = sg_dma_address(sg);
-		unsigned int dma_len = sg_dma_len(sg);
+		u64 dma_addr = sg_dma_address(sg) + sg_offset;
+		u64 prev_addr = dma_addr;
+		unsigned int dma_len = sg_dma_len(sg) - sg_offset;
 		u64 end_dma_addr = dma_addr + dma_len;
 		u64 page_addr = dma_addr & page_mask;
 
@@ -1685,8 +1729,14 @@ int ib_sg_to_pages(struct ib_mr *mr,
 
 		do {
 			ret = set_page(mr, page_addr);
-			if (unlikely(ret < 0))
-				return i ? : ret;
+			if (unlikely(ret < 0)) {
+				sg_offset = prev_addr - sg_dma_address(sg);
+				mr->length += prev_addr - dma_addr;
+				if (sg_offset_p)
+					*sg_offset_p = sg_offset;
+				return i || sg_offset ? i : ret;
+			}
+			prev_addr = page_addr;
 next_page:
 			page_addr += mr->page_size;
 		} while (page_addr < end_dma_addr);
@@ -1694,8 +1744,12 @@ int ib_sg_to_pages(struct ib_mr *mr,
 		mr->length += dma_len;
 		last_end_dma_addr = end_dma_addr;
 		last_page_off = end_dma_addr & ~page_mask;
+
+		sg_offset = 0;
 	}
 
+	if (sg_offset_p)
+		*sg_offset_p = 0;
 	return i;
 }
 EXPORT_SYMBOL(ib_sg_to_pages);

drivers/infiniband/hw/cxgb3/iwch_provider.c

@@ -783,15 +783,14 @@ static int iwch_set_page(struct ib_mr *ibmr, u64 addr)
 	return 0;
 }
 
-static int iwch_map_mr_sg(struct ib_mr *ibmr,
-			  struct scatterlist *sg,
-			  int sg_nents)
+static int iwch_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg,
+			  int sg_nents, unsigned int *sg_offset)
 {
 	struct iwch_mr *mhp = to_iwch_mr(ibmr);
 
 	mhp->npages = 0;
 
-	return ib_sg_to_pages(ibmr, sg, sg_nents, iwch_set_page);
+	return ib_sg_to_pages(ibmr, sg, sg_nents, sg_offset, iwch_set_page);
 }
 
 static int iwch_destroy_qp(struct ib_qp *ib_qp)

drivers/infiniband/hw/cxgb4/iw_cxgb4.h

@@ -755,6 +755,7 @@ enum c4iw_ep_flags {
 	CLOSE_SENT		= 3,
 	TIMEOUT                 = 4,
 	QP_REFERENCED           = 5,
+	STOP_MPA_TIMER		= 7,
 };
 
 enum c4iw_ep_history {
@@ -779,7 +780,13 @@ enum c4iw_ep_history {
 	EP_DISC_ABORT           = 18,
 	CONN_RPL_UPCALL         = 19,
 	ACT_RETRY_NOMEM         = 20,
-	ACT_RETRY_INUSE         = 21
+	ACT_RETRY_INUSE         = 21,
+	CLOSE_CON_RPL		= 22,
+	EP_DISC_FAIL		= 24,
+	QP_REFED		= 25,
+	QP_DEREFED		= 26,
+	CM_ID_REFED		= 27,
+	CM_ID_DEREFED		= 28,
 };
 
 struct c4iw_ep_common {
@@ -917,9 +924,8 @@ void c4iw_qp_rem_ref(struct ib_qp *qp);
 struct ib_mr *c4iw_alloc_mr(struct ib_pd *pd,
 			    enum ib_mr_type mr_type,
 			    u32 max_num_sg);
-int c4iw_map_mr_sg(struct ib_mr *ibmr,
-		   struct scatterlist *sg,
-		   int sg_nents);
+int c4iw_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg, int sg_nents,
+		   unsigned int *sg_offset);
 int c4iw_dealloc_mw(struct ib_mw *mw);
 struct ib_mw *c4iw_alloc_mw(struct ib_pd *pd, enum ib_mw_type type,
 			    struct ib_udata *udata);

drivers/infiniband/hw/cxgb4/mem.c

@@ -86,8 +86,9 @@ static int _c4iw_write_mem_dma_aligned(struct c4iw_rdev *rdev, u32 addr,
 			(wait ? FW_WR_COMPL_F : 0));
 	req->wr.wr_lo = wait ? (__force __be64)(unsigned long) &wr_wait : 0L;
 	req->wr.wr_mid = cpu_to_be32(FW_WR_LEN16_V(DIV_ROUND_UP(wr_len, 16)));
-	req->cmd = cpu_to_be32(ULPTX_CMD_V(ULP_TX_MEM_WRITE));
-	req->cmd |= cpu_to_be32(T5_ULP_MEMIO_ORDER_V(1));
+	req->cmd = cpu_to_be32(ULPTX_CMD_V(ULP_TX_MEM_WRITE) |
+			       T5_ULP_MEMIO_ORDER_V(1) |
+			       T5_ULP_MEMIO_FID_V(rdev->lldi.rxq_ids[0]));
 	req->dlen = cpu_to_be32(ULP_MEMIO_DATA_LEN_V(len>>5));
 	req->len16 = cpu_to_be32(DIV_ROUND_UP(wr_len-sizeof(req->wr), 16));
 	req->lock_addr = cpu_to_be32(ULP_MEMIO_ADDR_V(addr));
@@ -690,15 +691,14 @@ static int c4iw_set_page(struct ib_mr *ibmr, u64 addr)
 	return 0;
 }
 
-int c4iw_map_mr_sg(struct ib_mr *ibmr,
-		   struct scatterlist *sg,
-		   int sg_nents)
+int c4iw_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg, int sg_nents,
+		   unsigned int *sg_offset)
 {
 	struct c4iw_mr *mhp = to_c4iw_mr(ibmr);
 
 	mhp->mpl_len = 0;
 
-	return ib_sg_to_pages(ibmr, sg, sg_nents, c4iw_set_page);
+	return ib_sg_to_pages(ibmr, sg, sg_nents, sg_offset, c4iw_set_page);
 }
 
 int c4iw_dereg_mr(struct ib_mr *ib_mr)

drivers/infiniband/hw/i40iw/i40iw.h

@@ -50,8 +50,6 @@
 #include <rdma/ib_pack.h>
 #include <rdma/rdma_cm.h>
 #include <rdma/iw_cm.h>
-#include <rdma/iw_portmap.h>
-#include <rdma/rdma_netlink.h>
 #include <crypto/hash.h>
 
 #include "i40iw_status.h"
@@ -254,6 +252,7 @@ struct i40iw_device {
 	u32 arp_table_size;
 	u32 next_arp_index;
 	spinlock_t resource_lock; /* hw resource access */
+	spinlock_t qptable_lock;
 	u32 vendor_id;
 	u32 vendor_part_id;
 	u32 of_device_registered;
@@ -392,7 +391,7 @@ void i40iw_flush_wqes(struct i40iw_device *iwdev,
 
 void i40iw_manage_arp_cache(struct i40iw_device *iwdev,
 			    unsigned char *mac_addr,
-			    __be32 *ip_addr,
+			    u32 *ip_addr,
 			    bool ipv4,
 			    u32 action);
 
@@ -550,7 +549,7 @@ enum i40iw_status_code i40iw_hw_flush_wqes(struct i40iw_device *iwdev,
 					   struct i40iw_qp_flush_info *info,
 					   bool wait);
 
-void i40iw_copy_ip_ntohl(u32 *dst, u32 *src);
+void i40iw_copy_ip_ntohl(u32 *dst, __be32 *src);
 struct ib_mr *i40iw_reg_phys_mr(struct ib_pd *ib_pd,
 				u64 addr,
 				u64 size,

drivers/infiniband/hw/i40iw/i40iw_cm.c

@@ -771,6 +771,7 @@ static void i40iw_build_mpa_v2(struct i40iw_cm_node *cm_node,
 {
 	struct ietf_mpa_v2 *mpa_frame = (struct ietf_mpa_v2 *)start_addr;
 	struct ietf_rtr_msg *rtr_msg = &mpa_frame->rtr_msg;
+	u16 ctrl_ird, ctrl_ord;
 
 	/* initialize the upper 5 bytes of the frame */
 	i40iw_build_mpa_v1(cm_node, start_addr, mpa_key);
@@ -779,38 +780,38 @@ static void i40iw_build_mpa_v2(struct i40iw_cm_node *cm_node,
 
 	/* initialize RTR msg */
 	if (cm_node->mpav2_ird_ord == IETF_NO_IRD_ORD) {
-		rtr_msg->ctrl_ird = IETF_NO_IRD_ORD;
-		rtr_msg->ctrl_ord = IETF_NO_IRD_ORD;
+		ctrl_ird = IETF_NO_IRD_ORD;
+		ctrl_ord = IETF_NO_IRD_ORD;
 	} else {
-		rtr_msg->ctrl_ird = (cm_node->ird_size > IETF_NO_IRD_ORD) ?
+		ctrl_ird = (cm_node->ird_size > IETF_NO_IRD_ORD) ?
 			IETF_NO_IRD_ORD : cm_node->ird_size;
-		rtr_msg->ctrl_ord = (cm_node->ord_size > IETF_NO_IRD_ORD) ?
+		ctrl_ord = (cm_node->ord_size > IETF_NO_IRD_ORD) ?
 			IETF_NO_IRD_ORD : cm_node->ord_size;
 	}
 
-	rtr_msg->ctrl_ird |= IETF_PEER_TO_PEER;
-	rtr_msg->ctrl_ird |= IETF_FLPDU_ZERO_LEN;
+	ctrl_ird |= IETF_PEER_TO_PEER;
+	ctrl_ird |= IETF_FLPDU_ZERO_LEN;
 
 	switch (mpa_key) {
 	case MPA_KEY_REQUEST:
-		rtr_msg->ctrl_ord |= IETF_RDMA0_WRITE;
-		rtr_msg->ctrl_ord |= IETF_RDMA0_READ;
+		ctrl_ord |= IETF_RDMA0_WRITE;
+		ctrl_ord |= IETF_RDMA0_READ;
 		break;
 	case MPA_KEY_REPLY:
 		switch (cm_node->send_rdma0_op) {
 		case SEND_RDMA_WRITE_ZERO:
-			rtr_msg->ctrl_ord |= IETF_RDMA0_WRITE;
+			ctrl_ord |= IETF_RDMA0_WRITE;
 			break;
 		case SEND_RDMA_READ_ZERO:
-			rtr_msg->ctrl_ord |= IETF_RDMA0_READ;
+			ctrl_ord |= IETF_RDMA0_READ;
 			break;
 		}
 		break;
 	default:
 		break;
 	}
-	rtr_msg->ctrl_ird = htons(rtr_msg->ctrl_ird);
-	rtr_msg->ctrl_ord = htons(rtr_msg->ctrl_ord);
+	rtr_msg->ctrl_ird = htons(ctrl_ird);
+	rtr_msg->ctrl_ord = htons(ctrl_ord);
 }
 
 /**
@@ -2107,7 +2108,7 @@ static bool i40iw_ipv6_is_loopback(u32 *loc_addr, u32 *rem_addr)
 	struct in6_addr raddr6;
 
 	i40iw_copy_ip_htonl(raddr6.in6_u.u6_addr32, rem_addr);
-	return (!memcmp(loc_addr, rem_addr, 16) || ipv6_addr_loopback(&raddr6));
+	return !memcmp(loc_addr, rem_addr, 16) || ipv6_addr_loopback(&raddr6);
 }
 
 /**
@@ -2160,7 +2161,7 @@ static struct i40iw_cm_node *i40iw_make_cm_node(
 	cm_node->tcp_cntxt.rcv_wnd =
 			I40IW_CM_DEFAULT_RCV_WND_SCALED >> I40IW_CM_DEFAULT_RCV_WND_SCALE;
 	ts = current_kernel_time();
-	cm_node->tcp_cntxt.loc_seq_num = htonl(ts.tv_nsec);
+	cm_node->tcp_cntxt.loc_seq_num = ts.tv_nsec;
 	cm_node->tcp_cntxt.mss = iwdev->mss;
 
 	cm_node->iwdev = iwdev;
@@ -2234,7 +2235,7 @@ static void i40iw_rem_ref_cm_node(struct i40iw_cm_node *cm_node)
 	if (cm_node->listener) {
 		i40iw_dec_refcnt_listen(cm_core, cm_node->listener, 0, true);
 	} else {
-		if (!i40iw_listen_port_in_use(cm_core, htons(cm_node->loc_port)) &&
+		if (!i40iw_listen_port_in_use(cm_core, cm_node->loc_port) &&
 		    cm_node->apbvt_set && cm_node->iwdev) {
 			i40iw_manage_apbvt(cm_node->iwdev,
 					   cm_node->loc_port,
@@ -2852,7 +2853,6 @@ static struct i40iw_cm_node *i40iw_create_cm_node(
 					void *private_data,
 					struct i40iw_cm_info *cm_info)
 {
-	int ret;
 	struct i40iw_cm_node *cm_node;
 	struct i40iw_cm_listener *loopback_remotelistener;
 	struct i40iw_cm_node *loopback_remotenode;
@@ -2922,30 +2922,6 @@ static struct i40iw_cm_node *i40iw_create_cm_node(
 	memcpy(cm_node->pdata_buf, private_data, private_data_len);
 
 	cm_node->state = I40IW_CM_STATE_SYN_SENT;
-	ret = i40iw_send_syn(cm_node, 0);
-
-	if (ret) {
-		if (cm_node->ipv4)
-			i40iw_debug(cm_node->dev,
-				    I40IW_DEBUG_CM,
-				    "Api - connect() FAILED: dest addr=%pI4",
-				    cm_node->rem_addr);
-		else
-			i40iw_debug(cm_node->dev, I40IW_DEBUG_CM,
-				    "Api - connect() FAILED: dest addr=%pI6",
-				    cm_node->rem_addr);
-		i40iw_rem_ref_cm_node(cm_node);
-		cm_node = NULL;
-	}
-
-	if (cm_node)
-		i40iw_debug(cm_node->dev,
-			    I40IW_DEBUG_CM,
-			    "Api - connect(): port=0x%04x, cm_node=%p, cm_id = %p.\n",
-			    cm_node->rem_port,
-			    cm_node,
-			    cm_node->cm_id);
-
 	return cm_node;
 }
 
@@ -3266,11 +3242,13 @@ static void i40iw_init_tcp_ctx(struct i40iw_cm_node *cm_node,
 
 		tcp_info->dest_ip_addr3 = cpu_to_le32(cm_node->rem_addr[0]);
 		tcp_info->local_ipaddr3 = cpu_to_le32(cm_node->loc_addr[0]);
-		tcp_info->arp_idx = cpu_to_le32(i40iw_arp_table(iwqp->iwdev,
-								&tcp_info->dest_ip_addr3,
-								true,
-								NULL,
-								I40IW_ARP_RESOLVE));
+		tcp_info->arp_idx =
+			cpu_to_le16((u16)i40iw_arp_table(
+							 iwqp->iwdev,
+							 &tcp_info->dest_ip_addr3,
+							 true,
+							 NULL,
+							 I40IW_ARP_RESOLVE));
 	} else {
 		tcp_info->src_port = cpu_to_le16(cm_node->loc_port);
 		tcp_info->dst_port = cpu_to_le16(cm_node->rem_port);
@@ -3282,12 +3260,13 @@ static void i40iw_init_tcp_ctx(struct i40iw_cm_node *cm_node,
 		tcp_info->local_ipaddr1 = cpu_to_le32(cm_node->loc_addr[1]);
 		tcp_info->local_ipaddr2 = cpu_to_le32(cm_node->loc_addr[2]);
 		tcp_info->local_ipaddr3 = cpu_to_le32(cm_node->loc_addr[3]);
-		tcp_info->arp_idx = cpu_to_le32(i40iw_arp_table(
-							iwqp->iwdev,
-							&tcp_info->dest_ip_addr0,
-							false,
-							NULL,
-							I40IW_ARP_RESOLVE));
+		tcp_info->arp_idx =
+			cpu_to_le16((u16)i40iw_arp_table(
+							 iwqp->iwdev,
+							 &tcp_info->dest_ip_addr0,
+							 false,
+							 NULL,
+							 I40IW_ARP_RESOLVE));
 	}
 }
 
@@ -3564,7 +3543,6 @@ int i40iw_accept(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
 	struct i40iw_cm_node *cm_node;
 	struct ib_qp_attr attr;
 	int passive_state;
-	struct i40iw_ib_device *iwibdev;
 	struct ib_mr *ibmr;
 	struct i40iw_pd *iwpd;
 	u16 buf_len = 0;
@@ -3627,7 +3605,6 @@ int i40iw_accept(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
 	     !i40iw_ipv4_is_loopback(cm_node->loc_addr[0], cm_node->rem_addr[0])) ||
 	    (!cm_node->ipv4 &&
 	     !i40iw_ipv6_is_loopback(cm_node->loc_addr, cm_node->rem_addr))) {
-		iwibdev = iwdev->iwibdev;
 		iwpd = iwqp->iwpd;
 		tagged_offset = (uintptr_t)iwqp->ietf_mem.va;
 		ibmr = i40iw_reg_phys_mr(&iwpd->ibpd,
@@ -3752,6 +3729,7 @@ int i40iw_connect(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
 	struct sockaddr_in *raddr;
 	struct sockaddr_in6 *laddr6;
 	struct sockaddr_in6 *raddr6;
+	bool qhash_set = false;
 	int apbvt_set = 0;
 	enum i40iw_status_code status;
 
@@ -3810,6 +3788,7 @@ int i40iw_connect(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
 					    true);
 		if (status)
 			return -EINVAL;
+		qhash_set = true;
 	}
 	status = i40iw_manage_apbvt(iwdev, cm_info.loc_port, I40IW_MANAGE_APBVT_ADD);
 	if (status) {
@@ -3828,23 +3807,8 @@ int i40iw_connect(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
 				       conn_param->private_data_len,
 				       (void *)conn_param->private_data,
 				       &cm_info);
-	if (!cm_node) {
-		i40iw_manage_qhash(iwdev,
-				   &cm_info,
-				   I40IW_QHASH_TYPE_TCP_ESTABLISHED,
-				   I40IW_QHASH_MANAGE_TYPE_DELETE,
-				   NULL,
-				   false);
-
-		if (apbvt_set && !i40iw_listen_port_in_use(&iwdev->cm_core,
-							   cm_info.loc_port))
-			i40iw_manage_apbvt(iwdev,
-					   cm_info.loc_port,
-					   I40IW_MANAGE_APBVT_DEL);
-		cm_id->rem_ref(cm_id);
-		iwdev->cm_core.stats_connect_errs++;
-		return -ENOMEM;
-	}
+	if (!cm_node)
+		goto err;
 
 	i40iw_record_ird_ord(cm_node, (u16)conn_param->ird, (u16)conn_param->ord);
 	if (cm_node->send_rdma0_op == SEND_RDMA_READ_ZERO &&
@@ -3852,12 +3816,54 @@ int i40iw_connect(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
 		cm_node->ord_size = 1;
 
 	cm_node->apbvt_set = apbvt_set;
-	cm_node->qhash_set = true;
+	cm_node->qhash_set = qhash_set;
 	iwqp->cm_node = cm_node;
 	cm_node->iwqp = iwqp;
 	iwqp->cm_id = cm_id;
 	i40iw_add_ref(&iwqp->ibqp);
+
+	if (cm_node->state == I40IW_CM_STATE_SYN_SENT) {
+		if (i40iw_send_syn(cm_node, 0)) {
+			i40iw_rem_ref_cm_node(cm_node);
+			goto err;
+		}
+	}
+
+	i40iw_debug(cm_node->dev,
+		    I40IW_DEBUG_CM,
+		    "Api - connect(): port=0x%04x, cm_node=%p, cm_id = %p.\n",
+		    cm_node->rem_port,
+		    cm_node,
+		    cm_node->cm_id);
 	return 0;
+
+err:
+	if (cm_node) {
+		if (cm_node->ipv4)
+			i40iw_debug(cm_node->dev,
+				    I40IW_DEBUG_CM,
+				    "Api - connect() FAILED: dest addr=%pI4",
+				    cm_node->rem_addr);
+		else
+			i40iw_debug(cm_node->dev, I40IW_DEBUG_CM,
+				    "Api - connect() FAILED: dest addr=%pI6",
+				    cm_node->rem_addr);
+	}
+	i40iw_manage_qhash(iwdev,
+			   &cm_info,
+			   I40IW_QHASH_TYPE_TCP_ESTABLISHED,
+			   I40IW_QHASH_MANAGE_TYPE_DELETE,
+			   NULL,
+			   false);
+
+	if (apbvt_set && !i40iw_listen_port_in_use(&iwdev->cm_core,
+						   cm_info.loc_port))
+		i40iw_manage_apbvt(iwdev,
+				   cm_info.loc_port,
+				   I40IW_MANAGE_APBVT_DEL);
+	cm_id->rem_ref(cm_id);
+	iwdev->cm_core.stats_connect_errs++;
+	return -ENOMEM;
 }
 
 /**

drivers/infiniband/hw/i40iw/i40iw_cm.h

@@ -1,6 +1,6 @@
 /*******************************************************************************
 *
-* Copyright (c) 2015 Intel Corporation.  All rights reserved.
+* Copyright (c) 2015-2016 Intel Corporation.  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
@@ -291,8 +291,6 @@ struct i40iw_cm_listener {
 	u8 loc_mac[ETH_ALEN];
 	u32 loc_addr[4];
 	u16 loc_port;
-	u32 map_loc_addr[4];
-	u16 map_loc_port;
 	struct iw_cm_id *cm_id;
 	atomic_t ref_count;
 	struct i40iw_device *iwdev;
@@ -317,8 +315,6 @@ struct i40iw_kmem_info {
 struct i40iw_cm_node {
 	u32 loc_addr[4], rem_addr[4];
 	u16 loc_port, rem_port;
-	u32 map_loc_addr[4], map_rem_addr[4];
-	u16 map_loc_port, map_rem_port;
 	u16 vlan_id;
 	enum i40iw_cm_node_state state;
 	u8 loc_mac[ETH_ALEN];
@@ -370,10 +366,6 @@ struct i40iw_cm_info {
 	u16 rem_port;
 	u32 loc_addr[4];
 	u32 rem_addr[4];
-	u16 map_loc_port;
-	u16 map_rem_port;
-	u32 map_loc_addr[4];
-	u32 map_rem_addr[4];
 	u16 vlan_id;
 	int backlog;
 	u16 user_pri;

drivers/infiniband/hw/i40iw/i40iw_ctrl.c

@@ -114,16 +114,21 @@ static enum i40iw_status_code i40iw_cqp_poll_registers(
  * i40iw_sc_parse_fpm_commit_buf - parse fpm commit buffer
  * @buf: ptr to fpm commit buffer
  * @info: ptr to i40iw_hmc_obj_info struct
+ * @sd: number of SDs for HMC objects
  *
  * parses fpm commit info and copy base value
  * of hmc objects in hmc_info
  */
 static enum i40iw_status_code i40iw_sc_parse_fpm_commit_buf(
 				u64 *buf,
-				struct i40iw_hmc_obj_info *info)
+				struct i40iw_hmc_obj_info *info,
+				u32 *sd)
 {
 	u64 temp;
+	u64 size;
+	u64 base = 0;
 	u32 i, j;
+	u32 k = 0;
 	u32 low;
 
 	/* copy base values in obj_info */
@@ -131,10 +136,20 @@ static enum i40iw_status_code i40iw_sc_parse_fpm_commit_buf(
 			i <= I40IW_HMC_IW_PBLE; i++, j += 8) {
 		get_64bit_val(buf, j, &temp);
 		info[i].base = RS_64_1(temp, 32) * 512;
+		if (info[i].base > base) {
+			base = info[i].base;
+			k = i;
+		}
 		low = (u32)(temp);
 		if (low)
 			info[i].cnt = low;
 	}
+	size = info[k].cnt * info[k].size + info[k].base;
+	if (size & 0x1FFFFF)
+		*sd = (u32)((size >> 21) + 1); /* add 1 for remainder */
+	else
+		*sd = (u32)(size >> 21);
+
 	return 0;
 }
 
@@ -2908,6 +2923,65 @@ static enum i40iw_status_code i40iw_sc_mw_alloc(
 	return 0;
 }
 
+/**
+ * i40iw_sc_mr_fast_register - Posts RDMA fast register mr WR to iwarp qp
+ * @qp: sc qp struct
+ * @info: fast mr info
+ * @post_sq: flag for cqp db to ring
+ */
+enum i40iw_status_code i40iw_sc_mr_fast_register(
+				struct i40iw_sc_qp *qp,
+				struct i40iw_fast_reg_stag_info *info,
+				bool post_sq)
+{
+	u64 temp, header;
+	u64 *wqe;
+	u32 wqe_idx;
+
+	wqe = i40iw_qp_get_next_send_wqe(&qp->qp_uk, &wqe_idx, I40IW_QP_WQE_MIN_SIZE,
+					 0, info->wr_id);
+	if (!wqe)
+		return I40IW_ERR_QP_TOOMANY_WRS_POSTED;
+
+	i40iw_debug(qp->dev, I40IW_DEBUG_MR, "%s: wr_id[%llxh] wqe_idx[%04d] location[%p]\n",
+		    __func__, info->wr_id, wqe_idx,
+		    &qp->qp_uk.sq_wrtrk_array[wqe_idx].wrid);
+	temp = (info->addr_type == I40IW_ADDR_TYPE_VA_BASED) ? (uintptr_t)info->va : info->fbo;
+	set_64bit_val(wqe, 0, temp);
+
+	temp = RS_64(info->first_pm_pbl_index >> 16, I40IWQPSQ_FIRSTPMPBLIDXHI);
+	set_64bit_val(wqe,
+		      8,
+		      LS_64(temp, I40IWQPSQ_FIRSTPMPBLIDXHI) |
+		      LS_64(info->reg_addr_pa >> I40IWQPSQ_PBLADDR_SHIFT, I40IWQPSQ_PBLADDR));
+
+	set_64bit_val(wqe,
+		      16,
+		      info->total_len |
+		      LS_64(info->first_pm_pbl_index, I40IWQPSQ_FIRSTPMPBLIDXLO));
+
+	header = LS_64(info->stag_key, I40IWQPSQ_STAGKEY) |
+		 LS_64(info->stag_idx, I40IWQPSQ_STAGINDEX) |
+		 LS_64(I40IWQP_OP_FAST_REGISTER, I40IWQPSQ_OPCODE) |
+		 LS_64(info->chunk_size, I40IWQPSQ_LPBLSIZE) |
+		 LS_64(info->page_size, I40IWQPSQ_HPAGESIZE) |
+		 LS_64(info->access_rights, I40IWQPSQ_STAGRIGHTS) |
+		 LS_64(info->addr_type, I40IWQPSQ_VABASEDTO) |
+		 LS_64(info->read_fence, I40IWQPSQ_READFENCE) |
+		 LS_64(info->local_fence, I40IWQPSQ_LOCALFENCE) |
+		 LS_64(info->signaled, I40IWQPSQ_SIGCOMPL) |
+		 LS_64(qp->qp_uk.swqe_polarity, I40IWQPSQ_VALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(qp->dev, I40IW_DEBUG_WQE, "FAST_REG WQE",
+			wqe, I40IW_QP_WQE_MIN_SIZE);
+
+	if (post_sq)
+		i40iw_qp_post_wr(&qp->qp_uk);
+	return 0;
+}
+
 /**
  * i40iw_sc_send_lsmm - send last streaming mode message
  * @qp: sc qp struct
@@ -3147,7 +3221,7 @@ enum i40iw_status_code i40iw_sc_init_iw_hmc(struct i40iw_sc_dev *dev, u8 hmc_fn_
 		i40iw_cqp_commit_fpm_values_cmd(dev, &query_fpm_mem, hmc_fn_id);
 
 		/* parse the fpm_commit_buf and fill hmc obj info */
-		i40iw_sc_parse_fpm_commit_buf((u64 *)query_fpm_mem.va, hmc_info->hmc_obj);
+		i40iw_sc_parse_fpm_commit_buf((u64 *)query_fpm_mem.va, hmc_info->hmc_obj, &hmc_info->sd_table.sd_cnt);
 		mem_size = sizeof(struct i40iw_hmc_sd_entry) *
 			   (hmc_info->sd_table.sd_cnt + hmc_info->first_sd_index);
 		ret_code = i40iw_allocate_virt_mem(dev->hw, &virt_mem, mem_size);
@@ -3221,7 +3295,9 @@ static enum i40iw_status_code i40iw_sc_configure_iw_fpm(struct i40iw_sc_dev *dev
 
 	/* parse the fpm_commit_buf and fill hmc obj info */
 	if (!ret_code)
-		ret_code = i40iw_sc_parse_fpm_commit_buf(dev->fpm_commit_buf, hmc_info->hmc_obj);
+		ret_code = i40iw_sc_parse_fpm_commit_buf(dev->fpm_commit_buf,
+							 hmc_info->hmc_obj,
+							 &hmc_info->sd_table.sd_cnt);
 
 	i40iw_debug_buf(dev, I40IW_DEBUG_HMC, "COMMIT FPM BUFFER",
 			commit_fpm_mem.va, I40IW_COMMIT_FPM_BUF_SIZE);
@@ -3468,6 +3544,40 @@ static bool i40iw_ring_full(struct i40iw_sc_cqp *cqp)
 	return I40IW_RING_FULL_ERR(cqp->sq_ring);
 }
 
+/**
+ * i40iw_est_sd - returns approximate number of SDs for HMC
+ * @dev: sc device struct
+ * @hmc_info: hmc structure, size and count for HMC objects
+ */
+static u64 i40iw_est_sd(struct i40iw_sc_dev *dev, struct i40iw_hmc_info *hmc_info)
+{
+	int i;
+	u64 size = 0;
+	u64 sd;
+
+	for (i = I40IW_HMC_IW_QP; i < I40IW_HMC_IW_PBLE; i++)
+		size += hmc_info->hmc_obj[i].cnt * hmc_info->hmc_obj[i].size;
+
+	if (dev->is_pf)
+		size += hmc_info->hmc_obj[I40IW_HMC_IW_PBLE].cnt * hmc_info->hmc_obj[I40IW_HMC_IW_PBLE].size;
+
+	if (size & 0x1FFFFF)
+		sd = (size >> 21) + 1; /* add 1 for remainder */
+	else
+		sd = size >> 21;
+
+	if (!dev->is_pf) {
+		/* 2MB alignment for VF PBLE HMC */
+		size = hmc_info->hmc_obj[I40IW_HMC_IW_PBLE].cnt * hmc_info->hmc_obj[I40IW_HMC_IW_PBLE].size;
+		if (size & 0x1FFFFF)
+			sd += (size >> 21) + 1; /* add 1 for remainder */
+		else
+			sd += size >> 21;
+	}
+
+	return sd;
+}
+
 /**
  * i40iw_config_fpm_values - configure HMC objects
  * @dev: sc device struct
@@ -3479,7 +3589,7 @@ enum i40iw_status_code i40iw_config_fpm_values(struct i40iw_sc_dev *dev, u32 qp_
 	u32 i, mem_size;
 	u32 qpwantedoriginal, qpwanted, mrwanted, pblewanted;
 	u32 powerof2;
-	u64 sd_needed, bytes_needed;
+	u64 sd_needed;
 	u32 loop_count = 0;
 
 	struct i40iw_hmc_info *hmc_info;
@@ -3497,23 +3607,15 @@ enum i40iw_status_code i40iw_config_fpm_values(struct i40iw_sc_dev *dev, u32 qp_
 		return ret_code;
 	}
 
-	bytes_needed = 0;
-	for (i = I40IW_HMC_IW_QP; i < I40IW_HMC_IW_MAX; i++) {
+	for (i = I40IW_HMC_IW_QP; i < I40IW_HMC_IW_MAX; i++)
 		hmc_info->hmc_obj[i].cnt = hmc_info->hmc_obj[i].max_cnt;
-		bytes_needed +=
-		    (hmc_info->hmc_obj[i].max_cnt) * (hmc_info->hmc_obj[i].size);
-		i40iw_debug(dev, I40IW_DEBUG_HMC,
-			    "%s i[%04d] max_cnt[0x%04X] size[0x%04llx]\n",
-			    __func__, i, hmc_info->hmc_obj[i].max_cnt,
-			    hmc_info->hmc_obj[i].size);
-	}
-	sd_needed = (bytes_needed / I40IW_HMC_DIRECT_BP_SIZE) + 1; /* round up */
+	sd_needed = i40iw_est_sd(dev, hmc_info);
 	i40iw_debug(dev, I40IW_DEBUG_HMC,
 		    "%s: FW initial max sd_count[%08lld] first_sd_index[%04d]\n",
 		    __func__, sd_needed, hmc_info->first_sd_index);
 	i40iw_debug(dev, I40IW_DEBUG_HMC,
-		    "%s: bytes_needed=0x%llx sd count %d where max sd is %d\n",
-		    __func__, bytes_needed, hmc_info->sd_table.sd_cnt,
+		    "%s: sd count %d where max sd is %d\n",
+		    __func__, hmc_info->sd_table.sd_cnt,
 		    hmc_fpm_misc->max_sds);
 
 	qpwanted = min(qp_count, hmc_info->hmc_obj[I40IW_HMC_IW_QP].max_cnt);
@@ -3555,11 +3657,7 @@ enum i40iw_status_code i40iw_config_fpm_values(struct i40iw_sc_dev *dev, u32 qp_
 		hmc_info->hmc_obj[I40IW_HMC_IW_PBLE].cnt = pblewanted;
 
 		/* How much memory is needed for all the objects. */
-		bytes_needed = 0;
-		for (i = I40IW_HMC_IW_QP; i < I40IW_HMC_IW_MAX; i++)
-			bytes_needed +=
-			    (hmc_info->hmc_obj[i].cnt) * (hmc_info->hmc_obj[i].size);
-		sd_needed = (bytes_needed / I40IW_HMC_DIRECT_BP_SIZE) + 1;
+		sd_needed = i40iw_est_sd(dev, hmc_info);
 		if ((loop_count > 1000) ||
 		    ((!(loop_count % 10)) &&
 		    (qpwanted > qpwantedoriginal * 2 / 3))) {
@@ -3580,15 +3678,7 @@ enum i40iw_status_code i40iw_config_fpm_values(struct i40iw_sc_dev *dev, u32 qp_
 			pblewanted -= FPM_MULTIPLIER * 1000;
 	} while (sd_needed > hmc_fpm_misc->max_sds && loop_count < 2000);
 
-	bytes_needed = 0;
-	for (i = I40IW_HMC_IW_QP; i < I40IW_HMC_IW_MAX; i++) {
-		bytes_needed += (hmc_info->hmc_obj[i].cnt) * (hmc_info->hmc_obj[i].size);
-		i40iw_debug(dev, I40IW_DEBUG_HMC,
-			    "%s i[%04d] cnt[0x%04x] size[0x%04llx]\n",
-			    __func__, i, hmc_info->hmc_obj[i].cnt,
-			    hmc_info->hmc_obj[i].size);
-	}
-	sd_needed = (bytes_needed / I40IW_HMC_DIRECT_BP_SIZE) + 1;    /* round up not truncate. */
+	sd_needed = i40iw_est_sd(dev, hmc_info);
 
 	i40iw_debug(dev, I40IW_DEBUG_HMC,
 		    "loop_cnt=%d, sd_needed=%lld, qpcnt = %d, cqcnt=%d, mrcnt=%d, pblecnt=%d\n",
@@ -3606,8 +3696,6 @@ enum i40iw_status_code i40iw_config_fpm_values(struct i40iw_sc_dev *dev, u32 qp_
 		return ret_code;
 	}
 
-	hmc_info->sd_table.sd_cnt = (u32)sd_needed;
-
 	mem_size = sizeof(struct i40iw_hmc_sd_entry) *
 		   (hmc_info->sd_table.sd_cnt + hmc_info->first_sd_index + 1);
 	ret_code = i40iw_allocate_virt_mem(dev->hw, &virt_mem, mem_size);
@@ -3911,11 +3999,11 @@ enum i40iw_status_code i40iw_process_bh(struct i40iw_sc_dev *dev)
  */
 static u32 i40iw_iwarp_opcode(struct i40iw_aeqe_info *info, u8 *pkt)
 {
-	u16 *mpa;
+	__be16 *mpa;
 	u32 opcode = 0xffffffff;
 
 	if (info->q2_data_written) {
-		mpa = (u16 *)pkt;
+		mpa = (__be16 *)pkt;
 		opcode = ntohs(mpa[1]) & 0xf;
 	}
 	return opcode;
@@ -3977,7 +4065,7 @@ static int i40iw_bld_terminate_hdr(struct i40iw_sc_qp *qp,
 	if (info->q2_data_written) {
 		/* Use data from offending packet to fill in ddp & rdma hdrs */
 		pkt = i40iw_locate_mpa(pkt);
-		ddp_seg_len = ntohs(*(u16 *)pkt);
+		ddp_seg_len = ntohs(*(__be16 *)pkt);
 		if (ddp_seg_len) {
 			copy_len = 2;
 			termhdr->hdrct = DDP_LEN_FLAG;
@@ -4188,13 +4276,13 @@ void i40iw_terminate_connection(struct i40iw_sc_qp *qp, struct i40iw_aeqe_info *
 void i40iw_terminate_received(struct i40iw_sc_qp *qp, struct i40iw_aeqe_info *info)
 {
 	u8 *pkt = qp->q2_buf + Q2_BAD_FRAME_OFFSET;
-	u32 *mpa;
+	__be32 *mpa;
 	u8 ddp_ctl;
 	u8 rdma_ctl;
 	u16 aeq_id = 0;
 	struct i40iw_terminate_hdr *termhdr;
 
-	mpa = (u32 *)i40iw_locate_mpa(pkt);
+	mpa = (__be32 *)i40iw_locate_mpa(pkt);
 	if (info->q2_data_written) {
 		/* did not validate the frame - do it now */
 		ddp_ctl = (ntohl(mpa[0]) >> 8) & 0xff;
@@ -4559,17 +4647,18 @@ static struct i40iw_pd_ops iw_pd_ops = {
 };
 
 static struct i40iw_priv_qp_ops iw_priv_qp_ops = {
-	i40iw_sc_qp_init,
-	i40iw_sc_qp_create,
-	i40iw_sc_qp_modify,
-	i40iw_sc_qp_destroy,
-	i40iw_sc_qp_flush_wqes,
-	i40iw_sc_qp_upload_context,
-	i40iw_sc_qp_setctx,
-	i40iw_sc_send_lsmm,
-	i40iw_sc_send_lsmm_nostag,
-	i40iw_sc_send_rtt,
-	i40iw_sc_post_wqe0,
+	.qp_init = i40iw_sc_qp_init,
+	.qp_create = i40iw_sc_qp_create,
+	.qp_modify = i40iw_sc_qp_modify,
+	.qp_destroy = i40iw_sc_qp_destroy,
+	.qp_flush_wqes = i40iw_sc_qp_flush_wqes,
+	.qp_upload_context = i40iw_sc_qp_upload_context,
+	.qp_setctx = i40iw_sc_qp_setctx,
+	.qp_send_lsmm = i40iw_sc_send_lsmm,
+	.qp_send_lsmm_nostag = i40iw_sc_send_lsmm_nostag,
+	.qp_send_rtt = i40iw_sc_send_rtt,
+	.qp_post_wqe0 = i40iw_sc_post_wqe0,
+	.iw_mr_fast_register = i40iw_sc_mr_fast_register
 };
 
 static struct i40iw_priv_cq_ops iw_priv_cq_ops = {

drivers/infiniband/hw/i40iw/i40iw_d.h

@@ -1290,7 +1290,7 @@
 
 /* wqe size considering 32 bytes per wqe*/
 #define I40IWQP_SW_MIN_WQSIZE 4		/* 128 bytes */
-#define I40IWQP_SW_MAX_WQSIZE 16384	/* 524288 bytes */
+#define I40IWQP_SW_MAX_WQSIZE 2048	/* 2048 bytes */
 
 #define I40IWQP_OP_RDMA_WRITE 0
 #define I40IWQP_OP_RDMA_READ 1
@@ -1512,6 +1512,8 @@ enum i40iw_alignment {
 	I40IW_SD_BUF_ALIGNMENT =	0x100
 };
 
+#define I40IW_WQE_SIZE_64	64
+
 #define I40IW_QP_WQE_MIN_SIZE	32
 #define I40IW_QP_WQE_MAX_SIZE	128
 

drivers/infiniband/hw/i40iw/i40iw_hw.c

@@ -106,7 +106,9 @@ u32 i40iw_initialize_hw_resources(struct i40iw_device *iwdev)
 	set_bit(2, iwdev->allocated_pds);
 
 	spin_lock_init(&iwdev->resource_lock);
-	mrdrvbits = 24 - get_count_order(iwdev->max_mr);
+	spin_lock_init(&iwdev->qptable_lock);
+	/* stag index mask has a minimum of 14 bits */
+	mrdrvbits = 24 - max(get_count_order(iwdev->max_mr), 14);
 	iwdev->mr_stagmask = ~(((1 << mrdrvbits) - 1) << (32 - mrdrvbits));
 	return 0;
 }
@@ -301,11 +303,15 @@ void i40iw_process_aeq(struct i40iw_device *iwdev)
 			    "%s ae_id = 0x%x bool qp=%d qp_id = %d\n",
 			    __func__, info->ae_id, info->qp, info->qp_cq_id);
 		if (info->qp) {
+			spin_lock_irqsave(&iwdev->qptable_lock, flags);
 			iwqp = iwdev->qp_table[info->qp_cq_id];
 			if (!iwqp) {
+				spin_unlock_irqrestore(&iwdev->qptable_lock, flags);
 				i40iw_pr_err("qp_id %d is already freed\n", info->qp_cq_id);
 				continue;
 			}
+			i40iw_add_ref(&iwqp->ibqp);
+			spin_unlock_irqrestore(&iwdev->qptable_lock, flags);
 			qp = &iwqp->sc_qp;
 			spin_lock_irqsave(&iwqp->lock, flags);
 			iwqp->hw_tcp_state = info->tcp_state;
@@ -411,6 +417,8 @@ void i40iw_process_aeq(struct i40iw_device *iwdev)
 				i40iw_terminate_connection(qp, info);
 				break;
 		}
+		if (info->qp)
+			i40iw_rem_ref(&iwqp->ibqp);
 	} while (1);
 
 	if (aeqcnt)
@@ -460,7 +468,7 @@ int i40iw_manage_apbvt(struct i40iw_device *iwdev, u16 accel_local_port, bool ad
  */
 void i40iw_manage_arp_cache(struct i40iw_device *iwdev,
 			    unsigned char *mac_addr,
-			    __be32 *ip_addr,
+			    u32 *ip_addr,
 			    bool ipv4,
 			    u32 action)
 {
@@ -481,7 +489,7 @@ void i40iw_manage_arp_cache(struct i40iw_device *iwdev,
 		cqp_info->cqp_cmd = OP_ADD_ARP_CACHE_ENTRY;
 		info = &cqp_info->in.u.add_arp_cache_entry.info;
 		memset(info, 0, sizeof(*info));
-		info->arp_index = cpu_to_le32(arp_index);
+		info->arp_index = cpu_to_le16((u16)arp_index);
 		info->permanent = true;
 		ether_addr_copy(info->mac_addr, mac_addr);
 		cqp_info->in.u.add_arp_cache_entry.scratch = (uintptr_t)cqp_request;

drivers/infiniband/hw/i40iw/i40iw_main.c

@@ -270,7 +270,6 @@ static void i40iw_disable_irq(struct i40iw_sc_dev *dev,
 		i40iw_wr32(dev->hw, I40E_PFINT_DYN_CTLN(msix_vec->idx - 1), 0);
 	else
 		i40iw_wr32(dev->hw, I40E_VFINT_DYN_CTLN1(msix_vec->idx - 1), 0);
-	synchronize_irq(msix_vec->irq);
 	free_irq(msix_vec->irq, dev_id);
 }
 
@@ -1147,10 +1146,7 @@ static enum i40iw_status_code i40iw_alloc_set_mac_ipaddr(struct i40iw_device *iw
 	if (!status) {
 		status = i40iw_add_mac_ipaddr_entry(iwdev, macaddr,
 						    (u8)iwdev->mac_ip_table_idx);
-		if (!status)
-			status = i40iw_add_mac_ipaddr_entry(iwdev, macaddr,
-							    (u8)iwdev->mac_ip_table_idx);
-		else
+		if (status)
 			i40iw_del_macip_entry(iwdev, (u8)iwdev->mac_ip_table_idx);
 	}
 	return status;
@@ -1165,7 +1161,7 @@ static void i40iw_add_ipv6_addr(struct i40iw_device *iwdev)
 	struct net_device *ip_dev;
 	struct inet6_dev *idev;
 	struct inet6_ifaddr *ifp;
-	__be32 local_ipaddr6[4];
+	u32 local_ipaddr6[4];
 
 	rcu_read_lock();
 	for_each_netdev_rcu(&init_net, ip_dev) {
@@ -1512,6 +1508,7 @@ static enum i40iw_status_code i40iw_setup_init_state(struct i40iw_handler *hdl,
 	    I40IW_HMC_PROFILE_DEFAULT;
 	iwdev->max_rdma_vfs =
 		(iwdev->resource_profile != I40IW_HMC_PROFILE_DEFAULT) ?  max_rdma_vfs : 0;
+	iwdev->max_enabled_vfs = iwdev->max_rdma_vfs;
 	iwdev->netdev = ldev->netdev;
 	hdl->client = client;
 	iwdev->mss = (!ldev->params.mtu) ? I40IW_DEFAULT_MSS : ldev->params.mtu - I40IW_MTU_TO_MSS;
@@ -1531,7 +1528,10 @@ static enum i40iw_status_code i40iw_setup_init_state(struct i40iw_handler *hdl,
 		goto exit;
 	iwdev->obj_next = iwdev->obj_mem;
 	iwdev->push_mode = push_mode;
+
 	init_waitqueue_head(&iwdev->vchnl_waitq);
+	init_waitqueue_head(&dev->vf_reqs);
+
 	status = i40iw_initialize_dev(iwdev, ldev);
 exit:
 	if (status) {
@@ -1710,7 +1710,6 @@ static void i40iw_vf_reset(struct i40e_info *ldev, struct i40e_client *client, u
 	for (i = 0; i < I40IW_MAX_PE_ENABLED_VF_COUNT; i++) {
 		if (!dev->vf_dev[i] || (dev->vf_dev[i]->vf_id != vf_id))
 			continue;
-
 		/* free all resources allocated on behalf of vf */
 		tmp_vfdev = dev->vf_dev[i];
 		spin_lock_irqsave(&dev->dev_pestat.stats_lock, flags);
@@ -1819,8 +1818,6 @@ static int i40iw_virtchnl_receive(struct i40e_info *ldev,
 	dev = &hdl->device.sc_dev;
 	iwdev = dev->back_dev;
 
-	i40iw_debug(dev, I40IW_DEBUG_VIRT, "msg %p, message length %u\n", msg, len);
-
 	if (dev->vchnl_if.vchnl_recv) {
 		ret_code = dev->vchnl_if.vchnl_recv(dev, vf_id, msg, len);
 		if (!dev->is_pf) {
@@ -1831,6 +1828,39 @@ static int i40iw_virtchnl_receive(struct i40e_info *ldev,
 	return ret_code;
 }
 
+/**
+ * i40iw_vf_clear_to_send - wait to send virtual channel message
+ * @dev: iwarp device *
+ * Wait for until virtual channel is clear
+ * before sending the next message
+ *
+ * Returns false if error
+ * Returns true if clear to send
+ */
+bool i40iw_vf_clear_to_send(struct i40iw_sc_dev *dev)
+{
+	struct i40iw_device *iwdev;
+	wait_queue_t wait;
+
+	iwdev = dev->back_dev;
+
+	if (!wq_has_sleeper(&dev->vf_reqs) &&
+	    (atomic_read(&iwdev->vchnl_msgs) == 0))
+		return true; /* virtual channel is clear */
+
+	init_wait(&wait);
+	add_wait_queue_exclusive(&dev->vf_reqs, &wait);
+
+	if (!wait_event_timeout(dev->vf_reqs,
+				(atomic_read(&iwdev->vchnl_msgs) == 0),
+				I40IW_VCHNL_EVENT_TIMEOUT))
+		dev->vchnl_up = false;
+
+	remove_wait_queue(&dev->vf_reqs, &wait);
+
+	return dev->vchnl_up;
+}
+
 /**
  * i40iw_virtchnl_send - send a message through the virtual channel
  * @dev: iwarp device
@@ -1848,18 +1878,16 @@ static enum i40iw_status_code i40iw_virtchnl_send(struct i40iw_sc_dev *dev,
 {
 	struct i40iw_device *iwdev;
 	struct i40e_info *ldev;
-	enum i40iw_status_code ret_code = I40IW_ERR_BAD_PTR;
 
 	if (!dev || !dev->back_dev)
-		return ret_code;
+		return I40IW_ERR_BAD_PTR;
 
 	iwdev = dev->back_dev;
 	ldev = iwdev->ldev;
 
 	if (ldev && ldev->ops && ldev->ops->virtchnl_send)
-		ret_code = ldev->ops->virtchnl_send(ldev, &i40iw_client, vf_id, msg, len);
-
-	return ret_code;
+		return ldev->ops->virtchnl_send(ldev, &i40iw_client, vf_id, msg, len);
+	return I40IW_ERR_BAD_PTR;
 }
 
 /* client interface functions */

drivers/infiniband/hw/i40iw/i40iw_osdep.h

@@ -172,6 +172,7 @@ struct i40iw_hw;
 u8 __iomem *i40iw_get_hw_addr(void *dev);
 void i40iw_ieq_mpa_crc_ae(struct i40iw_sc_dev *dev, struct i40iw_sc_qp *qp);
 enum i40iw_status_code i40iw_vf_wait_vchnl_resp(struct i40iw_sc_dev *dev);
+bool i40iw_vf_clear_to_send(struct i40iw_sc_dev *dev);
 enum i40iw_status_code i40iw_ieq_check_mpacrc(struct shash_desc *desc, void *addr,
 					      u32 length, u32 value);
 struct i40iw_sc_qp *i40iw_ieq_get_qp(struct i40iw_sc_dev *dev, struct i40iw_puda_buf *buf);

drivers/infiniband/hw/i40iw/i40iw_pble.c

@@ -404,13 +404,14 @@ static enum i40iw_status_code add_pble_pool(struct i40iw_sc_dev *dev,
 			sd_entry->u.pd_table.pd_page_addr.pa : sd_entry->u.bp.addr.pa;
 	if (sd_entry->valid)
 		return 0;
-	if (dev->is_pf)
+	if (dev->is_pf) {
 		ret_code = i40iw_hmc_sd_one(dev, hmc_info->hmc_fn_id,
 					    sd_reg_val, idx->sd_idx,
 					    sd_entry->entry_type, true);
-	if (ret_code) {
-		i40iw_pr_err("cqp cmd failed for sd (pbles)\n");
-		goto error;
+		if (ret_code) {
+			i40iw_pr_err("cqp cmd failed for sd (pbles)\n");
+			goto error;
+		}
 	}
 
 	sd_entry->valid = true;

drivers/infiniband/hw/i40iw/i40iw_puda.c

@@ -1194,7 +1194,7 @@ static enum i40iw_status_code i40iw_ieq_process_buf(struct i40iw_puda_rsrc *ieq,
 
 	ioffset = (u16)(buf->data - (u8 *)buf->mem.va);
 	while (datalen) {
-		fpdu_len = i40iw_ieq_get_fpdu_length(ntohs(*(u16 *)datap));
+		fpdu_len = i40iw_ieq_get_fpdu_length(ntohs(*(__be16 *)datap));
 		if (fpdu_len > pfpdu->max_fpdu_data) {
 			i40iw_debug(ieq->dev, I40IW_DEBUG_IEQ,
 				    "%s: error bad fpdu_len\n", __func__);

drivers/infiniband/hw/i40iw/i40iw_status.h

@@ -95,6 +95,7 @@ enum i40iw_status_code {
 	I40IW_ERR_INVALID_MAC_ADDR = -65,
 	I40IW_ERR_BAD_STAG      = -66,
 	I40IW_ERR_CQ_COMPL_ERROR = -67,
+	I40IW_ERR_QUEUE_DESTROYED = -68
 
 };
 #endif

drivers/infiniband/hw/i40iw/i40iw_type.h

@@ -479,16 +479,17 @@ struct i40iw_sc_dev {
 	struct i40iw_virt_mem ieq_mem;
 	struct i40iw_puda_rsrc *ieq;
 
-	struct i40iw_vf_cqp_ops *iw_vf_cqp_ops;
+	const struct i40iw_vf_cqp_ops *iw_vf_cqp_ops;
 
 	struct i40iw_hmc_fpm_misc hmc_fpm_misc;
 	u16 qs_handle;
-	u32	debug_mask;
+	u32 debug_mask;
 	u16 exception_lan_queue;
 	u8 hmc_fn_id;
 	bool is_pf;
 	bool vchnl_up;
 	u8 vf_id;
+	wait_queue_head_t vf_reqs;
 	u64 cqp_cmd_stats[OP_SIZE_CQP_STAT_ARRAY];
 	struct i40iw_vchnl_vf_msg_buffer vchnl_vf_msg_buf;
 	u8 hw_rev;
@@ -889,8 +890,8 @@ struct i40iw_qhash_table_info {
 	u32 qp_num;
 	u32 dest_ip[4];
 	u32 src_ip[4];
-	u32 dest_port;
-	u32 src_port;
+	u16 dest_port;
+	u16 src_port;
 };
 
 struct i40iw_local_mac_ipaddr_entry_info {
@@ -1040,6 +1041,9 @@ struct i40iw_priv_qp_ops {
 	void (*qp_send_lsmm_nostag)(struct i40iw_sc_qp *, void *, u32);
 	void (*qp_send_rtt)(struct i40iw_sc_qp *, bool);
 	enum i40iw_status_code (*qp_post_wqe0)(struct i40iw_sc_qp *, u8);
+	enum i40iw_status_code (*iw_mr_fast_register)(struct i40iw_sc_qp *,
+						      struct i40iw_fast_reg_stag_info *,
+						      bool);
 };
 
 struct i40iw_priv_cq_ops {
@@ -1108,7 +1112,7 @@ struct i40iw_hmc_ops {
 	enum i40iw_status_code (*parse_fpm_query_buf)(u64 *, struct i40iw_hmc_info *,
 						      struct i40iw_hmc_fpm_misc *);
 	enum i40iw_status_code (*configure_iw_fpm)(struct i40iw_sc_dev *, u8);
-	enum i40iw_status_code (*parse_fpm_commit_buf)(u64 *, struct i40iw_hmc_obj_info *);
+	enum i40iw_status_code (*parse_fpm_commit_buf)(u64 *, struct i40iw_hmc_obj_info *, u32 *sd);
 	enum i40iw_status_code (*create_hmc_object)(struct i40iw_sc_dev *dev,
 						    struct i40iw_hmc_create_obj_info *);
 	enum i40iw_status_code (*del_hmc_object)(struct i40iw_sc_dev *dev,

drivers/infiniband/hw/i40iw/i40iw_uk.c

@@ -56,6 +56,9 @@ static enum i40iw_status_code i40iw_nop_1(struct i40iw_qp_uk *qp)
 
 	wqe_idx = I40IW_RING_GETCURRENT_HEAD(qp->sq_ring);
 	wqe = qp->sq_base[wqe_idx].elem;
+
+	qp->sq_wrtrk_array[wqe_idx].wqe_size = I40IW_QP_WQE_MIN_SIZE;
+
 	peek_head = (qp->sq_ring.head + 1) % qp->sq_ring.size;
 	wqe_0 = qp->sq_base[peek_head].elem;
 	if (peek_head)
@@ -130,7 +133,10 @@ static void i40iw_qp_ring_push_db(struct i40iw_qp_uk *qp, u32 wqe_idx)
  */
 u64 *i40iw_qp_get_next_send_wqe(struct i40iw_qp_uk *qp,
 				u32 *wqe_idx,
-				u8 wqe_size)
+				u8 wqe_size,
+				u32 total_size,
+				u64 wr_id
+				)
 {
 	u64 *wqe = NULL;
 	u64 wqe_ptr;
@@ -159,6 +165,17 @@ u64 *i40iw_qp_get_next_send_wqe(struct i40iw_qp_uk *qp,
 		if (!*wqe_idx)
 			qp->swqe_polarity = !qp->swqe_polarity;
 	}
+
+	if (((*wqe_idx & 3) == 1) && (wqe_size == I40IW_WQE_SIZE_64)) {
+		i40iw_nop_1(qp);
+		I40IW_RING_MOVE_HEAD(qp->sq_ring, ret_code);
+		if (ret_code)
+			return NULL;
+		*wqe_idx = I40IW_RING_GETCURRENT_HEAD(qp->sq_ring);
+		if (!*wqe_idx)
+			qp->swqe_polarity = !qp->swqe_polarity;
+	}
+
 	for (i = 0; i < wqe_size / I40IW_QP_WQE_MIN_SIZE; i++) {
 		I40IW_RING_MOVE_HEAD(qp->sq_ring, ret_code);
 		if (ret_code)
@@ -169,8 +186,15 @@ u64 *i40iw_qp_get_next_send_wqe(struct i40iw_qp_uk *qp,
 
 	peek_head = I40IW_RING_GETCURRENT_HEAD(qp->sq_ring);
 	wqe_0 = qp->sq_base[peek_head].elem;
-	if (peek_head & 0x3)
-		wqe_0[3] = LS_64(!qp->swqe_polarity, I40IWQPSQ_VALID);
+
+	if (((peek_head & 3) == 1) || ((peek_head & 3) == 3)) {
+		if (RS_64(wqe_0[3], I40IWQPSQ_VALID) != !qp->swqe_polarity)
+			wqe_0[3] = LS_64(!qp->swqe_polarity, I40IWQPSQ_VALID);
+	}
+
+	qp->sq_wrtrk_array[*wqe_idx].wrid = wr_id;
+	qp->sq_wrtrk_array[*wqe_idx].wr_len = total_size;
+	qp->sq_wrtrk_array[*wqe_idx].wqe_size = wqe_size;
 	return wqe;
 }
 
@@ -249,12 +273,9 @@ static enum i40iw_status_code i40iw_rdma_write(struct i40iw_qp_uk *qp,
 	if (ret_code)
 		return ret_code;
 
-	wqe = i40iw_qp_get_next_send_wqe(qp, &wqe_idx, wqe_size);
+	wqe = i40iw_qp_get_next_send_wqe(qp, &wqe_idx, wqe_size, total_size, info->wr_id);
 	if (!wqe)
 		return I40IW_ERR_QP_TOOMANY_WRS_POSTED;
-
-	qp->sq_wrtrk_array[wqe_idx].wrid = info->wr_id;
-	qp->sq_wrtrk_array[wqe_idx].wr_len = total_size;
 	set_64bit_val(wqe, 16,
 		      LS_64(op_info->rem_addr.tag_off, I40IWQPSQ_FRAG_TO));
 	if (!op_info->rem_addr.stag)
@@ -309,12 +330,9 @@ static enum i40iw_status_code i40iw_rdma_read(struct i40iw_qp_uk *qp,
 	ret_code = i40iw_fragcnt_to_wqesize_sq(1, &wqe_size);
 	if (ret_code)
 		return ret_code;
-	wqe = i40iw_qp_get_next_send_wqe(qp, &wqe_idx, wqe_size);
+	wqe = i40iw_qp_get_next_send_wqe(qp, &wqe_idx, wqe_size, op_info->lo_addr.len, info->wr_id);
 	if (!wqe)
 		return I40IW_ERR_QP_TOOMANY_WRS_POSTED;
-
-	qp->sq_wrtrk_array[wqe_idx].wrid = info->wr_id;
-	qp->sq_wrtrk_array[wqe_idx].wr_len = op_info->lo_addr.len;
 	local_fence |= info->local_fence;
 
 	set_64bit_val(wqe, 16, LS_64(op_info->rem_addr.tag_off, I40IWQPSQ_FRAG_TO));
@@ -366,13 +384,11 @@ static enum i40iw_status_code i40iw_send(struct i40iw_qp_uk *qp,
 	if (ret_code)
 		return ret_code;
 
-	wqe = i40iw_qp_get_next_send_wqe(qp, &wqe_idx, wqe_size);
+	wqe = i40iw_qp_get_next_send_wqe(qp, &wqe_idx, wqe_size, total_size, info->wr_id);
 	if (!wqe)
 		return I40IW_ERR_QP_TOOMANY_WRS_POSTED;
 
 	read_fence |= info->read_fence;
-	qp->sq_wrtrk_array[wqe_idx].wrid = info->wr_id;
-	qp->sq_wrtrk_array[wqe_idx].wr_len = total_size;
 	set_64bit_val(wqe, 16, 0);
 	header = LS_64(stag_to_inv, I40IWQPSQ_REMSTAG) |
 		 LS_64(info->op_type, I40IWQPSQ_OPCODE) |
@@ -427,13 +443,11 @@ static enum i40iw_status_code i40iw_inline_rdma_write(struct i40iw_qp_uk *qp,
 	if (ret_code)
 		return ret_code;
 
-	wqe = i40iw_qp_get_next_send_wqe(qp, &wqe_idx, wqe_size);
+	wqe = i40iw_qp_get_next_send_wqe(qp, &wqe_idx, wqe_size, op_info->len, info->wr_id);
 	if (!wqe)
 		return I40IW_ERR_QP_TOOMANY_WRS_POSTED;
 
 	read_fence |= info->read_fence;
-	qp->sq_wrtrk_array[wqe_idx].wrid = info->wr_id;
-	qp->sq_wrtrk_array[wqe_idx].wr_len = op_info->len;
 	set_64bit_val(wqe, 16,
 		      LS_64(op_info->rem_addr.tag_off, I40IWQPSQ_FRAG_TO));
 
@@ -507,14 +521,11 @@ static enum i40iw_status_code i40iw_inline_send(struct i40iw_qp_uk *qp,
 	if (ret_code)
 		return ret_code;
 
-	wqe = i40iw_qp_get_next_send_wqe(qp, &wqe_idx, wqe_size);
+	wqe = i40iw_qp_get_next_send_wqe(qp, &wqe_idx, wqe_size, op_info->len, info->wr_id);
 	if (!wqe)
 		return I40IW_ERR_QP_TOOMANY_WRS_POSTED;
 
 	read_fence |= info->read_fence;
-
-	qp->sq_wrtrk_array[wqe_idx].wrid = info->wr_id;
-	qp->sq_wrtrk_array[wqe_idx].wr_len = op_info->len;
 	header = LS_64(stag_to_inv, I40IWQPSQ_REMSTAG) |
 	    LS_64(info->op_type, I40IWQPSQ_OPCODE) |
 	    LS_64(op_info->len, I40IWQPSQ_INLINEDATALEN) |
@@ -574,12 +585,9 @@ static enum i40iw_status_code i40iw_stag_local_invalidate(struct i40iw_qp_uk *qp
 	op_info = &info->op.inv_local_stag;
 	local_fence = info->local_fence;
 
-	wqe = i40iw_qp_get_next_send_wqe(qp, &wqe_idx, I40IW_QP_WQE_MIN_SIZE);
+	wqe = i40iw_qp_get_next_send_wqe(qp, &wqe_idx, I40IW_QP_WQE_MIN_SIZE, 0, info->wr_id);
 	if (!wqe)
 		return I40IW_ERR_QP_TOOMANY_WRS_POSTED;
-
-	qp->sq_wrtrk_array[wqe_idx].wrid = info->wr_id;
-	qp->sq_wrtrk_array[wqe_idx].wr_len = 0;
 	set_64bit_val(wqe, 0, 0);
 	set_64bit_val(wqe, 8,
 		      LS_64(op_info->target_stag, I40IWQPSQ_LOCSTAG));
@@ -619,12 +627,9 @@ static enum i40iw_status_code i40iw_mw_bind(struct i40iw_qp_uk *qp,
 	op_info = &info->op.bind_window;
 
 	local_fence |= info->local_fence;
-	wqe = i40iw_qp_get_next_send_wqe(qp, &wqe_idx, I40IW_QP_WQE_MIN_SIZE);
+	wqe = i40iw_qp_get_next_send_wqe(qp, &wqe_idx, I40IW_QP_WQE_MIN_SIZE, 0, info->wr_id);
 	if (!wqe)
 		return I40IW_ERR_QP_TOOMANY_WRS_POSTED;
-
-	qp->sq_wrtrk_array[wqe_idx].wrid = info->wr_id;
-	qp->sq_wrtrk_array[wqe_idx].wr_len = 0;
 	set_64bit_val(wqe, 0, (uintptr_t)op_info->va);
 	set_64bit_val(wqe, 8,
 		      LS_64(op_info->mr_stag, I40IWQPSQ_PARENTMRSTAG) |
@@ -760,7 +765,7 @@ static enum i40iw_status_code i40iw_cq_poll_completion(struct i40iw_cq_uk *cq,
 	enum i40iw_status_code ret_code2 = 0;
 	bool move_cq_head = true;
 	u8 polarity;
-	u8 addl_frag_cnt, addl_wqes = 0;
+	u8 addl_wqes = 0;
 
 	if (cq->avoid_mem_cflct)
 		cqe = (u64 *)I40IW_GET_CURRENT_EXTENDED_CQ_ELEMENT(cq);
@@ -797,6 +802,10 @@ static enum i40iw_status_code i40iw_cq_poll_completion(struct i40iw_cq_uk *cq,
 	info->is_srq = (bool)RS_64(qword3, I40IWCQ_SRQ);
 
 	qp = (struct i40iw_qp_uk *)(unsigned long)comp_ctx;
+	if (!qp) {
+		ret_code = I40IW_ERR_QUEUE_DESTROYED;
+		goto exit;
+	}
 	wqe_idx = (u32)RS_64(qword3, I40IW_CQ_WQEIDX);
 	info->qp_handle = (i40iw_qp_handle)(unsigned long)qp;
 
@@ -827,11 +836,8 @@ static enum i40iw_status_code i40iw_cq_poll_completion(struct i40iw_cq_uk *cq,
 			info->op_type = (u8)RS_64(qword3, I40IWCQ_OP);
 			sw_wqe = qp->sq_base[wqe_idx].elem;
 			get_64bit_val(sw_wqe, 24, &wqe_qword);
-			addl_frag_cnt =
-			    (u8)RS_64(wqe_qword, I40IWQPSQ_ADDFRAGCNT);
-			i40iw_fragcnt_to_wqesize_sq(addl_frag_cnt + 1, &addl_wqes);
 
-			addl_wqes = (addl_wqes / I40IW_QP_WQE_MIN_SIZE);
+			addl_wqes = qp->sq_wrtrk_array[wqe_idx].wqe_size / I40IW_QP_WQE_MIN_SIZE;
 			I40IW_RING_SET_TAIL(qp->sq_ring, (wqe_idx + addl_wqes));
 		} else {
 			do {
@@ -843,9 +849,7 @@ static enum i40iw_status_code i40iw_cq_poll_completion(struct i40iw_cq_uk *cq,
 				get_64bit_val(sw_wqe, 24, &wqe_qword);
 				op_type = (u8)RS_64(wqe_qword, I40IWQPSQ_OPCODE);
 				info->op_type = op_type;
-				addl_frag_cnt = (u8)RS_64(wqe_qword, I40IWQPSQ_ADDFRAGCNT);
-				i40iw_fragcnt_to_wqesize_sq(addl_frag_cnt + 1, &addl_wqes);
-				addl_wqes = (addl_wqes / I40IW_QP_WQE_MIN_SIZE);
+				addl_wqes = qp->sq_wrtrk_array[tail].wqe_size / I40IW_QP_WQE_MIN_SIZE;
 				I40IW_RING_SET_TAIL(qp->sq_ring, (tail + addl_wqes));
 				if (op_type != I40IWQP_OP_NOP) {
 					info->wr_id = qp->sq_wrtrk_array[tail].wrid;
@@ -859,6 +863,7 @@ static enum i40iw_status_code i40iw_cq_poll_completion(struct i40iw_cq_uk *cq,
 
 	ret_code = 0;
 
+exit:
 	if (!ret_code &&
 	    (info->comp_status == I40IW_COMPL_STATUS_FLUSHED))
 		if (pring && (I40IW_RING_MORE_WORK(*pring)))
@@ -893,19 +898,21 @@ static enum i40iw_status_code i40iw_cq_poll_completion(struct i40iw_cq_uk *cq,
  * i40iw_get_wqe_shift - get shift count for maximum wqe size
  * @wqdepth: depth of wq required.
  * @sge: Maximum Scatter Gather Elements wqe
+ * @inline_data: Maximum inline data size
  * @shift: Returns the shift needed based on sge
  *
- * Shift can be used to left shift the wqe size based on sge.
- * If sge, == 1, shift =0 (wqe_size of 32 bytes), for sge=2 and 3, shift =1
- * (64 bytes wqes) and 2 otherwise (128 bytes wqe).
+ * Shift can be used to left shift the wqe size based on number of SGEs and inlind data size.
+ * For 1 SGE or inline data <= 16, shift = 0 (wqe size of 32 bytes).
+ * For 2 or 3 SGEs or inline data <= 48, shift = 1 (wqe size of 64 bytes).
+ * Shift of 2 otherwise (wqe size of 128 bytes).
  */
-enum i40iw_status_code i40iw_get_wqe_shift(u32 wqdepth, u8 sge, u8 *shift)
+enum i40iw_status_code i40iw_get_wqe_shift(u32 wqdepth, u32 sge, u32 inline_data, u8 *shift)
 {
 	u32 size;
 
 	*shift = 0;
-	if (sge > 1)
-		*shift = (sge < 4) ? 1 : 2;
+	if (sge > 1 || inline_data > 16)
+		*shift = (sge < 4 && inline_data <= 48) ? 1 : 2;
 
 	/* check if wqdepth is multiple of 2 or not */
 
@@ -968,11 +975,11 @@ enum i40iw_status_code i40iw_qp_uk_init(struct i40iw_qp_uk *qp,
 
 	if (info->max_rq_frag_cnt > I40IW_MAX_WQ_FRAGMENT_COUNT)
 		return I40IW_ERR_INVALID_FRAG_COUNT;
-	ret_code = i40iw_get_wqe_shift(info->sq_size, info->max_sq_frag_cnt, &sqshift);
+	ret_code = i40iw_get_wqe_shift(info->sq_size, info->max_sq_frag_cnt, info->max_inline_data, &sqshift);
 	if (ret_code)
 		return ret_code;
 
-	ret_code = i40iw_get_wqe_shift(info->rq_size, info->max_rq_frag_cnt, &rqshift);
+	ret_code = i40iw_get_wqe_shift(info->rq_size, info->max_rq_frag_cnt, 0, &rqshift);
 	if (ret_code)
 		return ret_code;
 
@@ -1097,12 +1104,9 @@ enum i40iw_status_code i40iw_nop(struct i40iw_qp_uk *qp,
 	u64 header, *wqe;
 	u32 wqe_idx;
 
-	wqe = i40iw_qp_get_next_send_wqe(qp, &wqe_idx, I40IW_QP_WQE_MIN_SIZE);
+	wqe = i40iw_qp_get_next_send_wqe(qp, &wqe_idx, I40IW_QP_WQE_MIN_SIZE, 0, wr_id);
 	if (!wqe)
 		return I40IW_ERR_QP_TOOMANY_WRS_POSTED;
-
-	qp->sq_wrtrk_array[wqe_idx].wrid = wr_id;
-	qp->sq_wrtrk_array[wqe_idx].wr_len = 0;
 	set_64bit_val(wqe, 0, 0);
 	set_64bit_val(wqe, 8, 0);
 	set_64bit_val(wqe, 16, 0);
@@ -1125,7 +1129,7 @@ enum i40iw_status_code i40iw_nop(struct i40iw_qp_uk *qp,
  * @frag_cnt: number of fragments
  * @wqe_size: size of sq wqe returned
  */
-enum i40iw_status_code i40iw_fragcnt_to_wqesize_sq(u8 frag_cnt, u8 *wqe_size)
+enum i40iw_status_code i40iw_fragcnt_to_wqesize_sq(u32 frag_cnt, u8 *wqe_size)
 {
 	switch (frag_cnt) {
 	case 0:
@@ -1156,7 +1160,7 @@ enum i40iw_status_code i40iw_fragcnt_to_wqesize_sq(u8 frag_cnt, u8 *wqe_size)
  * @frag_cnt: number of fragments
  * @wqe_size: size of rq wqe returned
  */
-enum i40iw_status_code i40iw_fragcnt_to_wqesize_rq(u8 frag_cnt, u8 *wqe_size)
+enum i40iw_status_code i40iw_fragcnt_to_wqesize_rq(u32 frag_cnt, u8 *wqe_size)
 {
 	switch (frag_cnt) {
 	case 0:

drivers/infiniband/hw/i40iw/i40iw_user.h

@@ -61,7 +61,7 @@ enum i40iw_device_capabilities_const {
 	I40IW_MAX_CQ_SIZE =			1048575,
 	I40IW_MAX_AEQ_ALLOCATE_COUNT =		255,
 	I40IW_DB_ID_ZERO =			0,
-	I40IW_MAX_WQ_FRAGMENT_COUNT =		6,
+	I40IW_MAX_WQ_FRAGMENT_COUNT =		3,
 	I40IW_MAX_SGE_RD =			1,
 	I40IW_MAX_OUTBOUND_MESSAGE_SIZE =	2147483647,
 	I40IW_MAX_INBOUND_MESSAGE_SIZE =	2147483647,
@@ -70,8 +70,8 @@ enum i40iw_device_capabilities_const {
 	I40IW_MAX_VF_FPM_ID =			47,
 	I40IW_MAX_VF_PER_PF =			127,
 	I40IW_MAX_SQ_PAYLOAD_SIZE =		2145386496,
-	I40IW_MAX_INLINE_DATA_SIZE =		112,
-	I40IW_MAX_PUSHMODE_INLINE_DATA_SIZE =	112,
+	I40IW_MAX_INLINE_DATA_SIZE =		48,
+	I40IW_MAX_PUSHMODE_INLINE_DATA_SIZE =	48,
 	I40IW_MAX_IRD_SIZE =			32,
 	I40IW_QPCTX_ENCD_MAXIRD =		3,
 	I40IW_MAX_WQ_ENTRIES =			2048,
@@ -102,6 +102,8 @@ enum i40iw_device_capabilities_const {
 
 #define I40IW_STAG_INDEX_FROM_STAG(stag)    (((stag) && 0xFFFFFF00) >> 8)
 
+#define	I40IW_MAX_MR_SIZE	0x10000000000L
+
 struct i40iw_qp_uk;
 struct i40iw_cq_uk;
 struct i40iw_srq_uk;
@@ -198,7 +200,7 @@ enum i40iw_completion_notify {
 
 struct i40iw_post_send {
 	i40iw_sgl sg_list;
-	u8 num_sges;
+	u32 num_sges;
 };
 
 struct i40iw_post_inline_send {
@@ -220,7 +222,7 @@ struct i40iw_post_inline_send_w_inv {
 
 struct i40iw_rdma_write {
 	i40iw_sgl lo_sg_list;
-	u8 num_lo_sges;
+	u32 num_lo_sges;
 	struct i40iw_sge rem_addr;
 };
 
@@ -345,7 +347,9 @@ struct i40iw_dev_uk {
 
 struct i40iw_sq_uk_wr_trk_info {
 	u64 wrid;
-	u64 wr_len;
+	u32 wr_len;
+	u8 wqe_size;
+	u8 reserved[3];
 };
 
 struct i40iw_qp_quanta {
@@ -367,6 +371,8 @@ struct i40iw_qp_uk {
 	u32 qp_id;
 	u32 sq_size;
 	u32 rq_size;
+	u32 max_sq_frag_cnt;
+	u32 max_rq_frag_cnt;
 	struct i40iw_qp_uk_ops ops;
 	bool use_srq;
 	u8 swqe_polarity;
@@ -374,8 +380,6 @@ struct i40iw_qp_uk {
 	u8 rwqe_polarity;
 	u8 rq_wqe_size;
 	u8 rq_wqe_size_multiplier;
-	u8 max_sq_frag_cnt;
-	u8 max_rq_frag_cnt;
 	bool deferred_flag;
 };
 
@@ -404,8 +408,9 @@ struct i40iw_qp_uk_init_info {
 	u32 qp_id;
 	u32 sq_size;
 	u32 rq_size;
-	u8 max_sq_frag_cnt;
-	u8 max_rq_frag_cnt;
+	u32 max_sq_frag_cnt;
+	u32 max_rq_frag_cnt;
+	u32 max_inline_data;
 
 };
 
@@ -422,7 +427,10 @@ void i40iw_device_init_uk(struct i40iw_dev_uk *dev);
 
 void i40iw_qp_post_wr(struct i40iw_qp_uk *qp);
 u64 *i40iw_qp_get_next_send_wqe(struct i40iw_qp_uk *qp, u32 *wqe_idx,
-				u8 wqe_size);
+				u8 wqe_size,
+				u32 total_size,
+				u64 wr_id
+				);
 u64 *i40iw_qp_get_next_recv_wqe(struct i40iw_qp_uk *qp, u32 *wqe_idx);
 u64 *i40iw_qp_get_next_srq_wqe(struct i40iw_srq_uk *srq, u32 *wqe_idx);
 
@@ -434,9 +442,9 @@ enum i40iw_status_code i40iw_qp_uk_init(struct i40iw_qp_uk *qp,
 void i40iw_clean_cq(void *queue, struct i40iw_cq_uk *cq);
 enum i40iw_status_code i40iw_nop(struct i40iw_qp_uk *qp, u64 wr_id,
 				 bool signaled, bool post_sq);
-enum i40iw_status_code i40iw_fragcnt_to_wqesize_sq(u8 frag_cnt, u8 *wqe_size);
-enum i40iw_status_code i40iw_fragcnt_to_wqesize_rq(u8 frag_cnt, u8 *wqe_size);
+enum i40iw_status_code i40iw_fragcnt_to_wqesize_sq(u32 frag_cnt, u8 *wqe_size);
+enum i40iw_status_code i40iw_fragcnt_to_wqesize_rq(u32 frag_cnt, u8 *wqe_size);
 enum i40iw_status_code i40iw_inline_data_size_to_wqesize(u32 data_size,
 							 u8 *wqe_size);
-enum i40iw_status_code i40iw_get_wqe_shift(u32 wqdepth, u8 sge, u8 *shift);
+enum i40iw_status_code i40iw_get_wqe_shift(u32 wqdepth, u32 sge, u32 inline_data, u8 *shift);
 #endif

drivers/infiniband/hw/i40iw/i40iw_utils.c

@@ -59,7 +59,7 @@
  * @action: modify, delete or add
  */
 int i40iw_arp_table(struct i40iw_device *iwdev,
-		    __be32 *ip_addr,
+		    u32 *ip_addr,
 		    bool ipv4,
 		    u8 *mac_addr,
 		    u32 action)
@@ -152,7 +152,7 @@ int i40iw_inetaddr_event(struct notifier_block *notifier,
 	struct net_device *upper_dev;
 	struct i40iw_device *iwdev;
 	struct i40iw_handler *hdl;
-	__be32 local_ipaddr;
+	u32 local_ipaddr;
 
 	hdl = i40iw_find_netdev(event_netdev);
 	if (!hdl)
@@ -167,11 +167,10 @@ int i40iw_inetaddr_event(struct notifier_block *notifier,
 	switch (event) {
 	case NETDEV_DOWN:
 		if (upper_dev)
-			local_ipaddr =
-				((struct in_device *)upper_dev->ip_ptr)->ifa_list->ifa_address;
+			local_ipaddr = ntohl(
+				((struct in_device *)upper_dev->ip_ptr)->ifa_list->ifa_address);
 		else
-			local_ipaddr = ifa->ifa_address;
-		local_ipaddr = ntohl(local_ipaddr);
+			local_ipaddr = ntohl(ifa->ifa_address);
 		i40iw_manage_arp_cache(iwdev,
 				       netdev->dev_addr,
 				       &local_ipaddr,
@@ -180,11 +179,10 @@ int i40iw_inetaddr_event(struct notifier_block *notifier,
 		return NOTIFY_OK;
 	case NETDEV_UP:
 		if (upper_dev)
-			local_ipaddr =
-				((struct in_device *)upper_dev->ip_ptr)->ifa_list->ifa_address;
+			local_ipaddr = ntohl(
+				((struct in_device *)upper_dev->ip_ptr)->ifa_list->ifa_address);
 		else
-			local_ipaddr = ifa->ifa_address;
-		local_ipaddr = ntohl(local_ipaddr);
+			local_ipaddr = ntohl(ifa->ifa_address);
 		i40iw_manage_arp_cache(iwdev,
 				       netdev->dev_addr,
 				       &local_ipaddr,
@@ -194,12 +192,11 @@ int i40iw_inetaddr_event(struct notifier_block *notifier,
 	case NETDEV_CHANGEADDR:
 		/* Add the address to the IP table */
 		if (upper_dev)
-			local_ipaddr =
-				((struct in_device *)upper_dev->ip_ptr)->ifa_list->ifa_address;
+			local_ipaddr = ntohl(
+				((struct in_device *)upper_dev->ip_ptr)->ifa_list->ifa_address);
 		else
-			local_ipaddr = ifa->ifa_address;
+			local_ipaddr = ntohl(ifa->ifa_address);
 
-		local_ipaddr = ntohl(local_ipaddr);
 		i40iw_manage_arp_cache(iwdev,
 				       netdev->dev_addr,
 				       &local_ipaddr,
@@ -227,7 +224,7 @@ int i40iw_inet6addr_event(struct notifier_block *notifier,
 	struct net_device *netdev;
 	struct i40iw_device *iwdev;
 	struct i40iw_handler *hdl;
-	__be32 local_ipaddr6[4];
+	u32 local_ipaddr6[4];
 
 	hdl = i40iw_find_netdev(event_netdev);
 	if (!hdl)
@@ -506,14 +503,19 @@ void i40iw_rem_ref(struct ib_qp *ibqp)
 	struct cqp_commands_info *cqp_info;
 	struct i40iw_device *iwdev;
 	u32 qp_num;
+	unsigned long flags;
 
 	iwqp = to_iwqp(ibqp);
-	if (!atomic_dec_and_test(&iwqp->refcount))
-		return;
-
 	iwdev = iwqp->iwdev;
+	spin_lock_irqsave(&iwdev->qptable_lock, flags);
+	if (!atomic_dec_and_test(&iwqp->refcount)) {
+		spin_unlock_irqrestore(&iwdev->qptable_lock, flags);
+		return;
+	}
+
 	qp_num = iwqp->ibqp.qp_num;
 	iwdev->qp_table[qp_num] = NULL;
+	spin_unlock_irqrestore(&iwdev->qptable_lock, flags);
 	cqp_request = i40iw_get_cqp_request(&iwdev->cqp, false);
 	if (!cqp_request)
 		return;
@@ -985,21 +987,24 @@ enum i40iw_status_code i40iw_cqp_commit_fpm_values_cmd(struct i40iw_sc_dev *dev,
 enum i40iw_status_code i40iw_vf_wait_vchnl_resp(struct i40iw_sc_dev *dev)
 {
 	struct i40iw_device *iwdev = dev->back_dev;
-	enum i40iw_status_code err_code = 0;
 	int timeout_ret;
 
 	i40iw_debug(dev, I40IW_DEBUG_VIRT, "%s[%u] dev %p, iwdev %p\n",
 		    __func__, __LINE__, dev, iwdev);
-	atomic_add(2, &iwdev->vchnl_msgs);
+
+	atomic_set(&iwdev->vchnl_msgs, 2);
 	timeout_ret = wait_event_timeout(iwdev->vchnl_waitq,
 					 (atomic_read(&iwdev->vchnl_msgs) == 1),
 					 I40IW_VCHNL_EVENT_TIMEOUT);
 	atomic_dec(&iwdev->vchnl_msgs);
 	if (!timeout_ret) {
 		i40iw_pr_err("virt channel completion timeout = 0x%x\n", timeout_ret);
-		err_code = I40IW_ERR_TIMEOUT;
+		atomic_set(&iwdev->vchnl_msgs, 0);
+		dev->vchnl_up = false;
+		return I40IW_ERR_TIMEOUT;
 	}
-	return err_code;
+	wake_up(&dev->vf_reqs);
+	return 0;
 }
 
 /**

drivers/infiniband/hw/i40iw/i40iw_verbs.c

@@ -63,8 +63,8 @@ static int i40iw_query_device(struct ib_device *ibdev,
 	ether_addr_copy((u8 *)&props->sys_image_guid, iwdev->netdev->dev_addr);
 	props->fw_ver = I40IW_FW_VERSION;
 	props->device_cap_flags = iwdev->device_cap_flags;
-	props->vendor_id = iwdev->vendor_id;
-	props->vendor_part_id = iwdev->vendor_part_id;
+	props->vendor_id = iwdev->ldev->pcidev->vendor;
+	props->vendor_part_id = iwdev->ldev->pcidev->device;
 	props->hw_ver = (u32)iwdev->sc_dev.hw_rev;
 	props->max_mr_size = I40IW_MAX_OUTBOUND_MESSAGE_SIZE;
 	props->max_qp = iwdev->max_qp;
@@ -74,7 +74,7 @@ static int i40iw_query_device(struct ib_device *ibdev,
 	props->max_cqe = iwdev->max_cqe;
 	props->max_mr = iwdev->max_mr;
 	props->max_pd = iwdev->max_pd;
-	props->max_sge_rd = 1;
+	props->max_sge_rd = I40IW_MAX_SGE_RD;
 	props->max_qp_rd_atom = I40IW_MAX_IRD_SIZE;
 	props->max_qp_init_rd_atom = props->max_qp_rd_atom;
 	props->atomic_cap = IB_ATOMIC_NONE;
@@ -120,7 +120,7 @@ static int i40iw_query_port(struct ib_device *ibdev,
 	props->pkey_tbl_len = 1;
 	props->active_width = IB_WIDTH_4X;
 	props->active_speed = 1;
-	props->max_msg_sz = 0x80000000;
+	props->max_msg_sz = I40IW_MAX_OUTBOUND_MESSAGE_SIZE;
 	return 0;
 }
 
@@ -437,7 +437,6 @@ void i40iw_free_qp_resources(struct i40iw_device *iwdev,
 	kfree(iwqp->kqp.wrid_mem);
 	iwqp->kqp.wrid_mem = NULL;
 	kfree(iwqp->allocated_buffer);
-	iwqp->allocated_buffer = NULL;
 }
 
 /**
@@ -521,14 +520,12 @@ static int i40iw_setup_kmode_qp(struct i40iw_device *iwdev,
 	enum i40iw_status_code status;
 	struct i40iw_qp_uk_init_info *ukinfo = &info->qp_uk_init_info;
 
-	ukinfo->max_sq_frag_cnt = I40IW_MAX_WQ_FRAGMENT_COUNT;
-
 	sq_size = i40iw_qp_roundup(ukinfo->sq_size + 1);
 	rq_size = i40iw_qp_roundup(ukinfo->rq_size + 1);
 
-	status = i40iw_get_wqe_shift(sq_size, ukinfo->max_sq_frag_cnt, &sqshift);
+	status = i40iw_get_wqe_shift(sq_size, ukinfo->max_sq_frag_cnt, ukinfo->max_inline_data, &sqshift);
 	if (!status)
-		status = i40iw_get_wqe_shift(rq_size, ukinfo->max_rq_frag_cnt, &rqshift);
+		status = i40iw_get_wqe_shift(rq_size, ukinfo->max_rq_frag_cnt, 0, &rqshift);
 
 	if (status)
 		return -ENOSYS;
@@ -609,6 +606,9 @@ static struct ib_qp *i40iw_create_qp(struct ib_pd *ibpd,
 	if (init_attr->cap.max_inline_data > I40IW_MAX_INLINE_DATA_SIZE)
 		init_attr->cap.max_inline_data = I40IW_MAX_INLINE_DATA_SIZE;
 
+	if (init_attr->cap.max_send_sge > I40IW_MAX_WQ_FRAGMENT_COUNT)
+		init_attr->cap.max_send_sge = I40IW_MAX_WQ_FRAGMENT_COUNT;
+
 	memset(&init_info, 0, sizeof(init_info));
 
 	sq_size = init_attr->cap.max_send_wr;
@@ -618,6 +618,7 @@ static struct ib_qp *i40iw_create_qp(struct ib_pd *ibpd,
 	init_info.qp_uk_init_info.rq_size = rq_size;
 	init_info.qp_uk_init_info.max_sq_frag_cnt = init_attr->cap.max_send_sge;
 	init_info.qp_uk_init_info.max_rq_frag_cnt = init_attr->cap.max_recv_sge;
+	init_info.qp_uk_init_info.max_inline_data = init_attr->cap.max_inline_data;
 
 	mem = kzalloc(sizeof(*iwqp), GFP_KERNEL);
 	if (!mem)
@@ -722,8 +723,10 @@ static struct ib_qp *i40iw_create_qp(struct ib_pd *ibpd,
 	iwarp_info = &iwqp->iwarp_info;
 	iwarp_info->rd_enable = true;
 	iwarp_info->wr_rdresp_en = true;
-	if (!iwqp->user_mode)
+	if (!iwqp->user_mode) {
+		iwarp_info->fast_reg_en = true;
 		iwarp_info->priv_mode_en = true;
+	}
 	iwarp_info->ddp_ver = 1;
 	iwarp_info->rdmap_ver = 1;
 
@@ -784,6 +787,8 @@ static struct ib_qp *i40iw_create_qp(struct ib_pd *ibpd,
 			return ERR_PTR(err_code);
 		}
 	}
+	init_completion(&iwqp->sq_drained);
+	init_completion(&iwqp->rq_drained);
 
 	return &iwqp->ibqp;
 error:
@@ -1443,6 +1448,166 @@ static int i40iw_handle_q_mem(struct i40iw_device *iwdev,
 	return err;
 }
 
+/**
+ * i40iw_hw_alloc_stag - cqp command to allocate stag
+ * @iwdev: iwarp device
+ * @iwmr: iwarp mr pointer
+ */
+static int i40iw_hw_alloc_stag(struct i40iw_device *iwdev, struct i40iw_mr *iwmr)
+{
+	struct i40iw_allocate_stag_info *info;
+	struct i40iw_pd *iwpd = to_iwpd(iwmr->ibmr.pd);
+	enum i40iw_status_code status;
+	int err = 0;
+	struct i40iw_cqp_request *cqp_request;
+	struct cqp_commands_info *cqp_info;
+
+	cqp_request = i40iw_get_cqp_request(&iwdev->cqp, true);
+	if (!cqp_request)
+		return -ENOMEM;
+
+	cqp_info = &cqp_request->info;
+	info = &cqp_info->in.u.alloc_stag.info;
+	memset(info, 0, sizeof(*info));
+	info->page_size = PAGE_SIZE;
+	info->stag_idx = iwmr->stag >> I40IW_CQPSQ_STAG_IDX_SHIFT;
+	info->pd_id = iwpd->sc_pd.pd_id;
+	info->total_len = iwmr->length;
+	cqp_info->cqp_cmd = OP_ALLOC_STAG;
+	cqp_info->post_sq = 1;
+	cqp_info->in.u.alloc_stag.dev = &iwdev->sc_dev;
+	cqp_info->in.u.alloc_stag.scratch = (uintptr_t)cqp_request;
+
+	status = i40iw_handle_cqp_op(iwdev, cqp_request);
+	if (status) {
+		err = -ENOMEM;
+		i40iw_pr_err("CQP-OP MR Reg fail");
+	}
+	return err;
+}
+
+/**
+ * i40iw_alloc_mr - register stag for fast memory registration
+ * @pd: ibpd pointer
+ * @mr_type: memory for stag registrion
+ * @max_num_sg: man number of pages
+ */
+static struct ib_mr *i40iw_alloc_mr(struct ib_pd *pd,
+				    enum ib_mr_type mr_type,
+				    u32 max_num_sg)
+{
+	struct i40iw_pd *iwpd = to_iwpd(pd);
+	struct i40iw_device *iwdev = to_iwdev(pd->device);
+	struct i40iw_pble_alloc *palloc;
+	struct i40iw_pbl *iwpbl;
+	struct i40iw_mr *iwmr;
+	enum i40iw_status_code status;
+	u32 stag;
+	int err_code = -ENOMEM;
+
+	iwmr = kzalloc(sizeof(*iwmr), GFP_KERNEL);
+	if (!iwmr)
+		return ERR_PTR(-ENOMEM);
+
+	stag = i40iw_create_stag(iwdev);
+	if (!stag) {
+		err_code = -EOVERFLOW;
+		goto err;
+	}
+	iwmr->stag = stag;
+	iwmr->ibmr.rkey = stag;
+	iwmr->ibmr.lkey = stag;
+	iwmr->ibmr.pd = pd;
+	iwmr->ibmr.device = pd->device;
+	iwpbl = &iwmr->iwpbl;
+	iwpbl->iwmr = iwmr;
+	iwmr->type = IW_MEMREG_TYPE_MEM;
+	palloc = &iwpbl->pble_alloc;
+	iwmr->page_cnt = max_num_sg;
+	mutex_lock(&iwdev->pbl_mutex);
+	status = i40iw_get_pble(&iwdev->sc_dev, iwdev->pble_rsrc, palloc, iwmr->page_cnt);
+	mutex_unlock(&iwdev->pbl_mutex);
+	if (!status)
+		goto err1;
+
+	if (palloc->level != I40IW_LEVEL_1)
+		goto err2;
+	err_code = i40iw_hw_alloc_stag(iwdev, iwmr);
+	if (err_code)
+		goto err2;
+	iwpbl->pbl_allocated = true;
+	i40iw_add_pdusecount(iwpd);
+	return &iwmr->ibmr;
+err2:
+	i40iw_free_pble(iwdev->pble_rsrc, palloc);
+err1:
+	i40iw_free_stag(iwdev, stag);
+err:
+	kfree(iwmr);
+	return ERR_PTR(err_code);
+}
+
+/**
+ * i40iw_set_page - populate pbl list for fmr
+ * @ibmr: ib mem to access iwarp mr pointer
+ * @addr: page dma address fro pbl list
+ */
+static int i40iw_set_page(struct ib_mr *ibmr, u64 addr)
+{
+	struct i40iw_mr *iwmr = to_iwmr(ibmr);
+	struct i40iw_pbl *iwpbl = &iwmr->iwpbl;
+	struct i40iw_pble_alloc *palloc = &iwpbl->pble_alloc;
+	u64 *pbl;
+
+	if (unlikely(iwmr->npages == iwmr->page_cnt))
+		return -ENOMEM;
+
+	pbl = (u64 *)palloc->level1.addr;
+	pbl[iwmr->npages++] = cpu_to_le64(addr);
+	return 0;
+}
+
+/**
+ * i40iw_map_mr_sg - map of sg list for fmr
+ * @ibmr: ib mem to access iwarp mr pointer
+ * @sg: scatter gather list for fmr
+ * @sg_nents: number of sg pages
+ */
+static int i40iw_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg,
+			   int sg_nents, unsigned int *sg_offset)
+{
+	struct i40iw_mr *iwmr = to_iwmr(ibmr);
+
+	iwmr->npages = 0;
+	return ib_sg_to_pages(ibmr, sg, sg_nents, sg_offset, i40iw_set_page);
+}
+
+/**
+ * i40iw_drain_sq - drain the send queue
+ * @ibqp: ib qp pointer
+ */
+static void i40iw_drain_sq(struct ib_qp *ibqp)
+{
+	struct i40iw_qp *iwqp = to_iwqp(ibqp);
+	struct i40iw_sc_qp *qp = &iwqp->sc_qp;
+
+	if (I40IW_RING_MORE_WORK(qp->qp_uk.sq_ring))
+		wait_for_completion(&iwqp->sq_drained);
+}
+
+/**
+ * i40iw_drain_rq - drain the receive queue
+ * @ibqp: ib qp pointer
+ */
+static void i40iw_drain_rq(struct ib_qp *ibqp)
+{
+	struct i40iw_qp *iwqp = to_iwqp(ibqp);
+	struct i40iw_sc_qp *qp = &iwqp->sc_qp;
+
+	if (I40IW_RING_MORE_WORK(qp->qp_uk.rq_ring))
+		wait_for_completion(&iwqp->rq_drained);
+}
+
 /**
  * i40iw_hwreg_mr - send cqp command for memory registration
  * @iwdev: iwarp device
@@ -1526,14 +1691,16 @@ static struct ib_mr *i40iw_reg_user_mr(struct ib_pd *pd,
 	struct i40iw_mr *iwmr;
 	struct ib_umem *region;
 	struct i40iw_mem_reg_req req;
-	u32 pbl_depth = 0;
+	u64 pbl_depth = 0;
 	u32 stag = 0;
 	u16 access;
-	u32 region_length;
+	u64 region_length;
 	bool use_pbles = false;
 	unsigned long flags;
 	int err = -ENOSYS;
 
+	if (length > I40IW_MAX_MR_SIZE)
+		return ERR_PTR(-EINVAL);
 	region = ib_umem_get(pd->uobject->context, start, length, acc, 0);
 	if (IS_ERR(region))
 		return (struct ib_mr *)region;
@@ -1564,7 +1731,7 @@ static struct ib_mr *i40iw_reg_user_mr(struct ib_pd *pd,
 	palloc = &iwpbl->pble_alloc;
 
 	iwmr->type = req.reg_type;
-	iwmr->page_cnt = pbl_depth;
+	iwmr->page_cnt = (u32)pbl_depth;
 
 	switch (req.reg_type) {
 	case IW_MEMREG_TYPE_QP:
@@ -1881,12 +2048,14 @@ static int i40iw_post_send(struct ib_qp *ibqp,
 	enum i40iw_status_code ret;
 	int err = 0;
 	unsigned long flags;
+	bool inv_stag;
 
 	iwqp = (struct i40iw_qp *)ibqp;
 	ukqp = &iwqp->sc_qp.qp_uk;
 
 	spin_lock_irqsave(&iwqp->lock, flags);
 	while (ib_wr) {
+		inv_stag = false;
 		memset(&info, 0, sizeof(info));
 		info.wr_id = (u64)(ib_wr->wr_id);
 		if ((ib_wr->send_flags & IB_SEND_SIGNALED) || iwqp->sig_all)
@@ -1896,19 +2065,28 @@ static int i40iw_post_send(struct ib_qp *ibqp,
 
 		switch (ib_wr->opcode) {
 		case IB_WR_SEND:
-			if (ib_wr->send_flags & IB_SEND_SOLICITED)
-				info.op_type = I40IW_OP_TYPE_SEND_SOL;
-			else
-				info.op_type = I40IW_OP_TYPE_SEND;
+			/* fall-through */
+		case IB_WR_SEND_WITH_INV:
+			if (ib_wr->opcode == IB_WR_SEND) {
+				if (ib_wr->send_flags & IB_SEND_SOLICITED)
+					info.op_type = I40IW_OP_TYPE_SEND_SOL;
+				else
+					info.op_type = I40IW_OP_TYPE_SEND;
+			} else {
+				if (ib_wr->send_flags & IB_SEND_SOLICITED)
+					info.op_type = I40IW_OP_TYPE_SEND_SOL_INV;
+				else
+					info.op_type = I40IW_OP_TYPE_SEND_INV;
+			}
 
 			if (ib_wr->send_flags & IB_SEND_INLINE) {
 				info.op.inline_send.data = (void *)(unsigned long)ib_wr->sg_list[0].addr;
 				info.op.inline_send.len = ib_wr->sg_list[0].length;
-				ret = ukqp->ops.iw_inline_send(ukqp, &info, rdma_wr(ib_wr)->rkey, false);
+				ret = ukqp->ops.iw_inline_send(ukqp, &info, ib_wr->ex.invalidate_rkey, false);
 			} else {
 				info.op.send.num_sges = ib_wr->num_sge;
 				info.op.send.sg_list = (struct i40iw_sge *)ib_wr->sg_list;
-				ret = ukqp->ops.iw_send(ukqp, &info, rdma_wr(ib_wr)->rkey, false);
+				ret = ukqp->ops.iw_send(ukqp, &info, ib_wr->ex.invalidate_rkey, false);
 			}
 
 			if (ret)
@@ -1936,7 +2114,14 @@ static int i40iw_post_send(struct ib_qp *ibqp,
 			if (ret)
 				err = -EIO;
 			break;
+		case IB_WR_RDMA_READ_WITH_INV:
+			inv_stag = true;
+			/* fall-through*/
 		case IB_WR_RDMA_READ:
+			if (ib_wr->num_sge > I40IW_MAX_SGE_RD) {
+				err = -EINVAL;
+				break;
+			}
 			info.op_type = I40IW_OP_TYPE_RDMA_READ;
 			info.op.rdma_read.rem_addr.tag_off = rdma_wr(ib_wr)->remote_addr;
 			info.op.rdma_read.rem_addr.stag = rdma_wr(ib_wr)->rkey;
@@ -1944,10 +2129,47 @@ static int i40iw_post_send(struct ib_qp *ibqp,
 			info.op.rdma_read.lo_addr.tag_off = ib_wr->sg_list->addr;
 			info.op.rdma_read.lo_addr.stag = ib_wr->sg_list->lkey;
 			info.op.rdma_read.lo_addr.len = ib_wr->sg_list->length;
-			ret = ukqp->ops.iw_rdma_read(ukqp, &info, false, false);
+			ret = ukqp->ops.iw_rdma_read(ukqp, &info, inv_stag, false);
 			if (ret)
 				err = -EIO;
 			break;
+		case IB_WR_LOCAL_INV:
+			info.op_type = I40IW_OP_TYPE_INV_STAG;
+			info.op.inv_local_stag.target_stag = ib_wr->ex.invalidate_rkey;
+			ret = ukqp->ops.iw_stag_local_invalidate(ukqp, &info, true);
+			if (ret)
+				err = -EIO;
+			break;
+		case IB_WR_REG_MR:
+		{
+			struct i40iw_mr *iwmr = to_iwmr(reg_wr(ib_wr)->mr);
+			int page_shift = ilog2(reg_wr(ib_wr)->mr->page_size);
+			int flags = reg_wr(ib_wr)->access;
+			struct i40iw_pble_alloc *palloc = &iwmr->iwpbl.pble_alloc;
+			struct i40iw_sc_dev *dev = &iwqp->iwdev->sc_dev;
+			struct i40iw_fast_reg_stag_info info;
+
+			info.access_rights = I40IW_ACCESS_FLAGS_LOCALREAD;
+			info.access_rights |= i40iw_get_user_access(flags);
+			info.stag_key = reg_wr(ib_wr)->key & 0xff;
+			info.stag_idx = reg_wr(ib_wr)->key >> 8;
+			info.wr_id = ib_wr->wr_id;
+
+			info.addr_type = I40IW_ADDR_TYPE_VA_BASED;
+			info.va = (void *)(uintptr_t)iwmr->ibmr.iova;
+			info.total_len = iwmr->ibmr.length;
+			info.first_pm_pbl_index = palloc->level1.idx;
+			info.local_fence = ib_wr->send_flags & IB_SEND_FENCE;
+			info.signaled = ib_wr->send_flags & IB_SEND_SIGNALED;
+
+			if (page_shift == 21)
+				info.page_size = 1; /* 2M page */
+
+			ret = dev->iw_priv_qp_ops->iw_mr_fast_register(&iwqp->sc_qp, &info, true);
+			if (ret)
+				err = -EIO;
+			break;
+		}
 		default:
 			err = -EINVAL;
 			i40iw_pr_err(" upost_send bad opcode = 0x%x\n",
@@ -2027,6 +2249,7 @@ static int i40iw_poll_cq(struct ib_cq *ibcq,
 	enum i40iw_status_code ret;
 	struct i40iw_cq_uk *ukcq;
 	struct i40iw_sc_qp *qp;
+	struct i40iw_qp *iwqp;
 	unsigned long flags;
 
 	iwcq = (struct i40iw_cq *)ibcq;
@@ -2037,6 +2260,8 @@ static int i40iw_poll_cq(struct ib_cq *ibcq,
 		ret = ukcq->ops.iw_cq_poll_completion(ukcq, &cq_poll_info, true);
 		if (ret == I40IW_ERR_QUEUE_EMPTY) {
 			break;
+		} else if (ret == I40IW_ERR_QUEUE_DESTROYED) {
+			continue;
 		} else if (ret) {
 			if (!cqe_count)
 				cqe_count = -1;
@@ -2044,10 +2269,12 @@ static int i40iw_poll_cq(struct ib_cq *ibcq,
 		}
 		entry->wc_flags = 0;
 		entry->wr_id = cq_poll_info.wr_id;
-		if (!cq_poll_info.error)
-			entry->status = IB_WC_SUCCESS;
-		else
+		if (cq_poll_info.error) {
 			entry->status = IB_WC_WR_FLUSH_ERR;
+			entry->vendor_err = cq_poll_info.major_err << 16 | cq_poll_info.minor_err;
+		} else {
+			entry->status = IB_WC_SUCCESS;
+		}
 
 		switch (cq_poll_info.op_type) {
 		case I40IW_OP_TYPE_RDMA_WRITE:
@@ -2071,12 +2298,17 @@ static int i40iw_poll_cq(struct ib_cq *ibcq,
 			break;
 		}
 
-		entry->vendor_err =
-		    cq_poll_info.major_err << 16 | cq_poll_info.minor_err;
 		entry->ex.imm_data = 0;
 		qp = (struct i40iw_sc_qp *)cq_poll_info.qp_handle;
 		entry->qp = (struct ib_qp *)qp->back_qp;
 		entry->src_qp = cq_poll_info.qp_id;
+		iwqp = (struct i40iw_qp *)qp->back_qp;
+		if (iwqp->iwarp_state > I40IW_QP_STATE_RTS) {
+			if (!I40IW_RING_MORE_WORK(qp->qp_uk.sq_ring))
+				complete(&iwqp->sq_drained);
+			if (!I40IW_RING_MORE_WORK(qp->qp_uk.rq_ring))
+				complete(&iwqp->rq_drained);
+		}
 		entry->byte_len = cq_poll_info.bytes_xfered;
 		entry++;
 		cqe_count++;
@@ -2143,7 +2375,6 @@ static int i40iw_get_protocol_stats(struct ib_device *ibdev,
 	struct i40iw_dev_hw_stats *hw_stats = &devstat->hw_stats;
 	struct timespec curr_time;
 	static struct timespec last_rd_time = {0, 0};
-	enum i40iw_status_code status = 0;
 	unsigned long flags;
 
 	curr_time = current_kernel_time();
@@ -2156,11 +2387,8 @@ static int i40iw_get_protocol_stats(struct ib_device *ibdev,
 		spin_unlock_irqrestore(&devstat->stats_lock, flags);
 	} else {
 		if (((u64)curr_time.tv_sec - (u64)last_rd_time.tv_sec) > 1)
-			status = i40iw_vchnl_vf_get_pe_stats(dev,
-							     &devstat->hw_stats);
-
-		if (status)
-			return -ENOSYS;
+			if (i40iw_vchnl_vf_get_pe_stats(dev, &devstat->hw_stats))
+				return -ENOSYS;
 	}
 
 	stats->iw.ipInReceives = hw_stats->stat_value_64[I40IW_HW_STAT_INDEX_IP4RXPKTS] +
@@ -2327,6 +2555,10 @@ static struct i40iw_ib_device *i40iw_init_rdma_device(struct i40iw_device *iwdev
 	iwibdev->ibdev.query_device = i40iw_query_device;
 	iwibdev->ibdev.create_ah = i40iw_create_ah;
 	iwibdev->ibdev.destroy_ah = i40iw_destroy_ah;
+	iwibdev->ibdev.drain_sq = i40iw_drain_sq;
+	iwibdev->ibdev.drain_rq = i40iw_drain_rq;
+	iwibdev->ibdev.alloc_mr = i40iw_alloc_mr;
+	iwibdev->ibdev.map_mr_sg = i40iw_map_mr_sg;
 	iwibdev->ibdev.iwcm = kzalloc(sizeof(*iwibdev->ibdev.iwcm), GFP_KERNEL);
 	if (!iwibdev->ibdev.iwcm) {
 		ib_dealloc_device(&iwibdev->ibdev);

drivers/infiniband/hw/i40iw/i40iw_verbs.h

@@ -92,6 +92,7 @@ struct i40iw_mr {
 	struct ib_umem *region;
 	u16 type;
 	u32 page_cnt;
+	u32 npages;
 	u32 stag;
 	u64 length;
 	u64 pgaddrmem[MAX_SAVE_PAGE_ADDRS];
@@ -169,5 +170,7 @@ struct i40iw_qp {
 	struct i40iw_pbl *iwpbl;
 	struct i40iw_dma_mem q2_ctx_mem;
 	struct i40iw_dma_mem ietf_mem;
+	struct completion sq_drained;
+	struct completion rq_drained;
 };
 #endif

drivers/infiniband/hw/i40iw/i40iw_vf.c

@@ -80,6 +80,6 @@ enum i40iw_status_code i40iw_manage_vf_pble_bp(struct i40iw_sc_cqp *cqp,
 	return 0;
 }
 
-struct i40iw_vf_cqp_ops iw_vf_cqp_ops = {
+const struct i40iw_vf_cqp_ops iw_vf_cqp_ops = {
 	i40iw_manage_vf_pble_bp
 };

drivers/infiniband/hw/i40iw/i40iw_vf.h

@@ -57,6 +57,6 @@ enum i40iw_status_code i40iw_manage_vf_pble_bp(struct i40iw_sc_cqp *cqp,
 					       u64 scratch,
 					       bool post_sq);
 
-extern struct i40iw_vf_cqp_ops iw_vf_cqp_ops;
+extern const struct i40iw_vf_cqp_ops iw_vf_cqp_ops;
 
 #endif

drivers/infiniband/hw/i40iw/i40iw_virtchnl.c

@@ -254,7 +254,7 @@ static void vchnl_pf_send_get_hmc_fcn_resp(struct i40iw_sc_dev *dev,
 static void vchnl_pf_send_get_pe_stats_resp(struct i40iw_sc_dev *dev,
 					    u32 vf_id,
 					    struct i40iw_virtchnl_op_buf *vchnl_msg,
-					    struct i40iw_dev_hw_stats hw_stats)
+					    struct i40iw_dev_hw_stats *hw_stats)
 {
 	enum i40iw_status_code ret_code;
 	u8 resp_buffer[sizeof(struct i40iw_virtchnl_resp_buf) + sizeof(struct i40iw_dev_hw_stats) - 1];
@@ -264,7 +264,7 @@ static void vchnl_pf_send_get_pe_stats_resp(struct i40iw_sc_dev *dev,
 	vchnl_msg_resp->iw_chnl_op_ctx = vchnl_msg->iw_chnl_op_ctx;
 	vchnl_msg_resp->iw_chnl_buf_len = sizeof(resp_buffer);
 	vchnl_msg_resp->iw_op_ret_code = I40IW_SUCCESS;
-	*((struct i40iw_dev_hw_stats *)vchnl_msg_resp->iw_chnl_buf) = hw_stats;
+	*((struct i40iw_dev_hw_stats *)vchnl_msg_resp->iw_chnl_buf) = *hw_stats;
 	ret_code = dev->vchnl_if.vchnl_send(dev, vf_id, resp_buffer, sizeof(resp_buffer));
 	if (ret_code)
 		i40iw_debug(dev, I40IW_DEBUG_VIRT,
@@ -437,11 +437,9 @@ enum i40iw_status_code i40iw_vchnl_recv_pf(struct i40iw_sc_dev *dev,
 			vchnl_pf_send_get_ver_resp(dev, vf_id, vchnl_msg);
 		return I40IW_SUCCESS;
 	}
-	for (iw_vf_idx = 0; iw_vf_idx < I40IW_MAX_PE_ENABLED_VF_COUNT;
-	     iw_vf_idx++) {
+	for (iw_vf_idx = 0; iw_vf_idx < I40IW_MAX_PE_ENABLED_VF_COUNT; iw_vf_idx++) {
 		if (!dev->vf_dev[iw_vf_idx]) {
-			if (first_avail_iw_vf ==
-			    I40IW_MAX_PE_ENABLED_VF_COUNT)
+			if (first_avail_iw_vf == I40IW_MAX_PE_ENABLED_VF_COUNT)
 				first_avail_iw_vf = iw_vf_idx;
 			continue;
 		}
@@ -541,7 +539,7 @@ enum i40iw_status_code i40iw_vchnl_recv_pf(struct i40iw_sc_dev *dev,
 		devstat->ops.iw_hw_stat_read_all(devstat, &devstat->hw_stats);
 		spin_unlock_irqrestore(&dev->dev_pestat.stats_lock, flags);
 		vf_dev->msg_count--;
-		vchnl_pf_send_get_pe_stats_resp(dev, vf_id, vchnl_msg, devstat->hw_stats);
+		vchnl_pf_send_get_pe_stats_resp(dev, vf_id, vchnl_msg, &devstat->hw_stats);
 		break;
 	default:
 		i40iw_debug(dev, I40IW_DEBUG_VIRT,
@@ -596,23 +594,25 @@ enum i40iw_status_code i40iw_vchnl_vf_get_ver(struct i40iw_sc_dev *dev,
 	struct i40iw_virtchnl_req vchnl_req;
 	enum i40iw_status_code ret_code;
 
+	if (!i40iw_vf_clear_to_send(dev))
+		return I40IW_ERR_TIMEOUT;
 	memset(&vchnl_req, 0, sizeof(vchnl_req));
 	vchnl_req.dev = dev;
 	vchnl_req.parm = vchnl_ver;
 	vchnl_req.parm_len = sizeof(*vchnl_ver);
 	vchnl_req.vchnl_msg = &dev->vchnl_vf_msg_buf.vchnl_msg;
+
 	ret_code = vchnl_vf_send_get_ver_req(dev, &vchnl_req);
-	if (!ret_code) {
-		ret_code = i40iw_vf_wait_vchnl_resp(dev);
-		if (!ret_code)
-			ret_code = vchnl_req.ret_code;
-		else
-			dev->vchnl_up = false;
-	} else {
+	if (ret_code) {
 		i40iw_debug(dev, I40IW_DEBUG_VIRT,
 			    "%s Send message failed 0x%0x\n", __func__, ret_code);
+		return ret_code;
 	}
-	return ret_code;
+	ret_code = i40iw_vf_wait_vchnl_resp(dev);
+	if (ret_code)
+		return ret_code;
+	else
+		return vchnl_req.ret_code;
 }
 
 /**
@@ -626,23 +626,25 @@ enum i40iw_status_code i40iw_vchnl_vf_get_hmc_fcn(struct i40iw_sc_dev *dev,
 	struct i40iw_virtchnl_req vchnl_req;
 	enum i40iw_status_code ret_code;
 
+	if (!i40iw_vf_clear_to_send(dev))
+		return I40IW_ERR_TIMEOUT;
 	memset(&vchnl_req, 0, sizeof(vchnl_req));
 	vchnl_req.dev = dev;
 	vchnl_req.parm = hmc_fcn;
 	vchnl_req.parm_len = sizeof(*hmc_fcn);
 	vchnl_req.vchnl_msg = &dev->vchnl_vf_msg_buf.vchnl_msg;
+
 	ret_code = vchnl_vf_send_get_hmc_fcn_req(dev, &vchnl_req);
-	if (!ret_code) {
-		ret_code = i40iw_vf_wait_vchnl_resp(dev);
-		if (!ret_code)
-			ret_code = vchnl_req.ret_code;
-		else
-			dev->vchnl_up = false;
-	} else {
+	if (ret_code) {
 		i40iw_debug(dev, I40IW_DEBUG_VIRT,
 			    "%s Send message failed 0x%0x\n", __func__, ret_code);
+		return ret_code;
 	}
-	return ret_code;
+	ret_code = i40iw_vf_wait_vchnl_resp(dev);
+	if (ret_code)
+		return ret_code;
+	else
+		return vchnl_req.ret_code;
 }
 
 /**
@@ -660,25 +662,27 @@ enum i40iw_status_code i40iw_vchnl_vf_add_hmc_objs(struct i40iw_sc_dev *dev,
 	struct i40iw_virtchnl_req vchnl_req;
 	enum i40iw_status_code ret_code;
 
+	if (!i40iw_vf_clear_to_send(dev))
+		return I40IW_ERR_TIMEOUT;
 	memset(&vchnl_req, 0, sizeof(vchnl_req));
 	vchnl_req.dev = dev;
 	vchnl_req.vchnl_msg = &dev->vchnl_vf_msg_buf.vchnl_msg;
+
 	ret_code = vchnl_vf_send_add_hmc_objs_req(dev,
 						  &vchnl_req,
 						  rsrc_type,
 						  start_index,
 						  rsrc_count);
-	if (!ret_code) {
-		ret_code = i40iw_vf_wait_vchnl_resp(dev);
-		if (!ret_code)
-			ret_code = vchnl_req.ret_code;
-		else
-			dev->vchnl_up = false;
-	} else {
+	if (ret_code) {
 		i40iw_debug(dev, I40IW_DEBUG_VIRT,
 			    "%s Send message failed 0x%0x\n", __func__, ret_code);
+		return ret_code;
 	}
-	return ret_code;
+	ret_code = i40iw_vf_wait_vchnl_resp(dev);
+	if (ret_code)
+		return ret_code;
+	else
+		return vchnl_req.ret_code;
 }
 
 /**
@@ -696,25 +700,27 @@ enum i40iw_status_code i40iw_vchnl_vf_del_hmc_obj(struct i40iw_sc_dev *dev,
 	struct i40iw_virtchnl_req vchnl_req;
 	enum i40iw_status_code ret_code;
 
+	if (!i40iw_vf_clear_to_send(dev))
+		return I40IW_ERR_TIMEOUT;
 	memset(&vchnl_req, 0, sizeof(vchnl_req));
 	vchnl_req.dev = dev;
 	vchnl_req.vchnl_msg = &dev->vchnl_vf_msg_buf.vchnl_msg;
+
 	ret_code = vchnl_vf_send_del_hmc_objs_req(dev,
 						  &vchnl_req,
 						  rsrc_type,
 						  start_index,
 						  rsrc_count);
-	if (!ret_code) {
-		ret_code = i40iw_vf_wait_vchnl_resp(dev);
-		if (!ret_code)
-			ret_code = vchnl_req.ret_code;
-		else
-			dev->vchnl_up = false;
-	} else {
+	if (ret_code) {
 		i40iw_debug(dev, I40IW_DEBUG_VIRT,
 			    "%s Send message failed 0x%0x\n", __func__, ret_code);
+		return ret_code;
 	}
-	return ret_code;
+	ret_code = i40iw_vf_wait_vchnl_resp(dev);
+	if (ret_code)
+		return ret_code;
+	else
+		return vchnl_req.ret_code;
 }
 
 /**
@@ -728,21 +734,23 @@ enum i40iw_status_code i40iw_vchnl_vf_get_pe_stats(struct i40iw_sc_dev *dev,
 	struct i40iw_virtchnl_req  vchnl_req;
 	enum i40iw_status_code ret_code;
 
+	if (!i40iw_vf_clear_to_send(dev))
+		return I40IW_ERR_TIMEOUT;
 	memset(&vchnl_req, 0, sizeof(vchnl_req));
 	vchnl_req.dev = dev;
 	vchnl_req.parm = hw_stats;
 	vchnl_req.parm_len = sizeof(*hw_stats);
 	vchnl_req.vchnl_msg = &dev->vchnl_vf_msg_buf.vchnl_msg;
+
 	ret_code = vchnl_vf_send_get_pe_stats_req(dev, &vchnl_req);
-	if (!ret_code) {
-		ret_code = i40iw_vf_wait_vchnl_resp(dev);
-		if (!ret_code)
-			ret_code = vchnl_req.ret_code;
-		else
-			dev->vchnl_up = false;
-	} else {
+	if (ret_code) {
 		i40iw_debug(dev, I40IW_DEBUG_VIRT,
 			    "%s Send message failed 0x%0x\n", __func__, ret_code);
+		return ret_code;
 	}
-	return ret_code;
+	ret_code = i40iw_vf_wait_vchnl_resp(dev);
+	if (ret_code)
+		return ret_code;
+	else
+		return vchnl_req.ret_code;
 }

drivers/infiniband/hw/mlx4/mcg.c

@@ -96,7 +96,7 @@ struct ib_sa_mcmember_data {
 	u8		scope_join_state;
 	u8		proxy_join;
 	u8		reserved[2];
-};
+} __packed __aligned(4);
 
 struct mcast_group {
 	struct ib_sa_mcmember_data rec;
@@ -747,14 +747,11 @@ static struct mcast_group *search_relocate_mgid0_group(struct mlx4_ib_demux_ctx
 						       __be64 tid,
 						       union ib_gid *new_mgid)
 {
-	struct mcast_group *group = NULL, *cur_group;
+	struct mcast_group *group = NULL, *cur_group, *n;
 	struct mcast_req *req;
-	struct list_head *pos;
-	struct list_head *n;
 
 	mutex_lock(&ctx->mcg_table_lock);
-	list_for_each_safe(pos, n, &ctx->mcg_mgid0_list) {
-		group = list_entry(pos, struct mcast_group, mgid0_list);
+	list_for_each_entry_safe(group, n, &ctx->mcg_mgid0_list, mgid0_list) {
 		mutex_lock(&group->lock);
 		if (group->last_req_tid == tid) {
 			if (memcmp(new_mgid, &mgid0, sizeof mgid0)) {

drivers/infiniband/hw/mlx4/mlx4_ib.h

@@ -717,9 +717,8 @@ int mlx4_ib_dealloc_mw(struct ib_mw *mw);
 struct ib_mr *mlx4_ib_alloc_mr(struct ib_pd *pd,
 			       enum ib_mr_type mr_type,
 			       u32 max_num_sg);
-int mlx4_ib_map_mr_sg(struct ib_mr *ibmr,
-		      struct scatterlist *sg,
-		      int sg_nents);
+int mlx4_ib_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg, int sg_nents,
+		      unsigned int *sg_offset);
 int mlx4_ib_modify_cq(struct ib_cq *cq, u16 cq_count, u16 cq_period);
 int mlx4_ib_resize_cq(struct ib_cq *ibcq, int entries, struct ib_udata *udata);
 struct ib_cq *mlx4_ib_create_cq(struct ib_device *ibdev,

drivers/infiniband/hw/mlx4/mr.c

@@ -528,9 +528,8 @@ static int mlx4_set_page(struct ib_mr *ibmr, u64 addr)
 	return 0;
 }
 
-int mlx4_ib_map_mr_sg(struct ib_mr *ibmr,
-		      struct scatterlist *sg,
-		      int sg_nents)
+int mlx4_ib_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg, int sg_nents,
+		      unsigned int *sg_offset)
 {
 	struct mlx4_ib_mr *mr = to_mmr(ibmr);
 	int rc;
@@ -541,7 +540,7 @@ int mlx4_ib_map_mr_sg(struct ib_mr *ibmr,
 				   sizeof(u64) * mr->max_pages,
 				   DMA_TO_DEVICE);
 
-	rc = ib_sg_to_pages(ibmr, sg, sg_nents, mlx4_set_page);
+	rc = ib_sg_to_pages(ibmr, sg, sg_nents, sg_offset, mlx4_set_page);
 
 	ib_dma_sync_single_for_device(ibmr->device, mr->page_map,
 				      sizeof(u64) * mr->max_pages,

drivers/infiniband/hw/mlx5/cq.c

@@ -879,7 +879,10 @@ struct ib_cq *mlx5_ib_create_cq(struct ib_device *ibdev,
 
 	mlx5_ib_dbg(dev, "cqn 0x%x\n", cq->mcq.cqn);
 	cq->mcq.irqn = irqn;
-	cq->mcq.comp  = mlx5_ib_cq_comp;
+	if (context)
+		cq->mcq.tasklet_ctx.comp = mlx5_ib_cq_comp;
+	else
+		cq->mcq.comp  = mlx5_ib_cq_comp;
 	cq->mcq.event = mlx5_ib_cq_event;
 
 	INIT_LIST_HEAD(&cq->wc_list);

drivers/infiniband/hw/mlx5/main.c

@@ -38,6 +38,9 @@
 #include <linux/dma-mapping.h>
 #include <linux/slab.h>
 #include <linux/io-mapping.h>
+#if defined(CONFIG_X86)
+#include <asm/pat.h>
+#endif
 #include <linux/sched.h>
 #include <rdma/ib_user_verbs.h>
 #include <rdma/ib_addr.h>
@@ -517,6 +520,10 @@ static int mlx5_ib_query_device(struct ib_device *ibdev,
 		props->device_cap_flags |= IB_DEVICE_UD_TSO;
 	}
 
+	if (MLX5_CAP_GEN(dev->mdev, eth_net_offloads) &&
+	    MLX5_CAP_ETH(dev->mdev, scatter_fcs))
+		props->device_cap_flags |= IB_DEVICE_RAW_SCATTER_FCS;
+
 	props->vendor_part_id	   = mdev->pdev->device;
 	props->hw_ver		   = mdev->pdev->revision;
 
@@ -1068,38 +1075,89 @@ static int get_index(unsigned long offset)
 	return get_arg(offset);
 }
 
+static inline char *mmap_cmd2str(enum mlx5_ib_mmap_cmd cmd)
+{
+	switch (cmd) {
+	case MLX5_IB_MMAP_WC_PAGE:
+		return "WC";
+	case MLX5_IB_MMAP_REGULAR_PAGE:
+		return "best effort WC";
+	case MLX5_IB_MMAP_NC_PAGE:
+		return "NC";
+	default:
+		return NULL;
+	}
+}
+
+static int uar_mmap(struct mlx5_ib_dev *dev, enum mlx5_ib_mmap_cmd cmd,
+		    struct vm_area_struct *vma, struct mlx5_uuar_info *uuari)
+{
+	int err;
+	unsigned long idx;
+	phys_addr_t pfn, pa;
+	pgprot_t prot;
+
+	switch (cmd) {
+	case MLX5_IB_MMAP_WC_PAGE:
+/* Some architectures don't support WC memory */
+#if defined(CONFIG_X86)
+		if (!pat_enabled())
+			return -EPERM;
+#elif !(defined(CONFIG_PPC) || (defined(CONFIG_ARM) && defined(CONFIG_MMU)))
+			return -EPERM;
+#endif
+	/* fall through */
+	case MLX5_IB_MMAP_REGULAR_PAGE:
+		/* For MLX5_IB_MMAP_REGULAR_PAGE do the best effort to get WC */
+		prot = pgprot_writecombine(vma->vm_page_prot);
+		break;
+	case MLX5_IB_MMAP_NC_PAGE:
+		prot = pgprot_noncached(vma->vm_page_prot);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	if (vma->vm_end - vma->vm_start != PAGE_SIZE)
+		return -EINVAL;
+
+	idx = get_index(vma->vm_pgoff);
+	if (idx >= uuari->num_uars)
+		return -EINVAL;
+
+	pfn = uar_index2pfn(dev, uuari->uars[idx].index);
+	mlx5_ib_dbg(dev, "uar idx 0x%lx, pfn %pa\n", idx, &pfn);
+
+	vma->vm_page_prot = prot;
+	err = io_remap_pfn_range(vma, vma->vm_start, pfn,
+				 PAGE_SIZE, vma->vm_page_prot);
+	if (err) {
+		mlx5_ib_err(dev, "io_remap_pfn_range failed with error=%d, vm_start=0x%lx, pfn=%pa, mmap_cmd=%s\n",
+			    err, vma->vm_start, &pfn, mmap_cmd2str(cmd));
+		return -EAGAIN;
+	}
+
+	pa = pfn << PAGE_SHIFT;
+	mlx5_ib_dbg(dev, "mapped %s at 0x%lx, PA %pa\n", mmap_cmd2str(cmd),
+		    vma->vm_start, &pa);
+
+	return 0;
+}
+
 static int mlx5_ib_mmap(struct ib_ucontext *ibcontext, struct vm_area_struct *vma)
 {
 	struct mlx5_ib_ucontext *context = to_mucontext(ibcontext);
 	struct mlx5_ib_dev *dev = to_mdev(ibcontext->device);
 	struct mlx5_uuar_info *uuari = &context->uuari;
 	unsigned long command;
-	unsigned long idx;
 	phys_addr_t pfn;
 
 	command = get_command(vma->vm_pgoff);
 	switch (command) {
+	case MLX5_IB_MMAP_WC_PAGE:
+	case MLX5_IB_MMAP_NC_PAGE:
 	case MLX5_IB_MMAP_REGULAR_PAGE:
-		if (vma->vm_end - vma->vm_start != PAGE_SIZE)
-			return -EINVAL;
-
-		idx = get_index(vma->vm_pgoff);
-		if (idx >= uuari->num_uars)
-			return -EINVAL;
-
-		pfn = uar_index2pfn(dev, uuari->uars[idx].index);
-		mlx5_ib_dbg(dev, "uar idx 0x%lx, pfn 0x%llx\n", idx,
-			    (unsigned long long)pfn);
-
-		vma->vm_page_prot = pgprot_writecombine(vma->vm_page_prot);
-		if (io_remap_pfn_range(vma, vma->vm_start, pfn,
-				       PAGE_SIZE, vma->vm_page_prot))
-			return -EAGAIN;
-
-		mlx5_ib_dbg(dev, "mapped WC at 0x%lx, PA 0x%llx\n",
-			    vma->vm_start,
-			    (unsigned long long)pfn << PAGE_SHIFT);
-		break;
+		return uar_mmap(dev, command, vma, uuari);
 
 	case MLX5_IB_MMAP_GET_CONTIGUOUS_PAGES:
 		return -ENOSYS;
@@ -1108,7 +1166,7 @@ static int mlx5_ib_mmap(struct ib_ucontext *ibcontext, struct vm_area_struct *vm
 		if (vma->vm_end - vma->vm_start != PAGE_SIZE)
 			return -EINVAL;
 
-		if (vma->vm_flags & (VM_WRITE | VM_EXEC))
+		if (vma->vm_flags & VM_WRITE)
 			return -EPERM;
 
 		/* Don't expose to user-space information it shouldn't have */

drivers/infiniband/hw/mlx5/mlx5_ib.h

@@ -70,6 +70,8 @@ enum {
 enum mlx5_ib_mmap_cmd {
 	MLX5_IB_MMAP_REGULAR_PAGE		= 0,
 	MLX5_IB_MMAP_GET_CONTIGUOUS_PAGES	= 1,
+	MLX5_IB_MMAP_WC_PAGE			= 2,
+	MLX5_IB_MMAP_NC_PAGE			= 3,
 	/* 5 is chosen in order to be compatible with old versions of libmlx5 */
 	MLX5_IB_MMAP_CORE_CLOCK			= 5,
 };
@@ -356,6 +358,7 @@ enum mlx5_ib_qp_flags {
 	MLX5_IB_QP_SIGNATURE_HANDLING           = 1 << 5,
 	/* QP uses 1 as its source QP number */
 	MLX5_IB_QP_SQPN_QP1			= 1 << 6,
+	MLX5_IB_QP_CAP_SCATTER_FCS		= 1 << 7,
 };
 
 struct mlx5_umr_wr {
@@ -712,9 +715,8 @@ int mlx5_ib_dereg_mr(struct ib_mr *ibmr);
 struct ib_mr *mlx5_ib_alloc_mr(struct ib_pd *pd,
 			       enum ib_mr_type mr_type,
 			       u32 max_num_sg);
-int mlx5_ib_map_mr_sg(struct ib_mr *ibmr,
-		      struct scatterlist *sg,
-		      int sg_nents);
+int mlx5_ib_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg, int sg_nents,
+		      unsigned int *sg_offset);
 int mlx5_ib_process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num,
 			const struct ib_wc *in_wc, const struct ib_grh *in_grh,
 			const struct ib_mad_hdr *in, size_t in_mad_size,

drivers/infiniband/hw/mlx5/mr.c

@@ -1751,26 +1751,33 @@ int mlx5_ib_check_mr_status(struct ib_mr *ibmr, u32 check_mask,
 static int
 mlx5_ib_sg_to_klms(struct mlx5_ib_mr *mr,
 		   struct scatterlist *sgl,
-		   unsigned short sg_nents)
+		   unsigned short sg_nents,
+		   unsigned int *sg_offset_p)
 {
 	struct scatterlist *sg = sgl;
 	struct mlx5_klm *klms = mr->descs;
+	unsigned int sg_offset = sg_offset_p ? *sg_offset_p : 0;
 	u32 lkey = mr->ibmr.pd->local_dma_lkey;
 	int i;
 
-	mr->ibmr.iova = sg_dma_address(sg);
+	mr->ibmr.iova = sg_dma_address(sg) + sg_offset;
 	mr->ibmr.length = 0;
 	mr->ndescs = sg_nents;
 
 	for_each_sg(sgl, sg, sg_nents, i) {
 		if (unlikely(i > mr->max_descs))
 			break;
-		klms[i].va = cpu_to_be64(sg_dma_address(sg));
-		klms[i].bcount = cpu_to_be32(sg_dma_len(sg));
+		klms[i].va = cpu_to_be64(sg_dma_address(sg) + sg_offset);
+		klms[i].bcount = cpu_to_be32(sg_dma_len(sg) - sg_offset);
 		klms[i].key = cpu_to_be32(lkey);
 		mr->ibmr.length += sg_dma_len(sg);
+
+		sg_offset = 0;
 	}
 
+	if (sg_offset_p)
+		*sg_offset_p = sg_offset;
+
 	return i;
 }
 
@@ -1788,9 +1795,8 @@ static int mlx5_set_page(struct ib_mr *ibmr, u64 addr)
 	return 0;
 }
 
-int mlx5_ib_map_mr_sg(struct ib_mr *ibmr,
-		      struct scatterlist *sg,
-		      int sg_nents)
+int mlx5_ib_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg, int sg_nents,
+		      unsigned int *sg_offset)
 {
 	struct mlx5_ib_mr *mr = to_mmr(ibmr);
 	int n;
@@ -1802,9 +1808,10 @@ int mlx5_ib_map_mr_sg(struct ib_mr *ibmr,
 				   DMA_TO_DEVICE);
 
 	if (mr->access_mode == MLX5_ACCESS_MODE_KLM)
-		n = mlx5_ib_sg_to_klms(mr, sg, sg_nents);
+		n = mlx5_ib_sg_to_klms(mr, sg, sg_nents, sg_offset);
 	else
-		n = ib_sg_to_pages(ibmr, sg, sg_nents, mlx5_set_page);
+		n = ib_sg_to_pages(ibmr, sg, sg_nents, sg_offset,
+				mlx5_set_page);
 
 	ib_dma_sync_single_for_device(ibmr->device, mr->desc_map,
 				      mr->desc_size * mr->max_descs,

drivers/infiniband/hw/mlx5/qp.c

@@ -1028,6 +1028,7 @@ static int get_rq_pas_size(void *qpc)
 static int create_raw_packet_qp_rq(struct mlx5_ib_dev *dev,
 				   struct mlx5_ib_rq *rq, void *qpin)
 {
+	struct mlx5_ib_qp *mqp = rq->base.container_mibqp;
 	__be64 *pas;
 	__be64 *qp_pas;
 	void *in;
@@ -1051,6 +1052,9 @@ static int create_raw_packet_qp_rq(struct mlx5_ib_dev *dev,
 	MLX5_SET(rqc, rqc, user_index, MLX5_GET(qpc, qpc, user_index));
 	MLX5_SET(rqc, rqc, cqn, MLX5_GET(qpc, qpc, cqn_rcv));
 
+	if (mqp->flags & MLX5_IB_QP_CAP_SCATTER_FCS)
+		MLX5_SET(rqc, rqc, scatter_fcs, 1);
+
 	wq = MLX5_ADDR_OF(rqc, rqc, wq);
 	MLX5_SET(wq, wq, wq_type, MLX5_WQ_TYPE_CYCLIC);
 	MLX5_SET(wq, wq, end_padding_mode,
@@ -1136,11 +1140,12 @@ static int create_raw_packet_qp(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
 	}
 
 	if (qp->rq.wqe_cnt) {
+		rq->base.container_mibqp = qp;
+
 		err = create_raw_packet_qp_rq(dev, rq, in);
 		if (err)
 			goto err_destroy_sq;
 
-		rq->base.container_mibqp = qp;
 
 		err = create_raw_packet_qp_tir(dev, rq, tdn);
 		if (err)
@@ -1252,6 +1257,19 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd,
 			return -EOPNOTSUPP;
 		}
 
+	if (init_attr->create_flags & IB_QP_CREATE_SCATTER_FCS) {
+		if (init_attr->qp_type != IB_QPT_RAW_PACKET) {
+			mlx5_ib_dbg(dev, "Scatter FCS is supported only for Raw Packet QPs");
+			return -EOPNOTSUPP;
+		}
+		if (!MLX5_CAP_GEN(dev->mdev, eth_net_offloads) ||
+		    !MLX5_CAP_ETH(dev->mdev, scatter_fcs)) {
+			mlx5_ib_dbg(dev, "Scatter FCS isn't supported\n");
+			return -EOPNOTSUPP;
+		}
+		qp->flags |= MLX5_IB_QP_CAP_SCATTER_FCS;
+	}
+
 	if (init_attr->sq_sig_type == IB_SIGNAL_ALL_WR)
 		qp->sq_signal_bits = MLX5_WQE_CTRL_CQ_UPDATE;
 

drivers/infiniband/hw/nes/nes_utils.c

@@ -44,6 +44,7 @@
 #include <linux/ip.h>
 #include <linux/tcp.h>
 #include <linux/init.h>
+#include <linux/kernel.h>
 
 #include <asm/io.h>
 #include <asm/irq.h>
@@ -903,70 +904,15 @@ void nes_clc(unsigned long parm)
  */
 void nes_dump_mem(unsigned int dump_debug_level, void *addr, int length)
 {
-	char  xlate[] = {'0', '1', '2', '3', '4', '5', '6', '7', '8', '9',
-		'a', 'b', 'c', 'd', 'e', 'f'};
-	char  *ptr;
-	char  hex_buf[80];
-	char  ascii_buf[20];
-	int   num_char;
-	int   num_ascii;
-	int   num_hex;
-
 	if (!(nes_debug_level & dump_debug_level)) {
 		return;
 	}
 
-	ptr = addr;
 	if (length > 0x100) {
 		nes_debug(dump_debug_level, "Length truncated from %x to %x\n", length, 0x100);
 		length = 0x100;
 	}
-	nes_debug(dump_debug_level, "Address=0x%p, length=0x%x (%d)\n", ptr, length, length);
+	nes_debug(dump_debug_level, "Address=0x%p, length=0x%x (%d)\n", addr, length, length);
 
-	memset(ascii_buf, 0, 20);
-	memset(hex_buf, 0, 80);
-
-	num_ascii = 0;
-	num_hex = 0;
-	for (num_char = 0; num_char < length; num_char++) {
-		if (num_ascii == 8) {
-			ascii_buf[num_ascii++] = ' ';
-			hex_buf[num_hex++] = '-';
-			hex_buf[num_hex++] = ' ';
-		}
-
-		if (*ptr < 0x20 || *ptr > 0x7e)
-			ascii_buf[num_ascii++] = '.';
-		else
-			ascii_buf[num_ascii++] = *ptr;
-		hex_buf[num_hex++] = xlate[((*ptr & 0xf0) >> 4)];
-		hex_buf[num_hex++] = xlate[*ptr & 0x0f];
-		hex_buf[num_hex++] = ' ';
-		ptr++;
-
-		if (num_ascii >= 17) {
-			/* output line and reset */
-			nes_debug(dump_debug_level, "   %s |  %s\n", hex_buf, ascii_buf);
-			memset(ascii_buf, 0, 20);
-			memset(hex_buf, 0, 80);
-			num_ascii = 0;
-			num_hex = 0;
-		}
-	}
-
-	/* output the rest */
-	if (num_ascii) {
-		while (num_ascii < 17) {
-			if (num_ascii == 8) {
-				hex_buf[num_hex++] = ' ';
-				hex_buf[num_hex++] = ' ';
-			}
-			hex_buf[num_hex++] = ' ';
-			hex_buf[num_hex++] = ' ';
-			hex_buf[num_hex++] = ' ';
-			num_ascii++;
-		}
-
-		nes_debug(dump_debug_level, "   %s |  %s\n", hex_buf, ascii_buf);
-	}
+	print_hex_dump(KERN_ERR, PFX, DUMP_PREFIX_NONE, 16, 1, addr, length, true);
 }

drivers/infiniband/hw/nes/nes_verbs.c

@@ -402,15 +402,14 @@ static int nes_set_page(struct ib_mr *ibmr, u64 addr)
 	return 0;
 }
 
-static int nes_map_mr_sg(struct ib_mr *ibmr,
-			 struct scatterlist *sg,
-			 int sg_nents)
+static int nes_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg,
+			 int sg_nents, unsigned int *sg_offset)
 {
 	struct nes_mr *nesmr = to_nesmr(ibmr);
 
 	nesmr->npages = 0;
 
-	return ib_sg_to_pages(ibmr, sg, sg_nents, nes_set_page);
+	return ib_sg_to_pages(ibmr, sg, sg_nents, sg_offset, nes_set_page);
 }
 
 /**
@@ -981,7 +980,7 @@ static int nes_setup_mmap_qp(struct nes_qp *nesqp, struct nes_vnic *nesvnic,
 /**
  * nes_free_qp_mem() is to free up the qp's pci_alloc_consistent() memory.
  */
-static inline void nes_free_qp_mem(struct nes_device *nesdev,
+static void nes_free_qp_mem(struct nes_device *nesdev,
 		struct nes_qp *nesqp, int virt_wqs)
 {
 	unsigned long flags;
@@ -1315,6 +1314,8 @@ static struct ib_qp *nes_create_qp(struct ib_pd *ibpd,
 			nes_debug(NES_DBG_QP, "Invalid QP type: %d\n", init_attr->qp_type);
 			return ERR_PTR(-EINVAL);
 	}
+	init_completion(&nesqp->sq_drained);
+	init_completion(&nesqp->rq_drained);
 
 	nesqp->sig_all = (init_attr->sq_sig_type == IB_SIGNAL_ALL_WR);
 	init_timer(&nesqp->terminate_timer);
@@ -3452,6 +3453,29 @@ static int nes_post_recv(struct ib_qp *ibqp, struct ib_recv_wr *ib_wr,
 	return err;
 }
 
+/**
+ * nes_drain_sq - drain sq
+ * @ibqp: pointer to ibqp
+ */
+static void nes_drain_sq(struct ib_qp *ibqp)
+{
+	struct nes_qp *nesqp = to_nesqp(ibqp);
+
+	if (nesqp->hwqp.sq_tail != nesqp->hwqp.sq_head)
+		wait_for_completion(&nesqp->sq_drained);
+}
+
+/**
+ * nes_drain_rq - drain rq
+ * @ibqp: pointer to ibqp
+ */
+static void nes_drain_rq(struct ib_qp *ibqp)
+{
+	struct nes_qp *nesqp = to_nesqp(ibqp);
+
+	if (nesqp->hwqp.rq_tail != nesqp->hwqp.rq_head)
+		wait_for_completion(&nesqp->rq_drained);
+}
 
 /**
  * nes_poll_cq
@@ -3582,6 +3606,13 @@ static int nes_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry)
 				}
 			}
 
+			if (nesqp->iwarp_state > NES_CQP_QP_IWARP_STATE_RTS) {
+				if (nesqp->hwqp.sq_tail == nesqp->hwqp.sq_head)
+					complete(&nesqp->sq_drained);
+				if (nesqp->hwqp.rq_tail == nesqp->hwqp.rq_head)
+					complete(&nesqp->rq_drained);
+			}
+
 			entry->wr_id = wrid;
 			entry++;
 			cqe_count++;
@@ -3754,6 +3785,8 @@ struct nes_ib_device *nes_init_ofa_device(struct net_device *netdev)
 	nesibdev->ibdev.req_notify_cq = nes_req_notify_cq;
 	nesibdev->ibdev.post_send = nes_post_send;
 	nesibdev->ibdev.post_recv = nes_post_recv;
+	nesibdev->ibdev.drain_sq = nes_drain_sq;
+	nesibdev->ibdev.drain_rq = nes_drain_rq;
 
 	nesibdev->ibdev.iwcm = kzalloc(sizeof(*nesibdev->ibdev.iwcm), GFP_KERNEL);
 	if (nesibdev->ibdev.iwcm == NULL) {

drivers/infiniband/hw/nes/nes_verbs.h

@@ -189,6 +189,8 @@ struct nes_qp {
 	u8                    pau_pending;
 	u8                    pau_state;
 	__u64                 nesuqp_addr;
+	struct completion     sq_drained;
+	struct completion     rq_drained;
 };
 
 struct ib_mr *nes_reg_phys_mr(struct ib_pd *ib_pd,

drivers/infiniband/hw/ocrdma/ocrdma_verbs.c

@@ -3081,13 +3081,12 @@ static int ocrdma_set_page(struct ib_mr *ibmr, u64 addr)
 	return 0;
 }
 
-int ocrdma_map_mr_sg(struct ib_mr *ibmr,
-		     struct scatterlist *sg,
-		     int sg_nents)
+int ocrdma_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg, int sg_nents,
+		     unsigned int *sg_offset)
 {
 	struct ocrdma_mr *mr = get_ocrdma_mr(ibmr);
 
 	mr->npages = 0;
 
-	return ib_sg_to_pages(ibmr, sg, sg_nents, ocrdma_set_page);
+	return ib_sg_to_pages(ibmr, sg, sg_nents, sg_offset, ocrdma_set_page);
 }

drivers/infiniband/hw/ocrdma/ocrdma_verbs.h

@@ -122,8 +122,7 @@ struct ib_mr *ocrdma_reg_user_mr(struct ib_pd *, u64 start, u64 length,
 struct ib_mr *ocrdma_alloc_mr(struct ib_pd *pd,
 			      enum ib_mr_type mr_type,
 			      u32 max_num_sg);
-int ocrdma_map_mr_sg(struct ib_mr *ibmr,
-		     struct scatterlist *sg,
-		     int sg_nents);
+int ocrdma_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg, int sg_nents,
+		     unsigned int *sg_offset);
 
 #endif				/* __OCRDMA_VERBS_H__ */

drivers/infiniband/hw/qib/qib_init.c

@@ -1090,7 +1090,7 @@ void qib_free_devdata(struct qib_devdata *dd)
 	qib_dbg_ibdev_exit(&dd->verbs_dev);
 #endif
 	free_percpu(dd->int_counter);
-	ib_dealloc_device(&dd->verbs_dev.rdi.ibdev);
+	rvt_dealloc_device(&dd->verbs_dev.rdi);
 }
 
 u64 qib_int_counter(struct qib_devdata *dd)
@@ -1183,7 +1183,7 @@ struct qib_devdata *qib_alloc_devdata(struct pci_dev *pdev, size_t extra)
 bail:
 	if (!list_empty(&dd->list))
 		list_del_init(&dd->list);
-	ib_dealloc_device(&dd->verbs_dev.rdi.ibdev);
+	rvt_dealloc_device(&dd->verbs_dev.rdi);
 	return ERR_PTR(ret);
 }
 

drivers/infiniband/hw/qib/qib_rc.c

@@ -230,7 +230,7 @@ static int qib_make_rc_ack(struct qib_ibdev *dev, struct rvt_qp *qp,
  *
  * Return 1 if constructed; otherwise, return 0.
  */
-int qib_make_rc_req(struct rvt_qp *qp)
+int qib_make_rc_req(struct rvt_qp *qp, unsigned long *flags)
 {
 	struct qib_qp_priv *priv = qp->priv;
 	struct qib_ibdev *dev = to_idev(qp->ibqp.device);

drivers/infiniband/hw/qib/qib_ruc.c

@@ -739,7 +739,7 @@ void qib_do_send(struct rvt_qp *qp)
 	struct qib_qp_priv *priv = qp->priv;
 	struct qib_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
 	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
-	int (*make_req)(struct rvt_qp *qp);
+	int (*make_req)(struct rvt_qp *qp, unsigned long *flags);
 	unsigned long flags;
 
 	if ((qp->ibqp.qp_type == IB_QPT_RC ||
@@ -781,7 +781,7 @@ void qib_do_send(struct rvt_qp *qp)
 			qp->s_hdrwords = 0;
 			spin_lock_irqsave(&qp->s_lock, flags);
 		}
-	} while (make_req(qp));
+	} while (make_req(qp, &flags));
 
 	spin_unlock_irqrestore(&qp->s_lock, flags);
 }

drivers/infiniband/hw/qib/qib_uc.c

@@ -45,7 +45,7 @@
  *
  * Return 1 if constructed; otherwise, return 0.
  */
-int qib_make_uc_req(struct rvt_qp *qp)
+int qib_make_uc_req(struct rvt_qp *qp, unsigned long *flags)
 {
 	struct qib_qp_priv *priv = qp->priv;
 	struct qib_other_headers *ohdr;

drivers/infiniband/hw/qib/qib_ud.c

@@ -238,7 +238,7 @@ static void qib_ud_loopback(struct rvt_qp *sqp, struct rvt_swqe *swqe)
  *
  * Return 1 if constructed; otherwise, return 0.
  */
-int qib_make_ud_req(struct rvt_qp *qp)
+int qib_make_ud_req(struct rvt_qp *qp, unsigned long *flags)
 {
 	struct qib_qp_priv *priv = qp->priv;
 	struct qib_other_headers *ohdr;
@@ -294,7 +294,7 @@ int qib_make_ud_req(struct rvt_qp *qp)
 		this_cpu_inc(ibp->pmastats->n_unicast_xmit);
 		lid = ah_attr->dlid & ~((1 << ppd->lmc) - 1);
 		if (unlikely(lid == ppd->lid)) {
-			unsigned long flags;
+			unsigned long tflags = *flags;
 			/*
 			 * If DMAs are in progress, we can't generate
 			 * a completion for the loopback packet since
@@ -307,10 +307,10 @@ int qib_make_ud_req(struct rvt_qp *qp)
 				goto bail;
 			}
 			qp->s_cur = next_cur;
-			local_irq_save(flags);
-			spin_unlock_irqrestore(&qp->s_lock, flags);
+			spin_unlock_irqrestore(&qp->s_lock, tflags);
 			qib_ud_loopback(qp, wqe);
-			spin_lock_irqsave(&qp->s_lock, flags);
+			spin_lock_irqsave(&qp->s_lock, tflags);
+			*flags = tflags;
 			qib_send_complete(qp, wqe, IB_WC_SUCCESS);
 			goto done;
 		}

drivers/infiniband/hw/qib/qib_verbs.h

@@ -430,11 +430,11 @@ void qib_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe,
 
 void qib_send_rc_ack(struct rvt_qp *qp);
 
-int qib_make_rc_req(struct rvt_qp *qp);
+int qib_make_rc_req(struct rvt_qp *qp, unsigned long *flags);
 
-int qib_make_uc_req(struct rvt_qp *qp);
+int qib_make_uc_req(struct rvt_qp *qp, unsigned long *flags);
 
-int qib_make_ud_req(struct rvt_qp *qp);
+int qib_make_ud_req(struct rvt_qp *qp, unsigned long *flags);
 
 int qib_register_ib_device(struct qib_devdata *);
 

drivers/infiniband/sw/rdmavt/qp.c

@@ -829,13 +829,13 @@ struct ib_qp *rvt_create_qp(struct ib_pd *ibpd,
 	case IB_QPT_SMI:
 	case IB_QPT_GSI:
 	case IB_QPT_UD:
-		qp->allowed_ops = IB_OPCODE_UD_SEND_ONLY & RVT_OPCODE_QP_MASK;
+		qp->allowed_ops = IB_OPCODE_UD;
 		break;
 	case IB_QPT_RC:
-		qp->allowed_ops = IB_OPCODE_RC_SEND_ONLY & RVT_OPCODE_QP_MASK;
+		qp->allowed_ops = IB_OPCODE_RC;
 		break;
 	case IB_QPT_UC:
-		qp->allowed_ops = IB_OPCODE_UC_SEND_ONLY & RVT_OPCODE_QP_MASK;
+		qp->allowed_ops = IB_OPCODE_UC;
 		break;
 	default:
 		ret = ERR_PTR(-EINVAL);

drivers/infiniband/sw/rdmavt/vt.c

@@ -106,6 +106,19 @@ struct rvt_dev_info *rvt_alloc_device(size_t size, int nports)
 }
 EXPORT_SYMBOL(rvt_alloc_device);
 
+/**
+ * rvt_dealloc_device - deallocate rdi
+ * @rdi: structure to free
+ *
+ * Free a structure allocated with rvt_alloc_device()
+ */
+void rvt_dealloc_device(struct rvt_dev_info *rdi)
+{
+	kfree(rdi->ports);
+	ib_dealloc_device(&rdi->ibdev);
+}
+EXPORT_SYMBOL(rvt_dealloc_device);
+
 static int rvt_query_device(struct ib_device *ibdev,
 			    struct ib_device_attr *props,
 			    struct ib_udata *uhw)

drivers/infiniband/ulp/ipoib/ipoib_ethtool.c

@@ -36,6 +36,27 @@
 
 #include "ipoib.h"
 
+struct ipoib_stats {
+	char stat_string[ETH_GSTRING_LEN];
+	int stat_offset;
+};
+
+#define IPOIB_NETDEV_STAT(m) { \
+		.stat_string = #m, \
+		.stat_offset = offsetof(struct rtnl_link_stats64, m) }
+
+static const struct ipoib_stats ipoib_gstrings_stats[] = {
+	IPOIB_NETDEV_STAT(rx_packets),
+	IPOIB_NETDEV_STAT(tx_packets),
+	IPOIB_NETDEV_STAT(rx_bytes),
+	IPOIB_NETDEV_STAT(tx_bytes),
+	IPOIB_NETDEV_STAT(tx_errors),
+	IPOIB_NETDEV_STAT(rx_dropped),
+	IPOIB_NETDEV_STAT(tx_dropped)
+};
+
+#define IPOIB_GLOBAL_STATS_LEN	ARRAY_SIZE(ipoib_gstrings_stats)
+
 static void ipoib_get_drvinfo(struct net_device *netdev,
 			      struct ethtool_drvinfo *drvinfo)
 {
@@ -92,11 +113,57 @@ static int ipoib_set_coalesce(struct net_device *dev,
 
 	return 0;
 }
+static void ipoib_get_ethtool_stats(struct net_device *dev,
+				    struct ethtool_stats __always_unused *stats,
+				    u64 *data)
+{
+	int i;
+	struct net_device_stats *net_stats = &dev->stats;
+	u8 *p = (u8 *)net_stats;
+
+	for (i = 0; i < IPOIB_GLOBAL_STATS_LEN; i++)
+		data[i] = *(u64 *)(p + ipoib_gstrings_stats[i].stat_offset);
+
+}
+static void ipoib_get_strings(struct net_device __always_unused *dev,
+			      u32 stringset, u8 *data)
+{
+	u8 *p = data;
+	int i;
+
+	switch (stringset) {
+	case ETH_SS_STATS:
+		for (i = 0; i < IPOIB_GLOBAL_STATS_LEN; i++) {
+			memcpy(p, ipoib_gstrings_stats[i].stat_string,
+				ETH_GSTRING_LEN);
+			p += ETH_GSTRING_LEN;
+		}
+		break;
+	case ETH_SS_TEST:
+	default:
+		break;
+	}
+}
+static int ipoib_get_sset_count(struct net_device __always_unused *dev,
+				 int sset)
+{
+	switch (sset) {
+	case ETH_SS_STATS:
+		return IPOIB_GLOBAL_STATS_LEN;
+	case ETH_SS_TEST:
+	default:
+		break;
+	}
+	return -EOPNOTSUPP;
+}
 
 static const struct ethtool_ops ipoib_ethtool_ops = {
 	.get_drvinfo		= ipoib_get_drvinfo,
 	.get_coalesce		= ipoib_get_coalesce,
 	.set_coalesce		= ipoib_set_coalesce,
+	.get_strings		= ipoib_get_strings,
+	.get_ethtool_stats	= ipoib_get_ethtool_stats,
+	.get_sset_count		= ipoib_get_sset_count,
 };
 
 void ipoib_set_ethtool_ops(struct net_device *dev)

drivers/infiniband/ulp/ipoib/ipoib_ib.c

@@ -51,8 +51,6 @@ MODULE_PARM_DESC(data_debug_level,
 		 "Enable data path debug tracing if > 0");
 #endif
 
-static DEFINE_MUTEX(pkey_mutex);
-
 struct ipoib_ah *ipoib_create_ah(struct net_device *dev,
 				 struct ib_pd *pd, struct ib_ah_attr *attr)
 {

drivers/infiniband/ulp/iser/iser_memory.c

@@ -236,7 +236,7 @@ int iser_fast_reg_fmr(struct iscsi_iser_task *iser_task,
 	page_vec->npages = 0;
 	page_vec->fake_mr.page_size = SIZE_4K;
 	plen = ib_sg_to_pages(&page_vec->fake_mr, mem->sg,
-			      mem->size, iser_set_page);
+			      mem->size, NULL, iser_set_page);
 	if (unlikely(plen < mem->size)) {
 		iser_err("page vec too short to hold this SG\n");
 		iser_data_buf_dump(mem, device->ib_device);
@@ -446,7 +446,7 @@ static int iser_fast_reg_mr(struct iscsi_iser_task *iser_task,
 
 	ib_update_fast_reg_key(mr, ib_inc_rkey(mr->rkey));
 
-	n = ib_map_mr_sg(mr, mem->sg, mem->size, SIZE_4K);
+	n = ib_map_mr_sg(mr, mem->sg, mem->size, NULL, SIZE_4K);
 	if (unlikely(n != mem->size)) {
 		iser_err("failed to map sg (%d/%d)\n",
 			 n, mem->size);

drivers/infiniband/ulp/isert/ib_isert.h

@@ -3,6 +3,7 @@
 #include <linux/in6.h>
 #include <rdma/ib_verbs.h>
 #include <rdma/rdma_cm.h>
+#include <rdma/rw.h>
 #include <scsi/iser.h>
 
 
@@ -53,10 +54,7 @@
 
 #define ISERT_MIN_POSTED_RX	(ISCSI_DEF_XMIT_CMDS_MAX >> 2)
 
-#define ISERT_INFLIGHT_DATAOUTS	8
-
-#define ISERT_QP_MAX_REQ_DTOS	(ISCSI_DEF_XMIT_CMDS_MAX *    \
-				(1 + ISERT_INFLIGHT_DATAOUTS) + \
+#define ISERT_QP_MAX_REQ_DTOS	(ISCSI_DEF_XMIT_CMDS_MAX +    \
 				ISERT_MAX_TX_MISC_PDUS	+ \
 				ISERT_MAX_RX_MISC_PDUS)
 
@@ -71,13 +69,6 @@ enum isert_desc_type {
 	ISCSI_TX_DATAIN
 };
 
-enum iser_ib_op_code {
-	ISER_IB_RECV,
-	ISER_IB_SEND,
-	ISER_IB_RDMA_WRITE,
-	ISER_IB_RDMA_READ,
-};
-
 enum iser_conn_state {
 	ISER_CONN_INIT,
 	ISER_CONN_UP,
@@ -118,42 +109,6 @@ static inline struct iser_tx_desc *cqe_to_tx_desc(struct ib_cqe *cqe)
 	return container_of(cqe, struct iser_tx_desc, tx_cqe);
 }
 
-
-enum isert_indicator {
-	ISERT_PROTECTED		= 1 << 0,
-	ISERT_DATA_KEY_VALID	= 1 << 1,
-	ISERT_PROT_KEY_VALID	= 1 << 2,
-	ISERT_SIG_KEY_VALID	= 1 << 3,
-};
-
-struct pi_context {
-	struct ib_mr		       *prot_mr;
-	struct ib_mr		       *sig_mr;
-};
-
-struct fast_reg_descriptor {
-	struct list_head		list;
-	struct ib_mr		       *data_mr;
-	u8				ind;
-	struct pi_context	       *pi_ctx;
-};
-
-struct isert_data_buf {
-	struct scatterlist     *sg;
-	int			nents;
-	u32			sg_off;
-	u32			len; /* cur_rdma_length */
-	u32			offset;
-	unsigned int		dma_nents;
-	enum dma_data_direction dma_dir;
-};
-
-enum {
-	DATA = 0,
-	PROT = 1,
-	SIG = 2,
-};
-
 struct isert_cmd {
 	uint32_t		read_stag;
 	uint32_t		write_stag;
@@ -166,16 +121,7 @@ struct isert_cmd {
 	struct iscsi_cmd	*iscsi_cmd;
 	struct iser_tx_desc	tx_desc;
 	struct iser_rx_desc	*rx_desc;
-	enum iser_ib_op_code	iser_ib_op;
-	struct ib_sge		*ib_sge;
-	struct ib_sge		s_ib_sge;
-	int			rdma_wr_num;
-	struct ib_rdma_wr	*rdma_wr;
-	struct ib_rdma_wr	s_rdma_wr;
-	struct ib_sge		ib_sg[3];
-	struct isert_data_buf	data;
-	struct isert_data_buf	prot;
-	struct fast_reg_descriptor *fr_desc;
+	struct rdma_rw_ctx	rw;
 	struct work_struct	comp_work;
 	struct scatterlist	sg;
 };
@@ -210,10 +156,6 @@ struct isert_conn {
 	struct isert_device	*device;
 	struct mutex		mutex;
 	struct kref		kref;
-	struct list_head	fr_pool;
-	int			fr_pool_size;
-	/* lock to protect fastreg pool */
-	spinlock_t		pool_lock;
 	struct work_struct	release_work;
 	bool                    logout_posted;
 	bool                    snd_w_inv;
@@ -236,7 +178,6 @@ struct isert_comp {
 };
 
 struct isert_device {
-	int			use_fastreg;
 	bool			pi_capable;
 	int			refcount;
 	struct ib_device	*ib_device;
@@ -244,10 +185,6 @@ struct isert_device {
 	struct isert_comp	*comps;
 	int                     comps_used;
 	struct list_head	dev_node;
-	int			(*reg_rdma_mem)(struct isert_cmd *isert_cmd,
-						struct iscsi_conn *conn);
-	void			(*unreg_rdma_mem)(struct isert_cmd *isert_cmd,
-						  struct isert_conn *isert_conn);
 };
 
 struct isert_np {

drivers/infiniband/ulp/srp/ib_srp.c

@@ -70,6 +70,7 @@ static unsigned int indirect_sg_entries;
 static bool allow_ext_sg;
 static bool prefer_fr = true;
 static bool register_always = true;
+static bool never_register;
 static int topspin_workarounds = 1;
 
 module_param(srp_sg_tablesize, uint, 0444);
@@ -99,6 +100,9 @@ module_param(register_always, bool, 0444);
 MODULE_PARM_DESC(register_always,
 		 "Use memory registration even for contiguous memory regions");
 
+module_param(never_register, bool, 0444);
+MODULE_PARM_DESC(never_register, "Never register memory");
+
 static const struct kernel_param_ops srp_tmo_ops;
 
 static int srp_reconnect_delay = 10;
@@ -316,7 +320,7 @@ static struct ib_fmr_pool *srp_alloc_fmr_pool(struct srp_target_port *target)
 	struct ib_fmr_pool_param fmr_param;
 
 	memset(&fmr_param, 0, sizeof(fmr_param));
-	fmr_param.pool_size	    = target->scsi_host->can_queue;
+	fmr_param.pool_size	    = target->mr_pool_size;
 	fmr_param.dirty_watermark   = fmr_param.pool_size / 4;
 	fmr_param.cache		    = 1;
 	fmr_param.max_pages_per_fmr = dev->max_pages_per_mr;
@@ -441,23 +445,22 @@ static struct srp_fr_pool *srp_alloc_fr_pool(struct srp_target_port *target)
 {
 	struct srp_device *dev = target->srp_host->srp_dev;
 
-	return srp_create_fr_pool(dev->dev, dev->pd,
-				  target->scsi_host->can_queue,
+	return srp_create_fr_pool(dev->dev, dev->pd, target->mr_pool_size,
 				  dev->max_pages_per_mr);
 }
 
 /**
  * srp_destroy_qp() - destroy an RDMA queue pair
- * @ch: SRP RDMA channel.
+ * @qp: RDMA queue pair.
  *
  * Drain the qp before destroying it.  This avoids that the receive
  * completion handler can access the queue pair while it is
  * being destroyed.
  */
-static void srp_destroy_qp(struct srp_rdma_ch *ch)
+static void srp_destroy_qp(struct ib_qp *qp)
 {
-	ib_drain_rq(ch->qp);
-	ib_destroy_qp(ch->qp);
+	ib_drain_rq(qp);
+	ib_destroy_qp(qp);
 }
 
 static int srp_create_ch_ib(struct srp_rdma_ch *ch)
@@ -469,7 +472,7 @@ static int srp_create_ch_ib(struct srp_rdma_ch *ch)
 	struct ib_qp *qp;
 	struct ib_fmr_pool *fmr_pool = NULL;
 	struct srp_fr_pool *fr_pool = NULL;
-	const int m = dev->use_fast_reg ? 3 : 1;
+	const int m = 1 + dev->use_fast_reg * target->mr_per_cmd * 2;
 	int ret;
 
 	init_attr = kzalloc(sizeof *init_attr, GFP_KERNEL);
@@ -530,7 +533,7 @@ static int srp_create_ch_ib(struct srp_rdma_ch *ch)
 	}
 
 	if (ch->qp)
-		srp_destroy_qp(ch);
+		srp_destroy_qp(ch->qp);
 	if (ch->recv_cq)
 		ib_free_cq(ch->recv_cq);
 	if (ch->send_cq)
@@ -554,7 +557,7 @@ static int srp_create_ch_ib(struct srp_rdma_ch *ch)
 	return 0;
 
 err_qp:
-	srp_destroy_qp(ch);
+	srp_destroy_qp(qp);
 
 err_send_cq:
 	ib_free_cq(send_cq);
@@ -597,7 +600,7 @@ static void srp_free_ch_ib(struct srp_target_port *target,
 			ib_destroy_fmr_pool(ch->fmr_pool);
 	}
 
-	srp_destroy_qp(ch);
+	srp_destroy_qp(ch->qp);
 	ib_free_cq(ch->send_cq);
 	ib_free_cq(ch->recv_cq);
 
@@ -850,7 +853,7 @@ static int srp_alloc_req_data(struct srp_rdma_ch *ch)
 
 	for (i = 0; i < target->req_ring_size; ++i) {
 		req = &ch->req_ring[i];
-		mr_list = kmalloc(target->cmd_sg_cnt * sizeof(void *),
+		mr_list = kmalloc(target->mr_per_cmd * sizeof(void *),
 				  GFP_KERNEL);
 		if (!mr_list)
 			goto out;
@@ -1112,7 +1115,7 @@ static struct scsi_cmnd *srp_claim_req(struct srp_rdma_ch *ch,
 }
 
 /**
- * srp_free_req() - Unmap data and add request to the free request list.
+ * srp_free_req() - Unmap data and adjust ch->req_lim.
  * @ch:     SRP RDMA channel.
  * @req:    Request to be freed.
  * @scmnd:  SCSI command associated with @req.
@@ -1299,9 +1302,16 @@ static void srp_reg_mr_err_done(struct ib_cq *cq, struct ib_wc *wc)
 	srp_handle_qp_err(cq, wc, "FAST REG");
 }
 
+/*
+ * Map up to sg_nents elements of state->sg where *sg_offset_p is the offset
+ * where to start in the first element. If sg_offset_p != NULL then
+ * *sg_offset_p is updated to the offset in state->sg[retval] of the first
+ * byte that has not yet been mapped.
+ */
 static int srp_map_finish_fr(struct srp_map_state *state,
 			     struct srp_request *req,
-			     struct srp_rdma_ch *ch, int sg_nents)
+			     struct srp_rdma_ch *ch, int sg_nents,
+			     unsigned int *sg_offset_p)
 {
 	struct srp_target_port *target = ch->target;
 	struct srp_device *dev = target->srp_host->srp_dev;
@@ -1316,13 +1326,14 @@ static int srp_map_finish_fr(struct srp_map_state *state,
 
 	WARN_ON_ONCE(!dev->use_fast_reg);
 
-	if (sg_nents == 0)
-		return 0;
-
 	if (sg_nents == 1 && target->global_mr) {
-		srp_map_desc(state, sg_dma_address(state->sg),
-			     sg_dma_len(state->sg),
+		unsigned int sg_offset = sg_offset_p ? *sg_offset_p : 0;
+
+		srp_map_desc(state, sg_dma_address(state->sg) + sg_offset,
+			     sg_dma_len(state->sg) - sg_offset,
 			     target->global_mr->rkey);
+		if (sg_offset_p)
+			*sg_offset_p = 0;
 		return 1;
 	}
 
@@ -1333,9 +1344,17 @@ static int srp_map_finish_fr(struct srp_map_state *state,
 	rkey = ib_inc_rkey(desc->mr->rkey);
 	ib_update_fast_reg_key(desc->mr, rkey);
 
-	n = ib_map_mr_sg(desc->mr, state->sg, sg_nents, dev->mr_page_size);
-	if (unlikely(n < 0))
+	n = ib_map_mr_sg(desc->mr, state->sg, sg_nents, sg_offset_p,
+			 dev->mr_page_size);
+	if (unlikely(n < 0)) {
+		srp_fr_pool_put(ch->fr_pool, &desc, 1);
+		pr_debug("%s: ib_map_mr_sg(%d, %d) returned %d.\n",
+			 dev_name(&req->scmnd->device->sdev_gendev), sg_nents,
+			 sg_offset_p ? *sg_offset_p : -1, n);
 		return n;
+	}
+
+	WARN_ON_ONCE(desc->mr->length == 0);
 
 	req->reg_cqe.done = srp_reg_mr_err_done;
 
@@ -1357,8 +1376,10 @@ static int srp_map_finish_fr(struct srp_map_state *state,
 		     desc->mr->length, desc->mr->rkey);
 
 	err = ib_post_send(ch->qp, &wr.wr, &bad_wr);
-	if (unlikely(err))
+	if (unlikely(err)) {
+		WARN_ON_ONCE(err == -ENOMEM);
 		return err;
+	}
 
 	return n;
 }
@@ -1398,7 +1419,7 @@ static int srp_map_sg_entry(struct srp_map_state *state,
 	/*
 	 * If the last entry of the MR wasn't a full page, then we need to
 	 * close it out and start a new one -- we can only merge at page
-	 * boundries.
+	 * boundaries.
 	 */
 	ret = 0;
 	if (len != dev->mr_page_size)
@@ -1413,10 +1434,9 @@ static int srp_map_sg_fmr(struct srp_map_state *state, struct srp_rdma_ch *ch,
 	struct scatterlist *sg;
 	int i, ret;
 
-	state->desc = req->indirect_desc;
 	state->pages = req->map_page;
 	state->fmr.next = req->fmr_list;
-	state->fmr.end = req->fmr_list + ch->target->cmd_sg_cnt;
+	state->fmr.end = req->fmr_list + ch->target->mr_per_cmd;
 
 	for_each_sg(scat, sg, count, i) {
 		ret = srp_map_sg_entry(state, ch, sg, i);
@@ -1428,8 +1448,6 @@ static int srp_map_sg_fmr(struct srp_map_state *state, struct srp_rdma_ch *ch,
 	if (ret)
 		return ret;
 
-	req->nmdesc = state->nmdesc;
-
 	return 0;
 }
 
@@ -1437,15 +1455,20 @@ static int srp_map_sg_fr(struct srp_map_state *state, struct srp_rdma_ch *ch,
 			 struct srp_request *req, struct scatterlist *scat,
 			 int count)
 {
+	unsigned int sg_offset = 0;
+
 	state->desc = req->indirect_desc;
 	state->fr.next = req->fr_list;
-	state->fr.end = req->fr_list + ch->target->cmd_sg_cnt;
+	state->fr.end = req->fr_list + ch->target->mr_per_cmd;
 	state->sg = scat;
 
+	if (count == 0)
+		return 0;
+
 	while (count) {
 		int i, n;
 
-		n = srp_map_finish_fr(state, req, ch, count);
+		n = srp_map_finish_fr(state, req, ch, count, &sg_offset);
 		if (unlikely(n < 0))
 			return n;
 
@@ -1454,8 +1477,6 @@ static int srp_map_sg_fr(struct srp_map_state *state, struct srp_rdma_ch *ch,
 			state->sg = sg_next(state->sg);
 	}
 
-	req->nmdesc = state->nmdesc;
-
 	return 0;
 }
 
@@ -1475,8 +1496,6 @@ static int srp_map_sg_dma(struct srp_map_state *state, struct srp_rdma_ch *ch,
 			     target->global_mr->rkey);
 	}
 
-	req->nmdesc = state->nmdesc;
-
 	return 0;
 }
 
@@ -1509,14 +1528,15 @@ static int srp_map_idb(struct srp_rdma_ch *ch, struct srp_request *req,
 
 	if (dev->use_fast_reg) {
 		state.sg = idb_sg;
-		sg_set_buf(idb_sg, req->indirect_desc, idb_len);
+		sg_init_one(idb_sg, req->indirect_desc, idb_len);
 		idb_sg->dma_address = req->indirect_dma_addr; /* hack! */
 #ifdef CONFIG_NEED_SG_DMA_LENGTH
 		idb_sg->dma_length = idb_sg->length;	      /* hack^2 */
 #endif
-		ret = srp_map_finish_fr(&state, req, ch, 1);
+		ret = srp_map_finish_fr(&state, req, ch, 1, NULL);
 		if (ret < 0)
 			return ret;
+		WARN_ON_ONCE(ret < 1);
 	} else if (dev->use_fmr) {
 		state.pages = idb_pages;
 		state.pages[0] = (req->indirect_dma_addr &
@@ -1534,6 +1554,41 @@ static int srp_map_idb(struct srp_rdma_ch *ch, struct srp_request *req,
 	return 0;
 }
 
+#if defined(DYNAMIC_DATA_DEBUG)
+static void srp_check_mapping(struct srp_map_state *state,
+			      struct srp_rdma_ch *ch, struct srp_request *req,
+			      struct scatterlist *scat, int count)
+{
+	struct srp_device *dev = ch->target->srp_host->srp_dev;
+	struct srp_fr_desc **pfr;
+	u64 desc_len = 0, mr_len = 0;
+	int i;
+
+	for (i = 0; i < state->ndesc; i++)
+		desc_len += be32_to_cpu(req->indirect_desc[i].len);
+	if (dev->use_fast_reg)
+		for (i = 0, pfr = req->fr_list; i < state->nmdesc; i++, pfr++)
+			mr_len += (*pfr)->mr->length;
+	else if (dev->use_fmr)
+		for (i = 0; i < state->nmdesc; i++)
+			mr_len += be32_to_cpu(req->indirect_desc[i].len);
+	if (desc_len != scsi_bufflen(req->scmnd) ||
+	    mr_len > scsi_bufflen(req->scmnd))
+		pr_err("Inconsistent: scsi len %d <> desc len %lld <> mr len %lld; ndesc %d; nmdesc = %d\n",
+		       scsi_bufflen(req->scmnd), desc_len, mr_len,
+		       state->ndesc, state->nmdesc);
+}
+#endif
+
+/**
+ * srp_map_data() - map SCSI data buffer onto an SRP request
+ * @scmnd: SCSI command to map
+ * @ch: SRP RDMA channel
+ * @req: SRP request
+ *
+ * Returns the length in bytes of the SRP_CMD IU or a negative value if
+ * mapping failed.
+ */
 static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_rdma_ch *ch,
 			struct srp_request *req)
 {
@@ -1601,11 +1656,23 @@ static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_rdma_ch *ch,
 
 	memset(&state, 0, sizeof(state));
 	if (dev->use_fast_reg)
-		srp_map_sg_fr(&state, ch, req, scat, count);
+		ret = srp_map_sg_fr(&state, ch, req, scat, count);
 	else if (dev->use_fmr)
-		srp_map_sg_fmr(&state, ch, req, scat, count);
+		ret = srp_map_sg_fmr(&state, ch, req, scat, count);
 	else
-		srp_map_sg_dma(&state, ch, req, scat, count);
+		ret = srp_map_sg_dma(&state, ch, req, scat, count);
+	req->nmdesc = state.nmdesc;
+	if (ret < 0)
+		goto unmap;
+
+#if defined(DYNAMIC_DEBUG)
+	{
+		DEFINE_DYNAMIC_DEBUG_METADATA(ddm,
+			"Memory mapping consistency check");
+		if (unlikely(ddm.flags & _DPRINTK_FLAGS_PRINT))
+			srp_check_mapping(&state, ch, req, scat, count);
+	}
+#endif
 
 	/* We've mapped the request, now pull as much of the indirect
 	 * descriptor table as we can into the command buffer. If this
@@ -1628,7 +1695,8 @@ static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_rdma_ch *ch,
 						!target->allow_ext_sg)) {
 		shost_printk(KERN_ERR, target->scsi_host,
 			     "Could not fit S/G list into SRP_CMD\n");
-		return -EIO;
+		ret = -EIO;
+		goto unmap;
 	}
 
 	count = min(state.ndesc, target->cmd_sg_cnt);
@@ -1646,7 +1714,7 @@ static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_rdma_ch *ch,
 		ret = srp_map_idb(ch, req, state.gen.next, state.gen.end,
 				  idb_len, &idb_rkey);
 		if (ret < 0)
-			return ret;
+			goto unmap;
 		req->nmdesc++;
 	} else {
 		idb_rkey = cpu_to_be32(target->global_mr->rkey);
@@ -1672,6 +1740,12 @@ static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_rdma_ch *ch,
 		cmd->buf_fmt = fmt;
 
 	return len;
+
+unmap:
+	srp_unmap_data(scmnd, ch, req);
+	if (ret == -ENOMEM && req->nmdesc >= target->mr_pool_size)
+		ret = -E2BIG;
+	return ret;
 }
 
 /*
@@ -2564,6 +2638,20 @@ static int srp_reset_host(struct scsi_cmnd *scmnd)
 	return srp_reconnect_rport(target->rport) == 0 ? SUCCESS : FAILED;
 }
 
+static int srp_slave_alloc(struct scsi_device *sdev)
+{
+	struct Scsi_Host *shost = sdev->host;
+	struct srp_target_port *target = host_to_target(shost);
+	struct srp_device *srp_dev = target->srp_host->srp_dev;
+	struct ib_device *ibdev = srp_dev->dev;
+
+	if (!(ibdev->attrs.device_cap_flags & IB_DEVICE_SG_GAPS_REG))
+		blk_queue_virt_boundary(sdev->request_queue,
+					~srp_dev->mr_page_mask);
+
+	return 0;
+}
+
 static int srp_slave_configure(struct scsi_device *sdev)
 {
 	struct Scsi_Host *shost = sdev->host;
@@ -2755,6 +2843,7 @@ static struct scsi_host_template srp_template = {
 	.module				= THIS_MODULE,
 	.name				= "InfiniBand SRP initiator",
 	.proc_name			= DRV_NAME,
+	.slave_alloc			= srp_slave_alloc,
 	.slave_configure		= srp_slave_configure,
 	.info				= srp_target_info,
 	.queuecommand			= srp_queuecommand,
@@ -2829,7 +2918,7 @@ static int srp_add_target(struct srp_host *host, struct srp_target_port *target)
 		goto out;
 	}
 
-	pr_debug(PFX "%s: SCSI scan succeeded - detected %d LUNs\n",
+	pr_debug("%s: SCSI scan succeeded - detected %d LUNs\n",
 		 dev_name(&target->scsi_host->shost_gendev),
 		 srp_sdev_count(target->scsi_host));
 
@@ -3161,6 +3250,7 @@ static ssize_t srp_create_target(struct device *dev,
 	struct srp_device *srp_dev = host->srp_dev;
 	struct ib_device *ibdev = srp_dev->dev;
 	int ret, node_idx, node, cpu, i;
+	unsigned int max_sectors_per_mr, mr_per_cmd = 0;
 	bool multich = false;
 
 	target_host = scsi_host_alloc(&srp_template,
@@ -3217,7 +3307,33 @@ static ssize_t srp_create_target(struct device *dev,
 		target->sg_tablesize = target->cmd_sg_cnt;
 	}
 
+	if (srp_dev->use_fast_reg || srp_dev->use_fmr) {
+		/*
+		 * FR and FMR can only map one HCA page per entry. If the
+		 * start address is not aligned on a HCA page boundary two
+		 * entries will be used for the head and the tail although
+		 * these two entries combined contain at most one HCA page of
+		 * data. Hence the "+ 1" in the calculation below.
+		 *
+		 * The indirect data buffer descriptor is contiguous so the
+		 * memory for that buffer will only be registered if
+		 * register_always is true. Hence add one to mr_per_cmd if
+		 * register_always has been set.
+		 */
+		max_sectors_per_mr = srp_dev->max_pages_per_mr <<
+				  (ilog2(srp_dev->mr_page_size) - 9);
+		mr_per_cmd = register_always +
+			(target->scsi_host->max_sectors + 1 +
+			 max_sectors_per_mr - 1) / max_sectors_per_mr;
+		pr_debug("max_sectors = %u; max_pages_per_mr = %u; mr_page_size = %u; max_sectors_per_mr = %u; mr_per_cmd = %u\n",
+			 target->scsi_host->max_sectors,
+			 srp_dev->max_pages_per_mr, srp_dev->mr_page_size,
+			 max_sectors_per_mr, mr_per_cmd);
+	}
+
 	target_host->sg_tablesize = target->sg_tablesize;
+	target->mr_pool_size = target->scsi_host->can_queue * mr_per_cmd;
+	target->mr_per_cmd = mr_per_cmd;
 	target->indirect_size = target->sg_tablesize *
 				sizeof (struct srp_direct_buf);
 	target->max_iu_len = sizeof (struct srp_cmd) +
@@ -3414,17 +3530,6 @@ static void srp_add_one(struct ib_device *device)
 	if (!srp_dev)
 		return;
 
-	srp_dev->has_fmr = (device->alloc_fmr && device->dealloc_fmr &&
-			    device->map_phys_fmr && device->unmap_fmr);
-	srp_dev->has_fr = (device->attrs.device_cap_flags &
-			   IB_DEVICE_MEM_MGT_EXTENSIONS);
-	if (!srp_dev->has_fmr && !srp_dev->has_fr)
-		dev_warn(&device->dev, "neither FMR nor FR is supported\n");
-
-	srp_dev->use_fast_reg = (srp_dev->has_fr &&
-				 (!srp_dev->has_fmr || prefer_fr));
-	srp_dev->use_fmr = !srp_dev->use_fast_reg && srp_dev->has_fmr;
-
 	/*
 	 * Use the smallest page size supported by the HCA, down to a
 	 * minimum of 4096 bytes. We're unlikely to build large sglists
@@ -3435,8 +3540,25 @@ static void srp_add_one(struct ib_device *device)
 	srp_dev->mr_page_mask	= ~((u64) srp_dev->mr_page_size - 1);
 	max_pages_per_mr	= device->attrs.max_mr_size;
 	do_div(max_pages_per_mr, srp_dev->mr_page_size);
+	pr_debug("%s: %llu / %u = %llu <> %u\n", __func__,
+		 device->attrs.max_mr_size, srp_dev->mr_page_size,
+		 max_pages_per_mr, SRP_MAX_PAGES_PER_MR);
 	srp_dev->max_pages_per_mr = min_t(u64, SRP_MAX_PAGES_PER_MR,
 					  max_pages_per_mr);
+
+	srp_dev->has_fmr = (device->alloc_fmr && device->dealloc_fmr &&
+			    device->map_phys_fmr && device->unmap_fmr);
+	srp_dev->has_fr = (device->attrs.device_cap_flags &
+			   IB_DEVICE_MEM_MGT_EXTENSIONS);
+	if (!never_register && !srp_dev->has_fmr && !srp_dev->has_fr) {
+		dev_warn(&device->dev, "neither FMR nor FR is supported\n");
+	} else if (!never_register &&
+		   device->attrs.max_mr_size >= 2 * srp_dev->mr_page_size) {
+		srp_dev->use_fast_reg = (srp_dev->has_fr &&
+					 (!srp_dev->has_fmr || prefer_fr));
+		srp_dev->use_fmr = !srp_dev->use_fast_reg && srp_dev->has_fmr;
+	}
+
 	if (srp_dev->use_fast_reg) {
 		srp_dev->max_pages_per_mr =
 			min_t(u32, srp_dev->max_pages_per_mr,
@@ -3456,7 +3578,8 @@ static void srp_add_one(struct ib_device *device)
 	if (IS_ERR(srp_dev->pd))
 		goto free_dev;
 
-	if (!register_always || (!srp_dev->has_fmr && !srp_dev->has_fr)) {
+	if (never_register || !register_always ||
+	    (!srp_dev->has_fmr && !srp_dev->has_fr)) {
 		srp_dev->global_mr = ib_get_dma_mr(srp_dev->pd,
 						   IB_ACCESS_LOCAL_WRITE |
 						   IB_ACCESS_REMOTE_READ |

drivers/infiniband/ulp/srp/ib_srp.h

@@ -202,6 +202,8 @@ struct srp_target_port {
 	char			target_name[32];
 	unsigned int		scsi_id;
 	unsigned int		sg_tablesize;
+	int			mr_pool_size;
+	int			mr_per_cmd;
 	int			queue_size;
 	int			req_ring_size;
 	int			comp_vector;

drivers/infiniband/ulp/srpt/ib_srpt.c

@@ -764,52 +764,6 @@ static int srpt_post_recv(struct srpt_device *sdev,
 	return ib_post_srq_recv(sdev->srq, &wr, &bad_wr);
 }
 
-/**
- * srpt_post_send() - Post an IB send request.
- *
- * Returns zero upon success and a non-zero value upon failure.
- */
-static int srpt_post_send(struct srpt_rdma_ch *ch,
-			  struct srpt_send_ioctx *ioctx, int len)
-{
-	struct ib_sge list;
-	struct ib_send_wr wr, *bad_wr;
-	struct srpt_device *sdev = ch->sport->sdev;
-	int ret;
-
-	atomic_inc(&ch->req_lim);
-
-	ret = -ENOMEM;
-	if (unlikely(atomic_dec_return(&ch->sq_wr_avail) < 0)) {
-		pr_warn("IB send queue full (needed 1)\n");
-		goto out;
-	}
-
-	ib_dma_sync_single_for_device(sdev->device, ioctx->ioctx.dma, len,
-				      DMA_TO_DEVICE);
-
-	list.addr = ioctx->ioctx.dma;
-	list.length = len;
-	list.lkey = sdev->pd->local_dma_lkey;
-
-	ioctx->ioctx.cqe.done = srpt_send_done;
-	wr.next = NULL;
-	wr.wr_cqe = &ioctx->ioctx.cqe;
-	wr.sg_list = &list;
-	wr.num_sge = 1;
-	wr.opcode = IB_WR_SEND;
-	wr.send_flags = IB_SEND_SIGNALED;
-
-	ret = ib_post_send(ch->qp, &wr, &bad_wr);
-
-out:
-	if (ret < 0) {
-		atomic_inc(&ch->sq_wr_avail);
-		atomic_dec(&ch->req_lim);
-	}
-	return ret;
-}
-
 /**
  * srpt_zerolength_write() - Perform a zero-length RDMA write.
  *
@@ -843,6 +797,110 @@ static void srpt_zerolength_write_done(struct ib_cq *cq, struct ib_wc *wc)
 	}
 }
 
+static int srpt_alloc_rw_ctxs(struct srpt_send_ioctx *ioctx,
+		struct srp_direct_buf *db, int nbufs, struct scatterlist **sg,
+		unsigned *sg_cnt)
+{
+	enum dma_data_direction dir = target_reverse_dma_direction(&ioctx->cmd);
+	struct srpt_rdma_ch *ch = ioctx->ch;
+	struct scatterlist *prev = NULL;
+	unsigned prev_nents;
+	int ret, i;
+
+	if (nbufs == 1) {
+		ioctx->rw_ctxs = &ioctx->s_rw_ctx;
+	} else {
+		ioctx->rw_ctxs = kmalloc_array(nbufs, sizeof(*ioctx->rw_ctxs),
+			GFP_KERNEL);
+		if (!ioctx->rw_ctxs)
+			return -ENOMEM;
+	}
+
+	for (i = ioctx->n_rw_ctx; i < nbufs; i++, db++) {
+		struct srpt_rw_ctx *ctx = &ioctx->rw_ctxs[i];
+		u64 remote_addr = be64_to_cpu(db->va);
+		u32 size = be32_to_cpu(db->len);
+		u32 rkey = be32_to_cpu(db->key);
+
+		ret = target_alloc_sgl(&ctx->sg, &ctx->nents, size, false,
+				i < nbufs - 1);
+		if (ret)
+			goto unwind;
+
+		ret = rdma_rw_ctx_init(&ctx->rw, ch->qp, ch->sport->port,
+				ctx->sg, ctx->nents, 0, remote_addr, rkey, dir);
+		if (ret < 0) {
+			target_free_sgl(ctx->sg, ctx->nents);
+			goto unwind;
+		}
+
+		ioctx->n_rdma += ret;
+		ioctx->n_rw_ctx++;
+
+		if (prev) {
+			sg_unmark_end(&prev[prev_nents - 1]);
+			sg_chain(prev, prev_nents + 1, ctx->sg);
+		} else {
+			*sg = ctx->sg;
+		}
+
+		prev = ctx->sg;
+		prev_nents = ctx->nents;
+
+		*sg_cnt += ctx->nents;
+	}
+
+	return 0;
+
+unwind:
+	while (--i >= 0) {
+		struct srpt_rw_ctx *ctx = &ioctx->rw_ctxs[i];
+
+		rdma_rw_ctx_destroy(&ctx->rw, ch->qp, ch->sport->port,
+				ctx->sg, ctx->nents, dir);
+		target_free_sgl(ctx->sg, ctx->nents);
+	}
+	if (ioctx->rw_ctxs != &ioctx->s_rw_ctx)
+		kfree(ioctx->rw_ctxs);
+	return ret;
+}
+
+static void srpt_free_rw_ctxs(struct srpt_rdma_ch *ch,
+				    struct srpt_send_ioctx *ioctx)
+{
+	enum dma_data_direction dir = target_reverse_dma_direction(&ioctx->cmd);
+	int i;
+
+	for (i = 0; i < ioctx->n_rw_ctx; i++) {
+		struct srpt_rw_ctx *ctx = &ioctx->rw_ctxs[i];
+
+		rdma_rw_ctx_destroy(&ctx->rw, ch->qp, ch->sport->port,
+				ctx->sg, ctx->nents, dir);
+		target_free_sgl(ctx->sg, ctx->nents);
+	}
+
+	if (ioctx->rw_ctxs != &ioctx->s_rw_ctx)
+		kfree(ioctx->rw_ctxs);
+}
+
+static inline void *srpt_get_desc_buf(struct srp_cmd *srp_cmd)
+{
+	/*
+	 * The pointer computations below will only be compiled correctly
+	 * if srp_cmd::add_data is declared as s8*, u8*, s8[] or u8[], so check
+	 * whether srp_cmd::add_data has been declared as a byte pointer.
+	 */
+	BUILD_BUG_ON(!__same_type(srp_cmd->add_data[0], (s8)0) &&
+		     !__same_type(srp_cmd->add_data[0], (u8)0));
+
+	/*
+	 * According to the SRP spec, the lower two bits of the 'ADDITIONAL
+	 * CDB LENGTH' field are reserved and the size in bytes of this field
+	 * is four times the value specified in bits 3..7. Hence the "& ~3".
+	 */
+	return srp_cmd->add_data + (srp_cmd->add_cdb_len & ~3);
+}
+
 /**
  * srpt_get_desc_tbl() - Parse the data descriptors of an SRP_CMD request.
  * @ioctx: Pointer to the I/O context associated with the request.
@@ -858,94 +916,59 @@ static void srpt_zerolength_write_done(struct ib_cq *cq, struct ib_wc *wc)
  * -ENOMEM when memory allocation fails and zero upon success.
  */
 static int srpt_get_desc_tbl(struct srpt_send_ioctx *ioctx,
-			     struct srp_cmd *srp_cmd,
-			     enum dma_data_direction *dir, u64 *data_len)
+		struct srp_cmd *srp_cmd, enum dma_data_direction *dir,
+		struct scatterlist **sg, unsigned *sg_cnt, u64 *data_len)
 {
-	struct srp_indirect_buf *idb;
-	struct srp_direct_buf *db;
-	unsigned add_cdb_offset;
-	int ret;
-
-	/*
-	 * The pointer computations below will only be compiled correctly
-	 * if srp_cmd::add_data is declared as s8*, u8*, s8[] or u8[], so check
-	 * whether srp_cmd::add_data has been declared as a byte pointer.
-	 */
-	BUILD_BUG_ON(!__same_type(srp_cmd->add_data[0], (s8)0)
-		     && !__same_type(srp_cmd->add_data[0], (u8)0));
-
 	BUG_ON(!dir);
 	BUG_ON(!data_len);
 
-	ret = 0;
-	*data_len = 0;
-
 	/*
 	 * The lower four bits of the buffer format field contain the DATA-IN
 	 * buffer descriptor format, and the highest four bits contain the
 	 * DATA-OUT buffer descriptor format.
 	 */
-	*dir = DMA_NONE;
 	if (srp_cmd->buf_fmt & 0xf)
 		/* DATA-IN: transfer data from target to initiator (read). */
 		*dir = DMA_FROM_DEVICE;
 	else if (srp_cmd->buf_fmt >> 4)
 		/* DATA-OUT: transfer data from initiator to target (write). */
 		*dir = DMA_TO_DEVICE;
+	else
+		*dir = DMA_NONE;
+
+	/* initialize data_direction early as srpt_alloc_rw_ctxs needs it */
+	ioctx->cmd.data_direction = *dir;
 
-	/*
-	 * According to the SRP spec, the lower two bits of the 'ADDITIONAL
-	 * CDB LENGTH' field are reserved and the size in bytes of this field
-	 * is four times the value specified in bits 3..7. Hence the "& ~3".
-	 */
-	add_cdb_offset = srp_cmd->add_cdb_len & ~3;
 	if (((srp_cmd->buf_fmt & 0xf) == SRP_DATA_DESC_DIRECT) ||
 	    ((srp_cmd->buf_fmt >> 4) == SRP_DATA_DESC_DIRECT)) {
-		ioctx->n_rbuf = 1;
-		ioctx->rbufs = &ioctx->single_rbuf;
+	    	struct srp_direct_buf *db = srpt_get_desc_buf(srp_cmd);
 
-		db = (struct srp_direct_buf *)(srp_cmd->add_data
-					       + add_cdb_offset);
-		memcpy(ioctx->rbufs, db, sizeof(*db));
 		*data_len = be32_to_cpu(db->len);
+		return srpt_alloc_rw_ctxs(ioctx, db, 1, sg, sg_cnt);
 	} else if (((srp_cmd->buf_fmt & 0xf) == SRP_DATA_DESC_INDIRECT) ||
 		   ((srp_cmd->buf_fmt >> 4) == SRP_DATA_DESC_INDIRECT)) {
-		idb = (struct srp_indirect_buf *)(srp_cmd->add_data
-						  + add_cdb_offset);
+		struct srp_indirect_buf *idb = srpt_get_desc_buf(srp_cmd);
+		int nbufs = be32_to_cpu(idb->table_desc.len) /
+				sizeof(struct srp_direct_buf);
 
-		ioctx->n_rbuf = be32_to_cpu(idb->table_desc.len) / sizeof(*db);
-
-		if (ioctx->n_rbuf >
+		if (nbufs >
 		    (srp_cmd->data_out_desc_cnt + srp_cmd->data_in_desc_cnt)) {
 			pr_err("received unsupported SRP_CMD request"
 			       " type (%u out + %u in != %u / %zu)\n",
 			       srp_cmd->data_out_desc_cnt,
 			       srp_cmd->data_in_desc_cnt,
 			       be32_to_cpu(idb->table_desc.len),
-			       sizeof(*db));
-			ioctx->n_rbuf = 0;
-			ret = -EINVAL;
-			goto out;
+			       sizeof(struct srp_direct_buf));
+			return -EINVAL;
 		}
 
-		if (ioctx->n_rbuf == 1)
-			ioctx->rbufs = &ioctx->single_rbuf;
-		else {
-			ioctx->rbufs =
-				kmalloc(ioctx->n_rbuf * sizeof(*db), GFP_ATOMIC);
-			if (!ioctx->rbufs) {
-				ioctx->n_rbuf = 0;
-				ret = -ENOMEM;
-				goto out;
-			}
-		}
-
-		db = idb->desc_list;
-		memcpy(ioctx->rbufs, db, ioctx->n_rbuf * sizeof(*db));
 		*data_len = be32_to_cpu(idb->len);
+		return srpt_alloc_rw_ctxs(ioctx, idb->desc_list, nbufs,
+				sg, sg_cnt);
+	} else {
+		*data_len = 0;
+		return 0;
 	}
-out:
-	return ret;
 }
 
 /**
@@ -1048,217 +1071,6 @@ static int srpt_ch_qp_err(struct srpt_rdma_ch *ch)
 	return ib_modify_qp(ch->qp, &qp_attr, IB_QP_STATE);
 }
 
-/**
- * srpt_unmap_sg_to_ib_sge() - Unmap an IB SGE list.
- */
-static void srpt_unmap_sg_to_ib_sge(struct srpt_rdma_ch *ch,
-				    struct srpt_send_ioctx *ioctx)
-{
-	struct scatterlist *sg;
-	enum dma_data_direction dir;
-
-	BUG_ON(!ch);
-	BUG_ON(!ioctx);
-	BUG_ON(ioctx->n_rdma && !ioctx->rdma_wrs);
-
-	while (ioctx->n_rdma)
-		kfree(ioctx->rdma_wrs[--ioctx->n_rdma].wr.sg_list);
-
-	kfree(ioctx->rdma_wrs);
-	ioctx->rdma_wrs = NULL;
-
-	if (ioctx->mapped_sg_count) {
-		sg = ioctx->sg;
-		WARN_ON(!sg);
-		dir = ioctx->cmd.data_direction;
-		BUG_ON(dir == DMA_NONE);
-		ib_dma_unmap_sg(ch->sport->sdev->device, sg, ioctx->sg_cnt,
-				target_reverse_dma_direction(&ioctx->cmd));
-		ioctx->mapped_sg_count = 0;
-	}
-}
-
-/**
- * srpt_map_sg_to_ib_sge() - Map an SG list to an IB SGE list.
- */
-static int srpt_map_sg_to_ib_sge(struct srpt_rdma_ch *ch,
-				 struct srpt_send_ioctx *ioctx)
-{
-	struct ib_device *dev = ch->sport->sdev->device;
-	struct se_cmd *cmd;
-	struct scatterlist *sg, *sg_orig;
-	int sg_cnt;
-	enum dma_data_direction dir;
-	struct ib_rdma_wr *riu;
-	struct srp_direct_buf *db;
-	dma_addr_t dma_addr;
-	struct ib_sge *sge;
-	u64 raddr;
-	u32 rsize;
-	u32 tsize;
-	u32 dma_len;
-	int count, nrdma;
-	int i, j, k;
-
-	BUG_ON(!ch);
-	BUG_ON(!ioctx);
-	cmd = &ioctx->cmd;
-	dir = cmd->data_direction;
-	BUG_ON(dir == DMA_NONE);
-
-	ioctx->sg = sg = sg_orig = cmd->t_data_sg;
-	ioctx->sg_cnt = sg_cnt = cmd->t_data_nents;
-
-	count = ib_dma_map_sg(ch->sport->sdev->device, sg, sg_cnt,
-			      target_reverse_dma_direction(cmd));
-	if (unlikely(!count))
-		return -EAGAIN;
-
-	ioctx->mapped_sg_count = count;
-
-	if (ioctx->rdma_wrs && ioctx->n_rdma_wrs)
-		nrdma = ioctx->n_rdma_wrs;
-	else {
-		nrdma = (count + SRPT_DEF_SG_PER_WQE - 1) / SRPT_DEF_SG_PER_WQE
-			+ ioctx->n_rbuf;
-
-		ioctx->rdma_wrs = kcalloc(nrdma, sizeof(*ioctx->rdma_wrs),
-				GFP_KERNEL);
-		if (!ioctx->rdma_wrs)
-			goto free_mem;
-
-		ioctx->n_rdma_wrs = nrdma;
-	}
-
-	db = ioctx->rbufs;
-	tsize = cmd->data_length;
-	dma_len = ib_sg_dma_len(dev, &sg[0]);
-	riu = ioctx->rdma_wrs;
-
-	/*
-	 * For each remote desc - calculate the #ib_sge.
-	 * If #ib_sge < SRPT_DEF_SG_PER_WQE per rdma operation then
-	 *      each remote desc rdma_iu is required a rdma wr;
-	 * else
-	 *      we need to allocate extra rdma_iu to carry extra #ib_sge in
-	 *      another rdma wr
-	 */
-	for (i = 0, j = 0;
-	     j < count && i < ioctx->n_rbuf && tsize > 0; ++i, ++riu, ++db) {
-		rsize = be32_to_cpu(db->len);
-		raddr = be64_to_cpu(db->va);
-		riu->remote_addr = raddr;
-		riu->rkey = be32_to_cpu(db->key);
-		riu->wr.num_sge = 0;
-
-		/* calculate how many sge required for this remote_buf */
-		while (rsize > 0 && tsize > 0) {
-
-			if (rsize >= dma_len) {
-				tsize -= dma_len;
-				rsize -= dma_len;
-				raddr += dma_len;
-
-				if (tsize > 0) {
-					++j;
-					if (j < count) {
-						sg = sg_next(sg);
-						dma_len = ib_sg_dma_len(
-								dev, sg);
-					}
-				}
-			} else {
-				tsize -= rsize;
-				dma_len -= rsize;
-				rsize = 0;
-			}
-
-			++riu->wr.num_sge;
-
-			if (rsize > 0 &&
-			    riu->wr.num_sge == SRPT_DEF_SG_PER_WQE) {
-				++ioctx->n_rdma;
-				riu->wr.sg_list = kmalloc_array(riu->wr.num_sge,
-						sizeof(*riu->wr.sg_list),
-						GFP_KERNEL);
-				if (!riu->wr.sg_list)
-					goto free_mem;
-
-				++riu;
-				riu->wr.num_sge = 0;
-				riu->remote_addr = raddr;
-				riu->rkey = be32_to_cpu(db->key);
-			}
-		}
-
-		++ioctx->n_rdma;
-		riu->wr.sg_list = kmalloc_array(riu->wr.num_sge,
-					sizeof(*riu->wr.sg_list),
-					GFP_KERNEL);
-		if (!riu->wr.sg_list)
-			goto free_mem;
-	}
-
-	db = ioctx->rbufs;
-	tsize = cmd->data_length;
-	riu = ioctx->rdma_wrs;
-	sg = sg_orig;
-	dma_len = ib_sg_dma_len(dev, &sg[0]);
-	dma_addr = ib_sg_dma_address(dev, &sg[0]);
-
-	/* this second loop is really mapped sg_addres to rdma_iu->ib_sge */
-	for (i = 0, j = 0;
-	     j < count && i < ioctx->n_rbuf && tsize > 0; ++i, ++riu, ++db) {
-		rsize = be32_to_cpu(db->len);
-		sge = riu->wr.sg_list;
-		k = 0;
-
-		while (rsize > 0 && tsize > 0) {
-			sge->addr = dma_addr;
-			sge->lkey = ch->sport->sdev->pd->local_dma_lkey;
-
-			if (rsize >= dma_len) {
-				sge->length =
-					(tsize < dma_len) ? tsize : dma_len;
-				tsize -= dma_len;
-				rsize -= dma_len;
-
-				if (tsize > 0) {
-					++j;
-					if (j < count) {
-						sg = sg_next(sg);
-						dma_len = ib_sg_dma_len(
-								dev, sg);
-						dma_addr = ib_sg_dma_address(
-								dev, sg);
-					}
-				}
-			} else {
-				sge->length = (tsize < rsize) ? tsize : rsize;
-				tsize -= rsize;
-				dma_len -= rsize;
-				dma_addr += rsize;
-				rsize = 0;
-			}
-
-			++k;
-			if (k == riu->wr.num_sge && rsize > 0 && tsize > 0) {
-				++riu;
-				sge = riu->wr.sg_list;
-				k = 0;
-			} else if (rsize > 0 && tsize > 0)
-				++sge;
-		}
-	}
-
-	return 0;
-
-free_mem:
-	srpt_unmap_sg_to_ib_sge(ch, ioctx);
-
-	return -ENOMEM;
-}
-
 /**
  * srpt_get_send_ioctx() - Obtain an I/O context for sending to the initiator.
  */
@@ -1284,12 +1096,8 @@ static struct srpt_send_ioctx *srpt_get_send_ioctx(struct srpt_rdma_ch *ch)
 	BUG_ON(ioctx->ch != ch);
 	spin_lock_init(&ioctx->spinlock);
 	ioctx->state = SRPT_STATE_NEW;
-	ioctx->n_rbuf = 0;
-	ioctx->rbufs = NULL;
 	ioctx->n_rdma = 0;
-	ioctx->n_rdma_wrs = 0;
-	ioctx->rdma_wrs = NULL;
-	ioctx->mapped_sg_count = 0;
+	ioctx->n_rw_ctx = 0;
 	init_completion(&ioctx->tx_done);
 	ioctx->queue_status_only = false;
 	/*
@@ -1359,7 +1167,6 @@ static int srpt_abort_cmd(struct srpt_send_ioctx *ioctx)
 		 * SRP_RSP sending failed or the SRP_RSP send completion has
 		 * not been received in time.
 		 */
-		srpt_unmap_sg_to_ib_sge(ioctx->ch, ioctx);
 		transport_generic_free_cmd(&ioctx->cmd, 0);
 		break;
 	case SRPT_STATE_MGMT_RSP_SENT:
@@ -1387,6 +1194,7 @@ static void srpt_rdma_read_done(struct ib_cq *cq, struct ib_wc *wc)
 
 	WARN_ON(ioctx->n_rdma <= 0);
 	atomic_add(ioctx->n_rdma, &ch->sq_wr_avail);
+	ioctx->n_rdma = 0;
 
 	if (unlikely(wc->status != IB_WC_SUCCESS)) {
 		pr_info("RDMA_READ for ioctx 0x%p failed with status %d\n",
@@ -1403,23 +1211,6 @@ static void srpt_rdma_read_done(struct ib_cq *cq, struct ib_wc *wc)
 		       __LINE__, srpt_get_cmd_state(ioctx));
 }
 
-static void srpt_rdma_write_done(struct ib_cq *cq, struct ib_wc *wc)
-{
-	struct srpt_send_ioctx *ioctx =
-		container_of(wc->wr_cqe, struct srpt_send_ioctx, rdma_cqe);
-
-	if (unlikely(wc->status != IB_WC_SUCCESS)) {
-		/*
-		 * Note: if an RDMA write error completion is received that
-		 * means that a SEND also has been posted. Defer further
-		 * processing of the associated command until the send error
-		 * completion has been received.
-		 */
-		pr_info("RDMA_WRITE for ioctx 0x%p failed with status %d\n",
-			ioctx, wc->status);
-	}
-}
-
 /**
  * srpt_build_cmd_rsp() - Build an SRP_RSP response.
  * @ch: RDMA channel through which the request has been received.
@@ -1537,6 +1328,8 @@ static void srpt_handle_cmd(struct srpt_rdma_ch *ch,
 {
 	struct se_cmd *cmd;
 	struct srp_cmd *srp_cmd;
+	struct scatterlist *sg = NULL;
+	unsigned sg_cnt = 0;
 	u64 data_len;
 	enum dma_data_direction dir;
 	int rc;
@@ -1563,16 +1356,21 @@ static void srpt_handle_cmd(struct srpt_rdma_ch *ch,
 		break;
 	}
 
-	if (srpt_get_desc_tbl(send_ioctx, srp_cmd, &dir, &data_len)) {
-		pr_err("0x%llx: parsing SRP descriptor table failed.\n",
-		       srp_cmd->tag);
+	rc = srpt_get_desc_tbl(send_ioctx, srp_cmd, &dir, &sg, &sg_cnt,
+			&data_len);
+	if (rc) {
+		if (rc != -EAGAIN) {
+			pr_err("0x%llx: parsing SRP descriptor table failed.\n",
+			       srp_cmd->tag);
+		}
 		goto release_ioctx;
 	}
 
-	rc = target_submit_cmd(cmd, ch->sess, srp_cmd->cdb,
+	rc = target_submit_cmd_map_sgls(cmd, ch->sess, srp_cmd->cdb,
 			       &send_ioctx->sense_data[0],
 			       scsilun_to_int(&srp_cmd->lun), data_len,
-			       TCM_SIMPLE_TAG, dir, TARGET_SCF_ACK_KREF);
+			       TCM_SIMPLE_TAG, dir, TARGET_SCF_ACK_KREF,
+			       sg, sg_cnt, NULL, 0, NULL, 0);
 	if (rc != 0) {
 		pr_debug("target_submit_cmd() returned %d for tag %#llx\n", rc,
 			 srp_cmd->tag);
@@ -1664,23 +1462,21 @@ static void srpt_handle_new_iu(struct srpt_rdma_ch *ch,
 				   recv_ioctx->ioctx.dma, srp_max_req_size,
 				   DMA_FROM_DEVICE);
 
-	if (unlikely(ch->state == CH_CONNECTING)) {
-		list_add_tail(&recv_ioctx->wait_list, &ch->cmd_wait_list);
-		goto out;
-	}
+	if (unlikely(ch->state == CH_CONNECTING))
+		goto out_wait;
 
 	if (unlikely(ch->state != CH_LIVE))
-		goto out;
+		return;
 
 	srp_cmd = recv_ioctx->ioctx.buf;
 	if (srp_cmd->opcode == SRP_CMD || srp_cmd->opcode == SRP_TSK_MGMT) {
-		if (!send_ioctx)
+		if (!send_ioctx) {
+			if (!list_empty(&ch->cmd_wait_list))
+				goto out_wait;
 			send_ioctx = srpt_get_send_ioctx(ch);
-		if (unlikely(!send_ioctx)) {
-			list_add_tail(&recv_ioctx->wait_list,
-				      &ch->cmd_wait_list);
-			goto out;
 		}
+		if (unlikely(!send_ioctx))
+			goto out_wait;
 	}
 
 	switch (srp_cmd->opcode) {
@@ -1709,8 +1505,10 @@ static void srpt_handle_new_iu(struct srpt_rdma_ch *ch,
 	}
 
 	srpt_post_recv(ch->sport->sdev, recv_ioctx);
-out:
 	return;
+
+out_wait:
+	list_add_tail(&recv_ioctx->wait_list, &ch->cmd_wait_list);
 }
 
 static void srpt_recv_done(struct ib_cq *cq, struct ib_wc *wc)
@@ -1779,14 +1577,13 @@ static void srpt_send_done(struct ib_cq *cq, struct ib_wc *wc)
 	WARN_ON(state != SRPT_STATE_CMD_RSP_SENT &&
 		state != SRPT_STATE_MGMT_RSP_SENT);
 
-	atomic_inc(&ch->sq_wr_avail);
+	atomic_add(1 + ioctx->n_rdma, &ch->sq_wr_avail);
 
 	if (wc->status != IB_WC_SUCCESS)
 		pr_info("sending response for ioctx 0x%p failed"
 			" with status %d\n", ioctx, wc->status);
 
 	if (state != SRPT_STATE_DONE) {
-		srpt_unmap_sg_to_ib_sge(ch, ioctx);
 		transport_generic_free_cmd(&ioctx->cmd, 0);
 	} else {
 		pr_err("IB completion has been received too late for"
@@ -1832,8 +1629,18 @@ static int srpt_create_ch_ib(struct srpt_rdma_ch *ch)
 	qp_init->srq = sdev->srq;
 	qp_init->sq_sig_type = IB_SIGNAL_REQ_WR;
 	qp_init->qp_type = IB_QPT_RC;
-	qp_init->cap.max_send_wr = srp_sq_size;
-	qp_init->cap.max_send_sge = SRPT_DEF_SG_PER_WQE;
+	/*
+	 * We divide up our send queue size into half SEND WRs to send the
+	 * completions, and half R/W contexts to actually do the RDMA
+	 * READ/WRITE transfers.  Note that we need to allocate CQ slots for
+	 * both both, as RDMA contexts will also post completions for the
+	 * RDMA READ case.
+	 */
+	qp_init->cap.max_send_wr = srp_sq_size / 2;
+	qp_init->cap.max_rdma_ctxs = srp_sq_size / 2;
+	qp_init->cap.max_send_sge = max(sdev->device->attrs.max_sge_rd,
+					sdev->device->attrs.max_sge);
+	qp_init->port_num = ch->sport->port;
 
 	ch->qp = ib_create_qp(sdev->pd, qp_init);
 	if (IS_ERR(ch->qp)) {
@@ -2386,95 +2193,6 @@ static int srpt_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event)
 	return ret;
 }
 
-/**
- * srpt_perform_rdmas() - Perform IB RDMA.
- *
- * Returns zero upon success or a negative number upon failure.
- */
-static int srpt_perform_rdmas(struct srpt_rdma_ch *ch,
-			      struct srpt_send_ioctx *ioctx)
-{
-	struct ib_send_wr *bad_wr;
-	int sq_wr_avail, ret, i;
-	enum dma_data_direction dir;
-	const int n_rdma = ioctx->n_rdma;
-
-	dir = ioctx->cmd.data_direction;
-	if (dir == DMA_TO_DEVICE) {
-		/* write */
-		ret = -ENOMEM;
-		sq_wr_avail = atomic_sub_return(n_rdma, &ch->sq_wr_avail);
-		if (sq_wr_avail < 0) {
-			pr_warn("IB send queue full (needed %d)\n",
-				n_rdma);
-			goto out;
-		}
-	}
-
-	for (i = 0; i < n_rdma; i++) {
-		struct ib_send_wr *wr = &ioctx->rdma_wrs[i].wr;
-
-		wr->opcode = (dir == DMA_FROM_DEVICE) ?
-				IB_WR_RDMA_WRITE : IB_WR_RDMA_READ;
-
-		if (i == n_rdma - 1) {
-			/* only get completion event for the last rdma read */
-			if (dir == DMA_TO_DEVICE) {
-				wr->send_flags = IB_SEND_SIGNALED;
-				ioctx->rdma_cqe.done = srpt_rdma_read_done;
-			} else {
-				ioctx->rdma_cqe.done = srpt_rdma_write_done;
-			}
-			wr->wr_cqe = &ioctx->rdma_cqe;
-			wr->next = NULL;
-		} else {
-			wr->wr_cqe = NULL;
-			wr->next = &ioctx->rdma_wrs[i + 1].wr;
-		}
-	}
-
-	ret = ib_post_send(ch->qp, &ioctx->rdma_wrs->wr, &bad_wr);
-	if (ret)
-		pr_err("%s[%d]: ib_post_send() returned %d for %d/%d\n",
-				 __func__, __LINE__, ret, i, n_rdma);
-out:
-	if (unlikely(dir == DMA_TO_DEVICE && ret < 0))
-		atomic_add(n_rdma, &ch->sq_wr_avail);
-	return ret;
-}
-
-/**
- * srpt_xfer_data() - Start data transfer from initiator to target.
- */
-static int srpt_xfer_data(struct srpt_rdma_ch *ch,
-			  struct srpt_send_ioctx *ioctx)
-{
-	int ret;
-
-	ret = srpt_map_sg_to_ib_sge(ch, ioctx);
-	if (ret) {
-		pr_err("%s[%d] ret=%d\n", __func__, __LINE__, ret);
-		goto out;
-	}
-
-	ret = srpt_perform_rdmas(ch, ioctx);
-	if (ret) {
-		if (ret == -EAGAIN || ret == -ENOMEM)
-			pr_info("%s[%d] queue full -- ret=%d\n",
-				__func__, __LINE__, ret);
-		else
-			pr_err("%s[%d] fatal error -- ret=%d\n",
-			       __func__, __LINE__, ret);
-		goto out_unmap;
-	}
-
-out:
-	return ret;
-out_unmap:
-	srpt_unmap_sg_to_ib_sge(ch, ioctx);
-	goto out;
-}
-
 static int srpt_write_pending_status(struct se_cmd *se_cmd)
 {
 	struct srpt_send_ioctx *ioctx;
@@ -2491,11 +2209,42 @@ static int srpt_write_pending(struct se_cmd *se_cmd)
 	struct srpt_send_ioctx *ioctx =
 		container_of(se_cmd, struct srpt_send_ioctx, cmd);
 	struct srpt_rdma_ch *ch = ioctx->ch;
+	struct ib_send_wr *first_wr = NULL, *bad_wr;
+	struct ib_cqe *cqe = &ioctx->rdma_cqe;
 	enum srpt_command_state new_state;
+	int ret, i;
 
 	new_state = srpt_set_cmd_state(ioctx, SRPT_STATE_NEED_DATA);
 	WARN_ON(new_state == SRPT_STATE_DONE);
-	return srpt_xfer_data(ch, ioctx);
+
+	if (atomic_sub_return(ioctx->n_rdma, &ch->sq_wr_avail) < 0) {
+		pr_warn("%s: IB send queue full (needed %d)\n",
+				__func__, ioctx->n_rdma);
+		ret = -ENOMEM;
+		goto out_undo;
+	}
+
+	cqe->done = srpt_rdma_read_done;
+	for (i = ioctx->n_rw_ctx - 1; i >= 0; i--) {
+		struct srpt_rw_ctx *ctx = &ioctx->rw_ctxs[i];
+
+		first_wr = rdma_rw_ctx_wrs(&ctx->rw, ch->qp, ch->sport->port,
+				cqe, first_wr);
+		cqe = NULL;
+	}
+	
+	ret = ib_post_send(ch->qp, first_wr, &bad_wr);
+	if (ret) {
+		pr_err("%s: ib_post_send() returned %d for %d (avail: %d)\n",
+			 __func__, ret, ioctx->n_rdma,
+			 atomic_read(&ch->sq_wr_avail));
+		goto out_undo;
+	}
+
+	return 0;
+out_undo:
+	atomic_add(ioctx->n_rdma, &ch->sq_wr_avail);
+	return ret;
 }
 
 static u8 tcm_to_srp_tsk_mgmt_status(const int tcm_mgmt_status)
@@ -2517,17 +2266,17 @@ static u8 tcm_to_srp_tsk_mgmt_status(const int tcm_mgmt_status)
  */
 static void srpt_queue_response(struct se_cmd *cmd)
 {
-	struct srpt_rdma_ch *ch;
-	struct srpt_send_ioctx *ioctx;
+	struct srpt_send_ioctx *ioctx =
+		container_of(cmd, struct srpt_send_ioctx, cmd);
+	struct srpt_rdma_ch *ch = ioctx->ch;
+	struct srpt_device *sdev = ch->sport->sdev;
+	struct ib_send_wr send_wr, *first_wr = NULL, *bad_wr;
+	struct ib_sge sge;
 	enum srpt_command_state state;
 	unsigned long flags;
-	int ret;
-	enum dma_data_direction dir;
-	int resp_len;
+	int resp_len, ret, i;
 	u8 srp_tm_status;
 
-	ioctx = container_of(cmd, struct srpt_send_ioctx, cmd);
-	ch = ioctx->ch;
 	BUG_ON(!ch);
 
 	spin_lock_irqsave(&ioctx->spinlock, flags);
@@ -2554,17 +2303,19 @@ static void srpt_queue_response(struct se_cmd *cmd)
 		return;
 	}
 
-	dir = ioctx->cmd.data_direction;
-
 	/* For read commands, transfer the data to the initiator. */
-	if (dir == DMA_FROM_DEVICE && ioctx->cmd.data_length &&
+	if (ioctx->cmd.data_direction == DMA_FROM_DEVICE &&
+	    ioctx->cmd.data_length &&
 	    !ioctx->queue_status_only) {
-		ret = srpt_xfer_data(ch, ioctx);
-		if (ret) {
-			pr_err("xfer_data failed for tag %llu\n",
-			       ioctx->cmd.tag);
-			return;
+		for (i = ioctx->n_rw_ctx - 1; i >= 0; i--) {
+			struct srpt_rw_ctx *ctx = &ioctx->rw_ctxs[i];
+
+			first_wr = rdma_rw_ctx_wrs(&ctx->rw, ch->qp,
+					ch->sport->port, NULL,
+					first_wr ? first_wr : &send_wr);
 		}
+	} else {
+		first_wr = &send_wr;
 	}
 
 	if (state != SRPT_STATE_MGMT)
@@ -2576,14 +2327,46 @@ static void srpt_queue_response(struct se_cmd *cmd)
 		resp_len = srpt_build_tskmgmt_rsp(ch, ioctx, srp_tm_status,
 						 ioctx->cmd.tag);
 	}
-	ret = srpt_post_send(ch, ioctx, resp_len);
-	if (ret) {
-		pr_err("sending cmd response failed for tag %llu\n",
-		       ioctx->cmd.tag);
-		srpt_unmap_sg_to_ib_sge(ch, ioctx);
-		srpt_set_cmd_state(ioctx, SRPT_STATE_DONE);
-		target_put_sess_cmd(&ioctx->cmd);
+
+	atomic_inc(&ch->req_lim);
+
+	if (unlikely(atomic_sub_return(1 + ioctx->n_rdma,
+			&ch->sq_wr_avail) < 0)) {
+		pr_warn("%s: IB send queue full (needed %d)\n",
+				__func__, ioctx->n_rdma);
+		ret = -ENOMEM;
+		goto out;
 	}
+
+	ib_dma_sync_single_for_device(sdev->device, ioctx->ioctx.dma, resp_len,
+				      DMA_TO_DEVICE);
+
+	sge.addr = ioctx->ioctx.dma;
+	sge.length = resp_len;
+	sge.lkey = sdev->pd->local_dma_lkey;
+
+	ioctx->ioctx.cqe.done = srpt_send_done;
+	send_wr.next = NULL;
+	send_wr.wr_cqe = &ioctx->ioctx.cqe;
+	send_wr.sg_list = &sge;
+	send_wr.num_sge = 1;
+	send_wr.opcode = IB_WR_SEND;
+	send_wr.send_flags = IB_SEND_SIGNALED;
+
+	ret = ib_post_send(ch->qp, first_wr, &bad_wr);
+	if (ret < 0) {
+		pr_err("%s: sending cmd response failed for tag %llu (%d)\n",
+			__func__, ioctx->cmd.tag, ret);
+		goto out;
+	}
+
+	return;
+
+out:
+	atomic_add(1 + ioctx->n_rdma, &ch->sq_wr_avail);
+	atomic_dec(&ch->req_lim);
+	srpt_set_cmd_state(ioctx, SRPT_STATE_DONE);
+	target_put_sess_cmd(&ioctx->cmd);
 }
 
 static int srpt_queue_data_in(struct se_cmd *cmd)
@@ -2599,10 +2382,6 @@ static void srpt_queue_tm_rsp(struct se_cmd *cmd)
 
 static void srpt_aborted_task(struct se_cmd *cmd)
 {
-	struct srpt_send_ioctx *ioctx = container_of(cmd,
-				struct srpt_send_ioctx, cmd);
-
-	srpt_unmap_sg_to_ib_sge(ioctx->ch, ioctx);
 }
 
 static int srpt_queue_status(struct se_cmd *cmd)
@@ -2903,12 +2682,10 @@ static void srpt_release_cmd(struct se_cmd *se_cmd)
 	unsigned long flags;
 
 	WARN_ON(ioctx->state != SRPT_STATE_DONE);
-	WARN_ON(ioctx->mapped_sg_count != 0);
 
-	if (ioctx->n_rbuf > 1) {
-		kfree(ioctx->rbufs);
-		ioctx->rbufs = NULL;
-		ioctx->n_rbuf = 0;
+	if (ioctx->n_rw_ctx) {
+		srpt_free_rw_ctxs(ch, ioctx);
+		ioctx->n_rw_ctx = 0;
 	}
 
 	spin_lock_irqsave(&ch->spinlock, flags);

drivers/infiniband/ulp/srpt/ib_srpt.h

@@ -42,6 +42,7 @@
 #include <rdma/ib_verbs.h>
 #include <rdma/ib_sa.h>
 #include <rdma/ib_cm.h>
+#include <rdma/rw.h>
 
 #include <scsi/srp.h>
 
@@ -105,7 +106,6 @@ enum {
 	SRP_LOGIN_RSP_MULTICHAN_MAINTAINED = 0x2,
 
 	SRPT_DEF_SG_TABLESIZE = 128,
-	SRPT_DEF_SG_PER_WQE = 16,
 
 	MIN_SRPT_SQ_SIZE = 16,
 	DEF_SRPT_SQ_SIZE = 4096,
@@ -174,21 +174,17 @@ struct srpt_recv_ioctx {
 	struct srpt_ioctx	ioctx;
 	struct list_head	wait_list;
 };
+	
+struct srpt_rw_ctx {
+	struct rdma_rw_ctx	rw;
+	struct scatterlist	*sg;
+	unsigned int		nents;
+};
 
 /**
  * struct srpt_send_ioctx - SRPT send I/O context.
  * @ioctx:       See above.
  * @ch:          Channel pointer.
- * @free_list:   Node in srpt_rdma_ch.free_list.
- * @n_rbuf:      Number of data buffers in the received SRP command.
- * @rbufs:       Pointer to SRP data buffer array.
- * @single_rbuf: SRP data buffer if the command has only a single buffer.
- * @sg:          Pointer to sg-list associated with this I/O context.
- * @sg_cnt:      SG-list size.
- * @mapped_sg_count: ib_dma_map_sg() return value.
- * @n_rdma_wrs:  Number of elements in the rdma_wrs array.
- * @rdma_wrs:    Array with information about the RDMA mapping.
- * @tag:         Tag of the received SRP information unit.
  * @spinlock:    Protects 'state'.
  * @state:       I/O context state.
  * @cmd:         Target core command data structure.
@@ -197,21 +193,18 @@ struct srpt_recv_ioctx {
 struct srpt_send_ioctx {
 	struct srpt_ioctx	ioctx;
 	struct srpt_rdma_ch	*ch;
-	struct ib_rdma_wr	*rdma_wrs;
+
+	struct srpt_rw_ctx	s_rw_ctx;
+	struct srpt_rw_ctx	*rw_ctxs;
+
 	struct ib_cqe		rdma_cqe;
-	struct srp_direct_buf	*rbufs;
-	struct srp_direct_buf	single_rbuf;
-	struct scatterlist	*sg;
 	struct list_head	free_list;
 	spinlock_t		spinlock;
 	enum srpt_command_state	state;
 	struct se_cmd		cmd;
 	struct completion	tx_done;
-	int			sg_cnt;
-	int			mapped_sg_count;
-	u16			n_rdma_wrs;
 	u8			n_rdma;
-	u8			n_rbuf;
+	u8			n_rw_ctx;
 	bool			queue_status_only;
 	u8			sense_data[TRANSPORT_SENSE_BUFFER];
 };

drivers/net/ethernet/chelsio/cxgb4/t4_msg.h

@@ -1392,6 +1392,10 @@ struct ulp_mem_io {
 #define T5_ULP_MEMIO_ORDER_V(x) ((x) << T5_ULP_MEMIO_ORDER_S)
 #define T5_ULP_MEMIO_ORDER_F    T5_ULP_MEMIO_ORDER_V(1U)
 
+#define T5_ULP_MEMIO_FID_S	4
+#define T5_ULP_MEMIO_FID_M	0x7ff
+#define T5_ULP_MEMIO_FID_V(x)	((x) << T5_ULP_MEMIO_FID_S)
+
 /* ulp_mem_io.lock_addr fields */
 #define ULP_MEMIO_ADDR_S    0
 #define ULP_MEMIO_ADDR_V(x) ((x) << ULP_MEMIO_ADDR_S)

drivers/net/ethernet/mellanox/mlx5/core/cq.c

@@ -39,6 +39,53 @@
 #include <linux/mlx5/cq.h>
 #include "mlx5_core.h"
 
+#define TASKLET_MAX_TIME 2
+#define TASKLET_MAX_TIME_JIFFIES msecs_to_jiffies(TASKLET_MAX_TIME)
+
+void mlx5_cq_tasklet_cb(unsigned long data)
+{
+	unsigned long flags;
+	unsigned long end = jiffies + TASKLET_MAX_TIME_JIFFIES;
+	struct mlx5_eq_tasklet *ctx = (struct mlx5_eq_tasklet *)data;
+	struct mlx5_core_cq *mcq;
+	struct mlx5_core_cq *temp;
+
+	spin_lock_irqsave(&ctx->lock, flags);
+	list_splice_tail_init(&ctx->list, &ctx->process_list);
+	spin_unlock_irqrestore(&ctx->lock, flags);
+
+	list_for_each_entry_safe(mcq, temp, &ctx->process_list,
+				 tasklet_ctx.list) {
+		list_del_init(&mcq->tasklet_ctx.list);
+		mcq->tasklet_ctx.comp(mcq);
+		if (atomic_dec_and_test(&mcq->refcount))
+			complete(&mcq->free);
+		if (time_after(jiffies, end))
+			break;
+	}
+
+	if (!list_empty(&ctx->process_list))
+		tasklet_schedule(&ctx->task);
+}
+
+static void mlx5_add_cq_to_tasklet(struct mlx5_core_cq *cq)
+{
+	unsigned long flags;
+	struct mlx5_eq_tasklet *tasklet_ctx = cq->tasklet_ctx.priv;
+
+	spin_lock_irqsave(&tasklet_ctx->lock, flags);
+	/* When migrating CQs between EQs will be implemented, please note
+	 * that you need to sync this point. It is possible that
+	 * while migrating a CQ, completions on the old EQs could
+	 * still arrive.
+	 */
+	if (list_empty_careful(&cq->tasklet_ctx.list)) {
+		atomic_inc(&cq->refcount);
+		list_add_tail(&cq->tasklet_ctx.list, &tasklet_ctx->list);
+	}
+	spin_unlock_irqrestore(&tasklet_ctx->lock, flags);
+}
+
 void mlx5_cq_completion(struct mlx5_core_dev *dev, u32 cqn)
 {
 	struct mlx5_core_cq *cq;
@@ -96,6 +143,13 @@ int mlx5_core_create_cq(struct mlx5_core_dev *dev, struct mlx5_core_cq *cq,
 	struct mlx5_create_cq_mbox_out out;
 	struct mlx5_destroy_cq_mbox_in din;
 	struct mlx5_destroy_cq_mbox_out dout;
+	int eqn = MLX5_GET(cqc, MLX5_ADDR_OF(create_cq_in, in, cq_context),
+			   c_eqn);
+	struct mlx5_eq *eq;
+
+	eq = mlx5_eqn2eq(dev, eqn);
+	if (IS_ERR(eq))
+		return PTR_ERR(eq);
 
 	in->hdr.opcode = cpu_to_be16(MLX5_CMD_OP_CREATE_CQ);
 	memset(&out, 0, sizeof(out));
@@ -111,6 +165,11 @@ int mlx5_core_create_cq(struct mlx5_core_dev *dev, struct mlx5_core_cq *cq,
 	cq->arm_sn     = 0;
 	atomic_set(&cq->refcount, 1);
 	init_completion(&cq->free);
+	if (!cq->comp)
+		cq->comp = mlx5_add_cq_to_tasklet;
+	/* assuming CQ will be deleted before the EQ */
+	cq->tasklet_ctx.priv = &eq->tasklet_ctx;
+	INIT_LIST_HEAD(&cq->tasklet_ctx.list);
 
 	spin_lock_irq(&table->lock);
 	err = radix_tree_insert(&table->tree, cq->cqn, cq);

drivers/net/ethernet/mellanox/mlx5/core/eq.c

@@ -202,7 +202,7 @@ static int mlx5_eq_int(struct mlx5_core_dev *dev, struct mlx5_eq *eq)
 	struct mlx5_eqe *eqe;
 	int eqes_found = 0;
 	int set_ci = 0;
-	u32 cqn;
+	u32 cqn = -1;
 	u32 rsn;
 	u8 port;
 
@@ -320,6 +320,9 @@ static int mlx5_eq_int(struct mlx5_core_dev *dev, struct mlx5_eq *eq)
 
 	eq_update_ci(eq, 1);
 
+	if (cqn != -1)
+		tasklet_schedule(&eq->tasklet_ctx.task);
+
 	return eqes_found;
 }
 
@@ -403,6 +406,12 @@ int mlx5_create_map_eq(struct mlx5_core_dev *dev, struct mlx5_eq *eq, u8 vecidx,
 	if (err)
 		goto err_irq;
 
+	INIT_LIST_HEAD(&eq->tasklet_ctx.list);
+	INIT_LIST_HEAD(&eq->tasklet_ctx.process_list);
+	spin_lock_init(&eq->tasklet_ctx.lock);
+	tasklet_init(&eq->tasklet_ctx.task, mlx5_cq_tasklet_cb,
+		     (unsigned long)&eq->tasklet_ctx);
+
 	/* EQs are created in ARMED state
 	 */
 	eq_update_ci(eq, 1);
@@ -436,6 +445,7 @@ int mlx5_destroy_unmap_eq(struct mlx5_core_dev *dev, struct mlx5_eq *eq)
 		mlx5_core_warn(dev, "failed to destroy a previously created eq: eqn %d\n",
 			       eq->eqn);
 	synchronize_irq(eq->irqn);
+	tasklet_disable(&eq->tasklet_ctx.task);
 	mlx5_buf_free(dev, &eq->buf);
 
 	return err;

drivers/net/ethernet/mellanox/mlx5/core/main.c

@@ -663,6 +663,23 @@ int mlx5_vector2eqn(struct mlx5_core_dev *dev, int vector, int *eqn,
 }
 EXPORT_SYMBOL(mlx5_vector2eqn);
 
+struct mlx5_eq *mlx5_eqn2eq(struct mlx5_core_dev *dev, int eqn)
+{
+	struct mlx5_eq_table *table = &dev->priv.eq_table;
+	struct mlx5_eq *eq;
+
+	spin_lock(&table->lock);
+	list_for_each_entry(eq, &table->comp_eqs_list, list)
+		if (eq->eqn == eqn) {
+			spin_unlock(&table->lock);
+			return eq;
+		}
+
+	spin_unlock(&table->lock);
+
+	return ERR_PTR(-ENOENT);
+}
+
 static void free_comp_eqs(struct mlx5_core_dev *dev)
 {
 	struct mlx5_eq_table *table = &dev->priv.eq_table;

drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h

@@ -102,6 +102,8 @@ int mlx5_core_disable_hca(struct mlx5_core_dev *dev, u16 func_id);
 int mlx5_wait_for_vf_pages(struct mlx5_core_dev *dev);
 cycle_t mlx5_read_internal_timer(struct mlx5_core_dev *dev);
 u32 mlx5_get_msix_vec(struct mlx5_core_dev *dev, int vecidx);
+struct mlx5_eq *mlx5_eqn2eq(struct mlx5_core_dev *dev, int eqn);
+void mlx5_cq_tasklet_cb(unsigned long data);
 
 void mlx5e_init(void);
 void mlx5e_cleanup(void);

drivers/staging/rdma/hfi1/affinity.c

@@ -53,20 +53,6 @@
 #include "sdma.h"
 #include "trace.h"
 
-struct cpu_mask_set {
-	struct cpumask mask;
-	struct cpumask used;
-	uint gen;
-};
-
-struct hfi1_affinity {
-	struct cpu_mask_set def_intr;
-	struct cpu_mask_set rcv_intr;
-	struct cpu_mask_set proc;
-	/* spin lock to protect affinity struct */
-	spinlock_t lock;
-};
-
 /* Name of IRQ types, indexed by enum irq_type */
 static const char * const irq_type_names[] = {
 	"SDMA",
@@ -82,6 +68,48 @@ static inline void init_cpu_mask_set(struct cpu_mask_set *set)
 	set->gen = 0;
 }
 
+/* Initialize non-HT cpu cores mask */
+int init_real_cpu_mask(struct hfi1_devdata *dd)
+{
+	struct hfi1_affinity *info;
+	int possible, curr_cpu, i, ht;
+
+	info = kzalloc(sizeof(*info), GFP_KERNEL);
+	if (!info)
+		return -ENOMEM;
+
+	cpumask_clear(&info->real_cpu_mask);
+
+	/* Start with cpu online mask as the real cpu mask */
+	cpumask_copy(&info->real_cpu_mask, cpu_online_mask);
+
+	/*
+	 * Remove HT cores from the real cpu mask.  Do this in two steps below.
+	 */
+	possible = cpumask_weight(&info->real_cpu_mask);
+	ht = cpumask_weight(topology_sibling_cpumask(
+					cpumask_first(&info->real_cpu_mask)));
+	/*
+	 * Step 1.  Skip over the first N HT siblings and use them as the
+	 * "real" cores.  Assumes that HT cores are not enumerated in
+	 * succession (except in the single core case).
+	 */
+	curr_cpu = cpumask_first(&info->real_cpu_mask);
+	for (i = 0; i < possible / ht; i++)
+		curr_cpu = cpumask_next(curr_cpu, &info->real_cpu_mask);
+	/*
+	 * Step 2.  Remove the remaining HT siblings.  Use cpumask_next() to
+	 * skip any gaps.
+	 */
+	for (; i < possible; i++) {
+		cpumask_clear_cpu(curr_cpu, &info->real_cpu_mask);
+		curr_cpu = cpumask_next(curr_cpu, &info->real_cpu_mask);
+	}
+
+	dd->affinity = info;
+	return 0;
+}
+
 /*
  * Interrupt affinity.
  *
@@ -93,20 +121,17 @@ static inline void init_cpu_mask_set(struct cpu_mask_set *set)
  * to the node relative 1 as necessary.
  *
  */
-int hfi1_dev_affinity_init(struct hfi1_devdata *dd)
+void hfi1_dev_affinity_init(struct hfi1_devdata *dd)
 {
 	int node = pcibus_to_node(dd->pcidev->bus);
-	struct hfi1_affinity *info;
+	struct hfi1_affinity *info = dd->affinity;
 	const struct cpumask *local_mask;
-	int curr_cpu, possible, i, ht;
+	int curr_cpu, possible, i;
 
 	if (node < 0)
 		node = numa_node_id();
 	dd->node = node;
 
-	info = kzalloc(sizeof(*info), GFP_KERNEL);
-	if (!info)
-		return -ENOMEM;
 	spin_lock_init(&info->lock);
 
 	init_cpu_mask_set(&info->def_intr);
@@ -116,30 +141,8 @@ int hfi1_dev_affinity_init(struct hfi1_devdata *dd)
 	local_mask = cpumask_of_node(dd->node);
 	if (cpumask_first(local_mask) >= nr_cpu_ids)
 		local_mask = topology_core_cpumask(0);
-	/* use local mask as default */
-	cpumask_copy(&info->def_intr.mask, local_mask);
-	/*
-	 * Remove HT cores from the default mask.  Do this in two steps below.
-	 */
-	possible = cpumask_weight(&info->def_intr.mask);
-	ht = cpumask_weight(topology_sibling_cpumask(
-					cpumask_first(&info->def_intr.mask)));
-	/*
-	 * Step 1.  Skip over the first N HT siblings and use them as the
-	 * "real" cores.  Assumes that HT cores are not enumerated in
-	 * succession (except in the single core case).
-	 */
-	curr_cpu = cpumask_first(&info->def_intr.mask);
-	for (i = 0; i < possible / ht; i++)
-		curr_cpu = cpumask_next(curr_cpu, &info->def_intr.mask);
-	/*
-	 * Step 2.  Remove the remaining HT siblings.  Use cpumask_next() to
-	 * skip any gaps.
-	 */
-	for (; i < possible; i++) {
-		cpumask_clear_cpu(curr_cpu, &info->def_intr.mask);
-		curr_cpu = cpumask_next(curr_cpu, &info->def_intr.mask);
-	}
+	/* Use the "real" cpu mask of this node as the default */
+	cpumask_and(&info->def_intr.mask, &info->real_cpu_mask, local_mask);
 
 	/*  fill in the receive list */
 	possible = cpumask_weight(&info->def_intr.mask);
@@ -167,8 +170,6 @@ int hfi1_dev_affinity_init(struct hfi1_devdata *dd)
 	}
 
 	cpumask_copy(&info->proc.mask, cpu_online_mask);
-	dd->affinity = info;
-	return 0;
 }
 
 void hfi1_dev_affinity_free(struct hfi1_devdata *dd)

drivers/staging/rdma/hfi1/affinity.h

@@ -64,10 +64,27 @@ enum affinity_flags {
 	AFF_IRQ_LOCAL
 };
 
+struct cpu_mask_set {
+	struct cpumask mask;
+	struct cpumask used;
+	uint gen;
+};
+
+struct hfi1_affinity {
+	struct cpu_mask_set def_intr;
+	struct cpu_mask_set rcv_intr;
+	struct cpu_mask_set proc;
+	struct cpumask real_cpu_mask;
+	/* spin lock to protect affinity struct */
+	spinlock_t lock;
+};
+
 struct hfi1_msix_entry;
 
+/* Initialize non-HT cpu cores mask */
+int init_real_cpu_mask(struct hfi1_devdata *);
 /* Initialize driver affinity data */
-int hfi1_dev_affinity_init(struct hfi1_devdata *);
+void hfi1_dev_affinity_init(struct hfi1_devdata *);
 /* Free driver affinity data */
 void hfi1_dev_affinity_free(struct hfi1_devdata *);
 /*

drivers/staging/rdma/hfi1/chip.c

@@ -123,6 +123,8 @@ struct flag_table {
 
 #define MIN_KERNEL_KCTXTS         2
 #define FIRST_KERNEL_KCTXT        1
+/* sizes for both the QP and RSM map tables */
+#define NUM_MAP_ENTRIES		256
 #define NUM_MAP_REGS             32
 
 /* Bit offset into the GUID which carries HFI id information */
@@ -1029,9 +1031,12 @@ static int thermal_init(struct hfi1_devdata *dd);
 static int wait_logical_linkstate(struct hfi1_pportdata *ppd, u32 state,
 				  int msecs);
 static void read_planned_down_reason_code(struct hfi1_devdata *dd, u8 *pdrrc);
+static void read_link_down_reason(struct hfi1_devdata *dd, u8 *ldr);
 static void handle_temp_err(struct hfi1_devdata *);
 static void dc_shutdown(struct hfi1_devdata *);
 static void dc_start(struct hfi1_devdata *);
+static int qos_rmt_entries(struct hfi1_devdata *dd, unsigned int *mp,
+			   unsigned int *np);
 
 /*
  * Error interrupt table entry.  This is used as input to the interrupt
@@ -5661,7 +5666,7 @@ static int sc_to_vl(struct hfi1_devdata *dd, int sw_index)
 	sci = &dd->send_contexts[sw_index];
 
 	/* there is no information for user (PSM) and ack contexts */
-	if (sci->type != SC_KERNEL)
+	if ((sci->type != SC_KERNEL) && (sci->type != SC_VL15))
 		return -1;
 
 	sc = sci->sc;
@@ -6199,18 +6204,13 @@ static void hreq_response(struct hfi1_devdata *dd, u8 return_code, u16 rsp_data)
 
 /*
  * Handle host requests from the 8051.
- *
- * This is a work-queue function outside of the interrupt.
  */
-void handle_8051_request(struct work_struct *work)
+static void handle_8051_request(struct hfi1_pportdata *ppd)
 {
-	struct hfi1_pportdata *ppd = container_of(work, struct hfi1_pportdata,
-							dc_host_req_work);
 	struct hfi1_devdata *dd = ppd->dd;
 	u64 reg;
 	u16 data = 0;
-	u8 type, i, lanes, *cache = ppd->qsfp_info.cache;
-	u8 cdr_ctrl_byte = cache[QSFP_CDR_CTRL_BYTE_OFFS];
+	u8 type;
 
 	reg = read_csr(dd, DC_DC8051_CFG_EXT_DEV_1);
 	if ((reg & DC_DC8051_CFG_EXT_DEV_1_REQ_NEW_SMASK) == 0)
@@ -6231,46 +6231,11 @@ void handle_8051_request(struct work_struct *work)
 	case HREQ_READ_CONFIG:
 	case HREQ_SET_TX_EQ_ABS:
 	case HREQ_SET_TX_EQ_REL:
+	case HREQ_ENABLE:
 		dd_dev_info(dd, "8051 request: request 0x%x not supported\n",
 			    type);
 		hreq_response(dd, HREQ_NOT_SUPPORTED, 0);
 		break;
-
-	case HREQ_ENABLE:
-		lanes = data & 0xF;
-		for (i = 0; lanes; lanes >>= 1, i++) {
-			if (!(lanes & 1))
-				continue;
-			if (data & 0x200) {
-				/* enable TX CDR */
-				if (cache[QSFP_MOD_PWR_OFFS] & 0x8 &&
-				    cache[QSFP_CDR_INFO_OFFS] & 0x80)
-					cdr_ctrl_byte |= (1 << (i + 4));
-			} else {
-				/* disable TX CDR */
-				if (cache[QSFP_MOD_PWR_OFFS] & 0x8 &&
-				    cache[QSFP_CDR_INFO_OFFS] & 0x80)
-					cdr_ctrl_byte &= ~(1 << (i + 4));
-			}
-
-			if (data & 0x800) {
-				/* enable RX CDR */
-				if (cache[QSFP_MOD_PWR_OFFS] & 0x4 &&
-				    cache[QSFP_CDR_INFO_OFFS] & 0x40)
-					cdr_ctrl_byte |= (1 << i);
-			} else {
-				/* disable RX CDR */
-				if (cache[QSFP_MOD_PWR_OFFS] & 0x4 &&
-				    cache[QSFP_CDR_INFO_OFFS] & 0x40)
-					cdr_ctrl_byte &= ~(1 << i);
-			}
-		}
-		one_qsfp_write(ppd, dd->hfi1_id, QSFP_CDR_CTRL_BYTE_OFFS,
-			       &cdr_ctrl_byte, 1);
-		hreq_response(dd, HREQ_SUCCESS, data);
-		refresh_qsfp_cache(ppd, &ppd->qsfp_info);
-		break;
-
 	case HREQ_CONFIG_DONE:
 		hreq_response(dd, HREQ_SUCCESS, 0);
 		break;
@@ -6278,7 +6243,6 @@ void handle_8051_request(struct work_struct *work)
 	case HREQ_INTERFACE_TEST:
 		hreq_response(dd, HREQ_SUCCESS, data);
 		break;
-
 	default:
 		dd_dev_err(dd, "8051 request: unknown request 0x%x\n", type);
 		hreq_response(dd, HREQ_NOT_SUPPORTED, 0);
@@ -6849,6 +6813,75 @@ static void reset_neighbor_info(struct hfi1_pportdata *ppd)
 	ppd->neighbor_fm_security = 0;
 }
 
+static const char * const link_down_reason_strs[] = {
+	[OPA_LINKDOWN_REASON_NONE] = "None",
+	[OPA_LINKDOWN_REASON_RCV_ERROR_0] = "Recive error 0",
+	[OPA_LINKDOWN_REASON_BAD_PKT_LEN] = "Bad packet length",
+	[OPA_LINKDOWN_REASON_PKT_TOO_LONG] = "Packet too long",
+	[OPA_LINKDOWN_REASON_PKT_TOO_SHORT] = "Packet too short",
+	[OPA_LINKDOWN_REASON_BAD_SLID] = "Bad SLID",
+	[OPA_LINKDOWN_REASON_BAD_DLID] = "Bad DLID",
+	[OPA_LINKDOWN_REASON_BAD_L2] = "Bad L2",
+	[OPA_LINKDOWN_REASON_BAD_SC] = "Bad SC",
+	[OPA_LINKDOWN_REASON_RCV_ERROR_8] = "Receive error 8",
+	[OPA_LINKDOWN_REASON_BAD_MID_TAIL] = "Bad mid tail",
+	[OPA_LINKDOWN_REASON_RCV_ERROR_10] = "Receive error 10",
+	[OPA_LINKDOWN_REASON_PREEMPT_ERROR] = "Preempt error",
+	[OPA_LINKDOWN_REASON_PREEMPT_VL15] = "Preempt vl15",
+	[OPA_LINKDOWN_REASON_BAD_VL_MARKER] = "Bad VL marker",
+	[OPA_LINKDOWN_REASON_RCV_ERROR_14] = "Receive error 14",
+	[OPA_LINKDOWN_REASON_RCV_ERROR_15] = "Receive error 15",
+	[OPA_LINKDOWN_REASON_BAD_HEAD_DIST] = "Bad head distance",
+	[OPA_LINKDOWN_REASON_BAD_TAIL_DIST] = "Bad tail distance",
+	[OPA_LINKDOWN_REASON_BAD_CTRL_DIST] = "Bad control distance",
+	[OPA_LINKDOWN_REASON_BAD_CREDIT_ACK] = "Bad credit ack",
+	[OPA_LINKDOWN_REASON_UNSUPPORTED_VL_MARKER] = "Unsupported VL marker",
+	[OPA_LINKDOWN_REASON_BAD_PREEMPT] = "Bad preempt",
+	[OPA_LINKDOWN_REASON_BAD_CONTROL_FLIT] = "Bad control flit",
+	[OPA_LINKDOWN_REASON_EXCEED_MULTICAST_LIMIT] = "Exceed multicast limit",
+	[OPA_LINKDOWN_REASON_RCV_ERROR_24] = "Receive error 24",
+	[OPA_LINKDOWN_REASON_RCV_ERROR_25] = "Receive error 25",
+	[OPA_LINKDOWN_REASON_RCV_ERROR_26] = "Receive error 26",
+	[OPA_LINKDOWN_REASON_RCV_ERROR_27] = "Receive error 27",
+	[OPA_LINKDOWN_REASON_RCV_ERROR_28] = "Receive error 28",
+	[OPA_LINKDOWN_REASON_RCV_ERROR_29] = "Receive error 29",
+	[OPA_LINKDOWN_REASON_RCV_ERROR_30] = "Receive error 30",
+	[OPA_LINKDOWN_REASON_EXCESSIVE_BUFFER_OVERRUN] =
+					"Excessive buffer overrun",
+	[OPA_LINKDOWN_REASON_UNKNOWN] = "Unknown",
+	[OPA_LINKDOWN_REASON_REBOOT] = "Reboot",
+	[OPA_LINKDOWN_REASON_NEIGHBOR_UNKNOWN] = "Neighbor unknown",
+	[OPA_LINKDOWN_REASON_FM_BOUNCE] = "FM bounce",
+	[OPA_LINKDOWN_REASON_SPEED_POLICY] = "Speed policy",
+	[OPA_LINKDOWN_REASON_WIDTH_POLICY] = "Width policy",
+	[OPA_LINKDOWN_REASON_DISCONNECTED] = "Disconnected",
+	[OPA_LINKDOWN_REASON_LOCAL_MEDIA_NOT_INSTALLED] =
+					"Local media not installed",
+	[OPA_LINKDOWN_REASON_NOT_INSTALLED] = "Not installed",
+	[OPA_LINKDOWN_REASON_CHASSIS_CONFIG] = "Chassis config",
+	[OPA_LINKDOWN_REASON_END_TO_END_NOT_INSTALLED] =
+					"End to end not installed",
+	[OPA_LINKDOWN_REASON_POWER_POLICY] = "Power policy",
+	[OPA_LINKDOWN_REASON_LINKSPEED_POLICY] = "Link speed policy",
+	[OPA_LINKDOWN_REASON_LINKWIDTH_POLICY] = "Link width policy",
+	[OPA_LINKDOWN_REASON_SWITCH_MGMT] = "Switch management",
+	[OPA_LINKDOWN_REASON_SMA_DISABLED] = "SMA disabled",
+	[OPA_LINKDOWN_REASON_TRANSIENT] = "Transient"
+};
+
+/* return the neighbor link down reason string */
+static const char *link_down_reason_str(u8 reason)
+{
+	const char *str = NULL;
+
+	if (reason < ARRAY_SIZE(link_down_reason_strs))
+		str = link_down_reason_strs[reason];
+	if (!str)
+		str = "(invalid)";
+
+	return str;
+}
+
 /*
  * Handle a link down interrupt from the 8051.
  *
@@ -6857,8 +6890,11 @@ static void reset_neighbor_info(struct hfi1_pportdata *ppd)
 void handle_link_down(struct work_struct *work)
 {
 	u8 lcl_reason, neigh_reason = 0;
+	u8 link_down_reason;
 	struct hfi1_pportdata *ppd = container_of(work, struct hfi1_pportdata,
-								link_down_work);
+						  link_down_work);
+	int was_up;
+	static const char ldr_str[] = "Link down reason: ";
 
 	if ((ppd->host_link_state &
 	     (HLS_DN_POLL | HLS_VERIFY_CAP | HLS_GOING_UP)) &&
@@ -6867,20 +6903,63 @@ void handle_link_down(struct work_struct *work)
 			HFI1_ODR_MASK(OPA_LINKDOWN_REASON_NOT_INSTALLED);
 
 	/* Go offline first, then deal with reading/writing through 8051 */
+	was_up = !!(ppd->host_link_state & HLS_UP);
 	set_link_state(ppd, HLS_DN_OFFLINE);
 
-	lcl_reason = 0;
-	read_planned_down_reason_code(ppd->dd, &neigh_reason);
+	if (was_up) {
+		lcl_reason = 0;
+		/* link down reason is only valid if the link was up */
+		read_link_down_reason(ppd->dd, &link_down_reason);
+		switch (link_down_reason) {
+		case LDR_LINK_TRANSFER_ACTIVE_LOW:
+			/* the link went down, no idle message reason */
+			dd_dev_info(ppd->dd, "%sUnexpected link down\n",
+				    ldr_str);
+			break;
+		case LDR_RECEIVED_LINKDOWN_IDLE_MSG:
+			/*
+			 * The neighbor reason is only valid if an idle message
+			 * was received for it.
+			 */
+			read_planned_down_reason_code(ppd->dd, &neigh_reason);
+			dd_dev_info(ppd->dd,
+				    "%sNeighbor link down message %d, %s\n",
+				    ldr_str, neigh_reason,
+				    link_down_reason_str(neigh_reason));
+			break;
+		case LDR_RECEIVED_HOST_OFFLINE_REQ:
+			dd_dev_info(ppd->dd,
+				    "%sHost requested link to go offline\n",
+				    ldr_str);
+			break;
+		default:
+			dd_dev_info(ppd->dd, "%sUnknown reason 0x%x\n",
+				    ldr_str, link_down_reason);
+			break;
+		}
 
-	/*
-	 * If no reason, assume peer-initiated but missed
-	 * LinkGoingDown idle flits.
-	 */
-	if (neigh_reason == 0)
-		lcl_reason = OPA_LINKDOWN_REASON_NEIGHBOR_UNKNOWN;
+		/*
+		 * If no reason, assume peer-initiated but missed
+		 * LinkGoingDown idle flits.
+		 */
+		if (neigh_reason == 0)
+			lcl_reason = OPA_LINKDOWN_REASON_NEIGHBOR_UNKNOWN;
+	} else {
+		/* went down while polling or going up */
+		lcl_reason = OPA_LINKDOWN_REASON_TRANSIENT;
+	}
 
 	set_link_down_reason(ppd, lcl_reason, neigh_reason, 0);
 
+	/* inform the SMA when the link transitions from up to down */
+	if (was_up && ppd->local_link_down_reason.sma == 0 &&
+	    ppd->neigh_link_down_reason.sma == 0) {
+		ppd->local_link_down_reason.sma =
+					ppd->local_link_down_reason.latest;
+		ppd->neigh_link_down_reason.sma =
+					ppd->neigh_link_down_reason.latest;
+	}
+
 	reset_neighbor_info(ppd);
 
 	/* disable the port */
@@ -6890,7 +6969,7 @@ void handle_link_down(struct work_struct *work)
 	 * If there is no cable attached, turn the DC off. Otherwise,
 	 * start the link bring up.
 	 */
-	if (!qsfp_mod_present(ppd)) {
+	if (ppd->port_type == PORT_TYPE_QSFP && !qsfp_mod_present(ppd)) {
 		dc_shutdown(ppd->dd);
 	} else {
 		tune_serdes(ppd);
@@ -7373,7 +7452,11 @@ void apply_link_downgrade_policy(struct hfi1_pportdata *ppd, int refresh_widths)
 		ppd->link_width_downgrade_rx_active = rx;
 	}
 
-	if (lwde == 0) {
+	if (ppd->link_width_downgrade_tx_active == 0 ||
+	    ppd->link_width_downgrade_rx_active == 0) {
+		/* the 8051 reported a dead link as a downgrade */
+		dd_dev_err(ppd->dd, "Link downgrade is really a link down, ignoring\n");
+	} else if (lwde == 0) {
 		/* downgrade is disabled */
 
 		/* bounce if not at starting active width */
@@ -7534,7 +7617,7 @@ static void handle_8051_interrupt(struct hfi1_devdata *dd, u32 unused, u64 reg)
 			host_msg &= ~(u64)LINKUP_ACHIEVED;
 		}
 		if (host_msg & EXT_DEVICE_CFG_REQ) {
-			queue_work(ppd->hfi1_wq, &ppd->dc_host_req_work);
+			handle_8051_request(ppd);
 			host_msg &= ~(u64)EXT_DEVICE_CFG_REQ;
 		}
 		if (host_msg & VERIFY_CAP_FRAME) {
@@ -8660,6 +8743,14 @@ static void read_planned_down_reason_code(struct hfi1_devdata *dd, u8 *pdrrc)
 	*pdrrc = (frame >> DOWN_REMOTE_REASON_SHIFT) & DOWN_REMOTE_REASON_MASK;
 }
 
+static void read_link_down_reason(struct hfi1_devdata *dd, u8 *ldr)
+{
+	u32 frame;
+
+	read_8051_config(dd, LINK_DOWN_REASON, GENERAL_CONFIG, &frame);
+	*ldr = (frame & 0xff);
+}
+
 static int read_tx_settings(struct hfi1_devdata *dd,
 			    u8 *enable_lane_tx,
 			    u8 *tx_polarity_inversion,
@@ -9049,9 +9140,9 @@ static int set_local_link_attributes(struct hfi1_pportdata *ppd)
 }
 
 /*
- * Call this to start the link.  Schedule a retry if the cable is not
- * present or if unable to start polling.  Do not do anything if the
- * link is disabled.  Returns 0 if link is disabled or moved to polling
+ * Call this to start the link.
+ * Do not do anything if the link is disabled.
+ * Returns 0 if link is disabled, moved to polling, or the driver is not ready.
  */
 int start_link(struct hfi1_pportdata *ppd)
 {
@@ -9068,15 +9159,7 @@ int start_link(struct hfi1_pportdata *ppd)
 		return 0;
 	}
 
-	if (qsfp_mod_present(ppd) || loopback == LOOPBACK_SERDES ||
-	    loopback == LOOPBACK_LCB ||
-	    ppd->dd->icode == ICODE_FUNCTIONAL_SIMULATOR)
-		return set_link_state(ppd, HLS_DN_POLL);
-
-	dd_dev_info(ppd->dd,
-		    "%s: stopping link start because no cable is present\n",
-		    __func__);
-	return -EAGAIN;
+	return set_link_state(ppd, HLS_DN_POLL);
 }
 
 static void wait_for_qsfp_init(struct hfi1_pportdata *ppd)
@@ -9247,7 +9330,7 @@ static int handle_qsfp_error_conditions(struct hfi1_pportdata *ppd,
 	return 0;
 }
 
-/* This routine will only be scheduled if the QSFP module is present */
+/* This routine will only be scheduled if the QSFP module present is asserted */
 void qsfp_event(struct work_struct *work)
 {
 	struct qsfp_data *qd;
@@ -9676,6 +9759,7 @@ static void set_send_length(struct hfi1_pportdata *ppd)
 			      & SEND_LEN_CHECK1_LEN_VL15_MASK) <<
 		SEND_LEN_CHECK1_LEN_VL15_SHIFT;
 	int i;
+	u32 thres;
 
 	for (i = 0; i < ppd->vls_supported; i++) {
 		if (dd->vld[i].mtu > maxvlmtu)
@@ -9694,16 +9778,17 @@ static void set_send_length(struct hfi1_pportdata *ppd)
 	/* adjust kernel credit return thresholds based on new MTUs */
 	/* all kernel receive contexts have the same hdrqentsize */
 	for (i = 0; i < ppd->vls_supported; i++) {
-		sc_set_cr_threshold(dd->vld[i].sc,
-				    sc_mtu_to_threshold(dd->vld[i].sc,
-							dd->vld[i].mtu,
-							dd->rcd[0]->
-							rcvhdrqentsize));
-	}
-	sc_set_cr_threshold(dd->vld[15].sc,
-			    sc_mtu_to_threshold(dd->vld[15].sc,
-						dd->vld[15].mtu,
+		thres = min(sc_percent_to_threshold(dd->vld[i].sc, 50),
+			    sc_mtu_to_threshold(dd->vld[i].sc,
+						dd->vld[i].mtu,
 						dd->rcd[0]->rcvhdrqentsize));
+		sc_set_cr_threshold(dd->vld[i].sc, thres);
+	}
+	thres = min(sc_percent_to_threshold(dd->vld[15].sc, 50),
+		    sc_mtu_to_threshold(dd->vld[15].sc,
+					dd->vld[15].mtu,
+					dd->rcd[0]->rcvhdrqentsize));
+	sc_set_cr_threshold(dd->vld[15].sc, thres);
 
 	/* Adjust maximum MTU for the port in DC */
 	dcmtu = maxvlmtu == 10240 ? DCC_CFG_PORT_MTU_CAP_10240 :
@@ -10030,7 +10115,6 @@ int set_link_state(struct hfi1_pportdata *ppd, u32 state)
 	struct hfi1_devdata *dd = ppd->dd;
 	struct ib_event event = {.device = NULL};
 	int ret1, ret = 0;
-	int was_up, is_down;
 	int orig_new_state, poll_bounce;
 
 	mutex_lock(&ppd->hls_lock);
@@ -10049,8 +10133,6 @@ int set_link_state(struct hfi1_pportdata *ppd, u32 state)
 		    poll_bounce ? "(bounce) " : "",
 		    link_state_reason_name(ppd, state));
 
-	was_up = !!(ppd->host_link_state & HLS_UP);
-
 	/*
 	 * If we're going to a (HLS_*) link state that implies the logical
 	 * link state is neither of (IB_PORT_ARMED, IB_PORT_ACTIVE), then
@@ -10261,17 +10343,6 @@ int set_link_state(struct hfi1_pportdata *ppd, u32 state)
 		break;
 	}
 
-	is_down = !!(ppd->host_link_state & (HLS_DN_POLL |
-			HLS_DN_DISABLE | HLS_DN_OFFLINE));
-
-	if (was_up && is_down && ppd->local_link_down_reason.sma == 0 &&
-	    ppd->neigh_link_down_reason.sma == 0) {
-		ppd->local_link_down_reason.sma =
-		  ppd->local_link_down_reason.latest;
-		ppd->neigh_link_down_reason.sma =
-		  ppd->neigh_link_down_reason.latest;
-	}
-
 	goto done;
 
 unexpected:
@@ -12673,22 +12744,24 @@ static int set_up_context_variables(struct hfi1_devdata *dd)
 	int total_contexts;
 	int ret;
 	unsigned ngroups;
+	int qos_rmt_count;
+	int user_rmt_reduced;
 
 	/*
-	 * Kernel contexts: (to be fixed later):
-	 * - min or 2 or 1 context/numa
+	 * Kernel receive contexts:
+	 * - min of 2 or 1 context/numa (excluding control context)
 	 * - Context 0 - control context (VL15/multicast/error)
-	 * - Context 1 - default context
+	 * - Context 1 - first kernel context
+	 * - Context 2 - second kernel context
+	 * ...
 	 */
 	if (n_krcvqs)
 		/*
-		 * Don't count context 0 in n_krcvqs since
-		 * is isn't used for normal verbs traffic.
-		 *
-		 * krcvqs will reflect number of kernel
-		 * receive contexts above 0.
+		 * n_krcvqs is the sum of module parameter kernel receive
+		 * contexts, krcvqs[].  It does not include the control
+		 * context, so add that.
 		 */
-		num_kernel_contexts = n_krcvqs + MIN_KERNEL_KCTXTS - 1;
+		num_kernel_contexts = n_krcvqs + 1;
 	else
 		num_kernel_contexts = num_online_nodes() + 1;
 	num_kernel_contexts =
@@ -12705,12 +12778,13 @@ static int set_up_context_variables(struct hfi1_devdata *dd)
 		num_kernel_contexts = dd->chip_send_contexts - num_vls - 1;
 	}
 	/*
-	 * User contexts: (to be fixed later)
-	 *	- default to 1 user context per CPU if num_user_contexts is
-	 *	  negative
+	 * User contexts:
+	 *	- default to 1 user context per real (non-HT) CPU core if
+	 *	  num_user_contexts is negative
 	 */
 	if (num_user_contexts < 0)
-		num_user_contexts = num_online_cpus();
+		num_user_contexts =
+			cpumask_weight(&dd->affinity->real_cpu_mask);
 
 	total_contexts = num_kernel_contexts + num_user_contexts;
 
@@ -12727,6 +12801,19 @@ static int set_up_context_variables(struct hfi1_devdata *dd)
 		total_contexts = num_kernel_contexts + num_user_contexts;
 	}
 
+	/* each user context requires an entry in the RMT */
+	qos_rmt_count = qos_rmt_entries(dd, NULL, NULL);
+	if (qos_rmt_count + num_user_contexts > NUM_MAP_ENTRIES) {
+		user_rmt_reduced = NUM_MAP_ENTRIES - qos_rmt_count;
+		dd_dev_err(dd,
+			   "RMT size is reducing the number of user receive contexts from %d to %d\n",
+			   (int)num_user_contexts,
+			   user_rmt_reduced);
+		/* recalculate */
+		num_user_contexts = user_rmt_reduced;
+		total_contexts = num_kernel_contexts + num_user_contexts;
+	}
+
 	/* the first N are kernel contexts, the rest are user contexts */
 	dd->num_rcv_contexts = total_contexts;
 	dd->n_krcv_queues = num_kernel_contexts;
@@ -12776,12 +12863,13 @@ static int set_up_context_variables(struct hfi1_devdata *dd)
 		dd->num_send_contexts = ret;
 		dd_dev_info(
 			dd,
-			"send contexts: chip %d, used %d (kernel %d, ack %d, user %d)\n",
+			"send contexts: chip %d, used %d (kernel %d, ack %d, user %d, vl15 %d)\n",
 			dd->chip_send_contexts,
 			dd->num_send_contexts,
 			dd->sc_sizes[SC_KERNEL].count,
 			dd->sc_sizes[SC_ACK].count,
-			dd->sc_sizes[SC_USER].count);
+			dd->sc_sizes[SC_USER].count,
+			dd->sc_sizes[SC_VL15].count);
 		ret = 0;	/* success */
 	}
 
@@ -13451,122 +13539,224 @@ static void init_qpmap_table(struct hfi1_devdata *dd,
 	int i;
 	u64 ctxt = first_ctxt;
 
-	for (i = 0; i < 256;) {
+	for (i = 0; i < 256; i++) {
 		reg |= ctxt << (8 * (i % 8));
-		i++;
 		ctxt++;
 		if (ctxt > last_ctxt)
 			ctxt = first_ctxt;
-		if (i % 8 == 0) {
+		if (i % 8 == 7) {
 			write_csr(dd, regno, reg);
 			reg = 0;
 			regno += 8;
 		}
 	}
-	if (i % 8)
-		write_csr(dd, regno, reg);
 
 	add_rcvctrl(dd, RCV_CTRL_RCV_QP_MAP_ENABLE_SMASK
 			| RCV_CTRL_RCV_BYPASS_ENABLE_SMASK);
 }
 
-/**
- * init_qos - init RX qos
- * @dd - device data
- * @first_context
- *
- * This routine initializes Rule 0 and the
- * RSM map table to implement qos.
- *
- * If all of the limit tests succeed,
- * qos is applied based on the array
- * interpretation of krcvqs where
- * entry 0 is VL0.
- *
- * The number of vl bits (n) and the number of qpn
- * bits (m) are computed to feed both the RSM map table
- * and the single rule.
- *
- */
-static void init_qos(struct hfi1_devdata *dd, u32 first_ctxt)
-{
-	u8 max_by_vl = 0;
-	unsigned qpns_per_vl, ctxt, i, qpn, n = 1, m;
-	u64 *rsmmap;
-	u64 reg;
-	u8  rxcontext = is_ax(dd) ? 0 : 0xff;  /* 0 is default if a0 ver. */
+struct rsm_map_table {
+	u64 map[NUM_MAP_REGS];
+	unsigned int used;
+};
 
-	/* validate */
+struct rsm_rule_data {
+	u8 offset;
+	u8 pkt_type;
+	u32 field1_off;
+	u32 field2_off;
+	u32 index1_off;
+	u32 index1_width;
+	u32 index2_off;
+	u32 index2_width;
+	u32 mask1;
+	u32 value1;
+	u32 mask2;
+	u32 value2;
+};
+
+/*
+ * Return an initialized RMT map table for users to fill in.  OK if it
+ * returns NULL, indicating no table.
+ */
+static struct rsm_map_table *alloc_rsm_map_table(struct hfi1_devdata *dd)
+{
+	struct rsm_map_table *rmt;
+	u8 rxcontext = is_ax(dd) ? 0 : 0xff;  /* 0 is default if a0 ver. */
+
+	rmt = kmalloc(sizeof(*rmt), GFP_KERNEL);
+	if (rmt) {
+		memset(rmt->map, rxcontext, sizeof(rmt->map));
+		rmt->used = 0;
+	}
+
+	return rmt;
+}
+
+/*
+ * Write the final RMT map table to the chip and free the table.  OK if
+ * table is NULL.
+ */
+static void complete_rsm_map_table(struct hfi1_devdata *dd,
+				   struct rsm_map_table *rmt)
+{
+	int i;
+
+	if (rmt) {
+		/* write table to chip */
+		for (i = 0; i < NUM_MAP_REGS; i++)
+			write_csr(dd, RCV_RSM_MAP_TABLE + (8 * i), rmt->map[i]);
+
+		/* enable RSM */
+		add_rcvctrl(dd, RCV_CTRL_RCV_RSM_ENABLE_SMASK);
+	}
+}
+
+/*
+ * Add a receive side mapping rule.
+ */
+static void add_rsm_rule(struct hfi1_devdata *dd, u8 rule_index,
+			 struct rsm_rule_data *rrd)
+{
+	write_csr(dd, RCV_RSM_CFG + (8 * rule_index),
+		  (u64)rrd->offset << RCV_RSM_CFG_OFFSET_SHIFT |
+		  1ull << rule_index | /* enable bit */
+		  (u64)rrd->pkt_type << RCV_RSM_CFG_PACKET_TYPE_SHIFT);
+	write_csr(dd, RCV_RSM_SELECT + (8 * rule_index),
+		  (u64)rrd->field1_off << RCV_RSM_SELECT_FIELD1_OFFSET_SHIFT |
+		  (u64)rrd->field2_off << RCV_RSM_SELECT_FIELD2_OFFSET_SHIFT |
+		  (u64)rrd->index1_off << RCV_RSM_SELECT_INDEX1_OFFSET_SHIFT |
+		  (u64)rrd->index1_width << RCV_RSM_SELECT_INDEX1_WIDTH_SHIFT |
+		  (u64)rrd->index2_off << RCV_RSM_SELECT_INDEX2_OFFSET_SHIFT |
+		  (u64)rrd->index2_width << RCV_RSM_SELECT_INDEX2_WIDTH_SHIFT);
+	write_csr(dd, RCV_RSM_MATCH + (8 * rule_index),
+		  (u64)rrd->mask1 << RCV_RSM_MATCH_MASK1_SHIFT |
+		  (u64)rrd->value1 << RCV_RSM_MATCH_VALUE1_SHIFT |
+		  (u64)rrd->mask2 << RCV_RSM_MATCH_MASK2_SHIFT |
+		  (u64)rrd->value2 << RCV_RSM_MATCH_VALUE2_SHIFT);
+}
+
+/* return the number of RSM map table entries that will be used for QOS */
+static int qos_rmt_entries(struct hfi1_devdata *dd, unsigned int *mp,
+			   unsigned int *np)
+{
+	int i;
+	unsigned int m, n;
+	u8 max_by_vl = 0;
+
+	/* is QOS active at all? */
 	if (dd->n_krcv_queues <= MIN_KERNEL_KCTXTS ||
 	    num_vls == 1 ||
 	    krcvqsset <= 1)
-		goto bail;
-	for (i = 0; i < min_t(unsigned, num_vls, krcvqsset); i++)
+		goto no_qos;
+
+	/* determine bits for qpn */
+	for (i = 0; i < min_t(unsigned int, num_vls, krcvqsset); i++)
 		if (krcvqs[i] > max_by_vl)
 			max_by_vl = krcvqs[i];
 	if (max_by_vl > 32)
-		goto bail;
-	qpns_per_vl = __roundup_pow_of_two(max_by_vl);
-	/* determine bits vl */
-	n = ilog2(num_vls);
-	/* determine bits for qpn */
-	m = ilog2(qpns_per_vl);
+		goto no_qos;
+	m = ilog2(__roundup_pow_of_two(max_by_vl));
+
+	/* determine bits for vl */
+	n = ilog2(__roundup_pow_of_two(num_vls));
+
+	/* reject if too much is used */
 	if ((m + n) > 7)
+		goto no_qos;
+
+	if (mp)
+		*mp = m;
+	if (np)
+		*np = n;
+
+	return 1 << (m + n);
+
+no_qos:
+	if (mp)
+		*mp = 0;
+	if (np)
+		*np = 0;
+	return 0;
+}
+
+/**
+ * init_qos - init RX qos
+ * @dd - device data
+ * @rmt - RSM map table
+ *
+ * This routine initializes Rule 0 and the RSM map table to implement
+ * quality of service (qos).
+ *
+ * If all of the limit tests succeed, qos is applied based on the array
+ * interpretation of krcvqs where entry 0 is VL0.
+ *
+ * The number of vl bits (n) and the number of qpn bits (m) are computed to
+ * feed both the RSM map table and the single rule.
+ */
+static void init_qos(struct hfi1_devdata *dd, struct rsm_map_table *rmt)
+{
+	struct rsm_rule_data rrd;
+	unsigned qpns_per_vl, ctxt, i, qpn, n = 1, m;
+	unsigned int rmt_entries;
+	u64 reg;
+
+	if (!rmt)
 		goto bail;
-	if (num_vls * qpns_per_vl > dd->chip_rcv_contexts)
+	rmt_entries = qos_rmt_entries(dd, &m, &n);
+	if (rmt_entries == 0)
 		goto bail;
-	rsmmap = kmalloc_array(NUM_MAP_REGS, sizeof(u64), GFP_KERNEL);
-	if (!rsmmap)
+	qpns_per_vl = 1 << m;
+
+	/* enough room in the map table? */
+	rmt_entries = 1 << (m + n);
+	if (rmt->used + rmt_entries >= NUM_MAP_ENTRIES)
 		goto bail;
-	memset(rsmmap, rxcontext, NUM_MAP_REGS * sizeof(u64));
-	/* init the local copy of the table */
-	for (i = 0, ctxt = first_ctxt; i < num_vls; i++) {
+
+	/* add qos entries to the the RSM map table */
+	for (i = 0, ctxt = FIRST_KERNEL_KCTXT; i < num_vls; i++) {
 		unsigned tctxt;
 
 		for (qpn = 0, tctxt = ctxt;
 		     krcvqs[i] && qpn < qpns_per_vl; qpn++) {
 			unsigned idx, regoff, regidx;
 
-			/* generate index <= 128 */
-			idx = (qpn << n) ^ i;
+			/* generate the index the hardware will produce */
+			idx = rmt->used + ((qpn << n) ^ i);
 			regoff = (idx % 8) * 8;
 			regidx = idx / 8;
-			reg = rsmmap[regidx];
-			/* replace 0xff with context number */
+			/* replace default with context number */
+			reg = rmt->map[regidx];
 			reg &= ~(RCV_RSM_MAP_TABLE_RCV_CONTEXT_A_MASK
 				<< regoff);
 			reg |= (u64)(tctxt++) << regoff;
-			rsmmap[regidx] = reg;
+			rmt->map[regidx] = reg;
 			if (tctxt == ctxt + krcvqs[i])
 				tctxt = ctxt;
 		}
 		ctxt += krcvqs[i];
 	}
-	/* flush cached copies to chip */
-	for (i = 0; i < NUM_MAP_REGS; i++)
-		write_csr(dd, RCV_RSM_MAP_TABLE + (8 * i), rsmmap[i]);
-	/* add rule0 */
-	write_csr(dd, RCV_RSM_CFG /* + (8 * 0) */,
-		  RCV_RSM_CFG_ENABLE_OR_CHAIN_RSM0_MASK <<
-		  RCV_RSM_CFG_ENABLE_OR_CHAIN_RSM0_SHIFT |
-		  2ull << RCV_RSM_CFG_PACKET_TYPE_SHIFT);
-	write_csr(dd, RCV_RSM_SELECT /* + (8 * 0) */,
-		  LRH_BTH_MATCH_OFFSET << RCV_RSM_SELECT_FIELD1_OFFSET_SHIFT |
-		  LRH_SC_MATCH_OFFSET << RCV_RSM_SELECT_FIELD2_OFFSET_SHIFT |
-		  LRH_SC_SELECT_OFFSET << RCV_RSM_SELECT_INDEX1_OFFSET_SHIFT |
-		  ((u64)n) << RCV_RSM_SELECT_INDEX1_WIDTH_SHIFT |
-		  QPN_SELECT_OFFSET << RCV_RSM_SELECT_INDEX2_OFFSET_SHIFT |
-		  ((u64)m + (u64)n) << RCV_RSM_SELECT_INDEX2_WIDTH_SHIFT);
-	write_csr(dd, RCV_RSM_MATCH /* + (8 * 0) */,
-		  LRH_BTH_MASK << RCV_RSM_MATCH_MASK1_SHIFT |
-		  LRH_BTH_VALUE << RCV_RSM_MATCH_VALUE1_SHIFT |
-		  LRH_SC_MASK << RCV_RSM_MATCH_MASK2_SHIFT |
-		  LRH_SC_VALUE << RCV_RSM_MATCH_VALUE2_SHIFT);
-	/* Enable RSM */
-	add_rcvctrl(dd, RCV_CTRL_RCV_RSM_ENABLE_SMASK);
-	kfree(rsmmap);
-	/* map everything else to first context */
-	init_qpmap_table(dd, FIRST_KERNEL_KCTXT, MIN_KERNEL_KCTXTS - 1);
+
+	rrd.offset = rmt->used;
+	rrd.pkt_type = 2;
+	rrd.field1_off = LRH_BTH_MATCH_OFFSET;
+	rrd.field2_off = LRH_SC_MATCH_OFFSET;
+	rrd.index1_off = LRH_SC_SELECT_OFFSET;
+	rrd.index1_width = n;
+	rrd.index2_off = QPN_SELECT_OFFSET;
+	rrd.index2_width = m + n;
+	rrd.mask1 = LRH_BTH_MASK;
+	rrd.value1 = LRH_BTH_VALUE;
+	rrd.mask2 = LRH_SC_MASK;
+	rrd.value2 = LRH_SC_VALUE;
+
+	/* add rule 0 */
+	add_rsm_rule(dd, 0, &rrd);
+
+	/* mark RSM map entries as used */
+	rmt->used += rmt_entries;
+	/* map everything else to the mcast/err/vl15 context */
+	init_qpmap_table(dd, HFI1_CTRL_CTXT, HFI1_CTRL_CTXT);
 	dd->qos_shift = n + 1;
 	return;
 bail:
@@ -13574,13 +13764,86 @@ static void init_qos(struct hfi1_devdata *dd, u32 first_ctxt)
 	init_qpmap_table(dd, FIRST_KERNEL_KCTXT, dd->n_krcv_queues - 1);
 }
 
+static void init_user_fecn_handling(struct hfi1_devdata *dd,
+				    struct rsm_map_table *rmt)
+{
+	struct rsm_rule_data rrd;
+	u64 reg;
+	int i, idx, regoff, regidx;
+	u8 offset;
+
+	/* there needs to be enough room in the map table */
+	if (rmt->used + dd->num_user_contexts >= NUM_MAP_ENTRIES) {
+		dd_dev_err(dd, "User FECN handling disabled - too many user contexts allocated\n");
+		return;
+	}
+
+	/*
+	 * RSM will extract the destination context as an index into the
+	 * map table.  The destination contexts are a sequential block
+	 * in the range first_user_ctxt...num_rcv_contexts-1 (inclusive).
+	 * Map entries are accessed as offset + extracted value.  Adjust
+	 * the added offset so this sequence can be placed anywhere in
+	 * the table - as long as the entries themselves do not wrap.
+	 * There are only enough bits in offset for the table size, so
+	 * start with that to allow for a "negative" offset.
+	 */
+	offset = (u8)(NUM_MAP_ENTRIES + (int)rmt->used -
+						(int)dd->first_user_ctxt);
+
+	for (i = dd->first_user_ctxt, idx = rmt->used;
+				i < dd->num_rcv_contexts; i++, idx++) {
+		/* replace with identity mapping */
+		regoff = (idx % 8) * 8;
+		regidx = idx / 8;
+		reg = rmt->map[regidx];
+		reg &= ~(RCV_RSM_MAP_TABLE_RCV_CONTEXT_A_MASK << regoff);
+		reg |= (u64)i << regoff;
+		rmt->map[regidx] = reg;
+	}
+
+	/*
+	 * For RSM intercept of Expected FECN packets:
+	 * o packet type 0 - expected
+	 * o match on F (bit 95), using select/match 1, and
+	 * o match on SH (bit 133), using select/match 2.
+	 *
+	 * Use index 1 to extract the 8-bit receive context from DestQP
+	 * (start at bit 64).  Use that as the RSM map table index.
+	 */
+	rrd.offset = offset;
+	rrd.pkt_type = 0;
+	rrd.field1_off = 95;
+	rrd.field2_off = 133;
+	rrd.index1_off = 64;
+	rrd.index1_width = 8;
+	rrd.index2_off = 0;
+	rrd.index2_width = 0;
+	rrd.mask1 = 1;
+	rrd.value1 = 1;
+	rrd.mask2 = 1;
+	rrd.value2 = 1;
+
+	/* add rule 1 */
+	add_rsm_rule(dd, 1, &rrd);
+
+	rmt->used += dd->num_user_contexts;
+}
+
 static void init_rxe(struct hfi1_devdata *dd)
 {
+	struct rsm_map_table *rmt;
+
 	/* enable all receive errors */
 	write_csr(dd, RCV_ERR_MASK, ~0ull);
-	/* setup QPN map table - start where VL15 context leaves off */
-	init_qos(dd, dd->n_krcv_queues > MIN_KERNEL_KCTXTS ?
-		 MIN_KERNEL_KCTXTS : 0);
+
+	rmt = alloc_rsm_map_table(dd);
+	/* set up QOS, including the QPN map table */
+	init_qos(dd, rmt);
+	init_user_fecn_handling(dd, rmt);
+	complete_rsm_map_table(dd, rmt);
+	kfree(rmt);
+
 	/*
 	 * make sure RcvCtrl.RcvWcb <= PCIe Device Control
 	 * Register Max_Payload_Size (PCI_EXP_DEVCTL in Linux PCIe config
@@ -13762,6 +14025,7 @@ int hfi1_set_ctxt_pkey(struct hfi1_devdata *dd, unsigned ctxt, u16 pkey)
 	write_kctxt_csr(dd, sctxt, SEND_CTXT_CHECK_PARTITION_KEY, reg);
 	reg = read_kctxt_csr(dd, sctxt, SEND_CTXT_CHECK_ENABLE);
 	reg |= SEND_CTXT_CHECK_ENABLE_CHECK_PARTITION_KEY_SMASK;
+	reg &= ~SEND_CTXT_CHECK_ENABLE_DISALLOW_KDETH_PACKETS_SMASK;
 	write_kctxt_csr(dd, sctxt, SEND_CTXT_CHECK_ENABLE, reg);
 done:
 	return ret;
@@ -14148,6 +14412,19 @@ struct hfi1_devdata *hfi1_init_dd(struct pci_dev *pdev,
 		 (dd->revision >> CCE_REVISION_SW_SHIFT)
 		    & CCE_REVISION_SW_MASK);
 
+	/*
+	 * The real cpu mask is part of the affinity struct but has to be
+	 * initialized earlier than the rest of the affinity struct because it
+	 * is needed to calculate the number of user contexts in
+	 * set_up_context_variables(). However, hfi1_dev_affinity_init(),
+	 * which initializes the rest of the affinity struct members,
+	 * depends on set_up_context_variables() for the number of kernel
+	 * contexts, so it cannot be called before set_up_context_variables().
+	 */
+	ret = init_real_cpu_mask(dd);
+	if (ret)
+		goto bail_cleanup;
+
 	ret = set_up_context_variables(dd);
 	if (ret)
 		goto bail_cleanup;
@@ -14161,9 +14438,7 @@ struct hfi1_devdata *hfi1_init_dd(struct pci_dev *pdev,
 	/* set up KDETH QP prefix in both RX and TX CSRs */
 	init_kdeth_qp(dd);
 
-	ret = hfi1_dev_affinity_init(dd);
-	if (ret)
-		goto bail_cleanup;
+	hfi1_dev_affinity_init(dd);
 
 	/* send contexts must be set up before receive contexts */
 	ret = init_send_contexts(dd);

drivers/staging/rdma/hfi1/chip.h

@@ -389,6 +389,7 @@
 #define LAST_REMOTE_STATE_COMPLETE   0x13
 #define LINK_QUALITY_INFO            0x14
 #define REMOTE_DEVICE_ID	     0x15
+#define LINK_DOWN_REASON	     0x16
 
 /* 8051 lane specific register field IDs */
 #define TX_EQ_SETTINGS		0x00
@@ -497,6 +498,11 @@
 #define PWRM_BER_CONTROL	0x1
 #define PWRM_BANDWIDTH_CONTROL	0x2
 
+/* 8051 link down reasons */
+#define LDR_LINK_TRANSFER_ACTIVE_LOW   0xa
+#define LDR_RECEIVED_LINKDOWN_IDLE_MSG 0xb
+#define LDR_RECEIVED_HOST_OFFLINE_REQ  0xc
+
 /* verify capability fabric CRC size bits */
 enum {
 	CAP_CRC_14B = (1 << 0), /* 14b CRC */
@@ -691,7 +697,6 @@ void handle_verify_cap(struct work_struct *work);
 void handle_freeze(struct work_struct *work);
 void handle_link_up(struct work_struct *work);
 void handle_link_down(struct work_struct *work);
-void handle_8051_request(struct work_struct *work);
 void handle_link_downgrade(struct work_struct *work);
 void handle_link_bounce(struct work_struct *work);
 void handle_sma_message(struct work_struct *work);

drivers/staging/rdma/hfi1/chip_registers.h

@@ -771,6 +771,7 @@
 #define RCV_RSM_CFG_ENABLE_OR_CHAIN_RSM0_MASK 0x1ull
 #define RCV_RSM_CFG_ENABLE_OR_CHAIN_RSM0_SHIFT 0
 #define RCV_RSM_CFG_PACKET_TYPE_SHIFT 60
+#define RCV_RSM_CFG_OFFSET_SHIFT 32
 #define RCV_RSM_MAP_TABLE (RXE + 0x000000000900)
 #define RCV_RSM_MAP_TABLE_RCV_CONTEXT_A_MASK 0xFFull
 #define RCV_RSM_MATCH (RXE + 0x000000000800)

drivers/staging/rdma/hfi1/diag.c

@@ -413,7 +413,8 @@ static ssize_t diagpkt_send(struct diag_pkt *dp)
 		goto bail;
 	}
 	/* can only use kernel contexts */
-	if (dd->send_contexts[dp->sw_index].type != SC_KERNEL) {
+	if (dd->send_contexts[dp->sw_index].type != SC_KERNEL &&
+	    dd->send_contexts[dp->sw_index].type != SC_VL15) {
 		ret = -EINVAL;
 		goto bail;
 	}

drivers/staging/rdma/hfi1/driver.c

@@ -75,7 +75,8 @@ DEFINE_MUTEX(hfi1_mutex);	/* general driver use */
 
 unsigned int hfi1_max_mtu = HFI1_DEFAULT_MAX_MTU;
 module_param_named(max_mtu, hfi1_max_mtu, uint, S_IRUGO);
-MODULE_PARM_DESC(max_mtu, "Set max MTU bytes, default is 8192");
+MODULE_PARM_DESC(max_mtu, "Set max MTU bytes, default is " __stringify(
+		 HFI1_DEFAULT_MAX_MTU));
 
 unsigned int hfi1_cu = 1;
 module_param_named(cu, hfi1_cu, uint, S_IRUGO);

drivers/staging/rdma/hfi1/firmware.c

@@ -1413,8 +1413,15 @@ static int __acquire_chip_resource(struct hfi1_devdata *dd, u32 resource)
 
 	if (resource & CR_DYN_MASK) {
 		/* a dynamic resource is in use if either HFI has set the bit */
-		all_bits = resource_mask(0, resource) |
+		if (dd->pcidev->device == PCI_DEVICE_ID_INTEL0 &&
+		    (resource & (CR_I2C1 | CR_I2C2))) {
+			/* discrete devices must serialize across both chains */
+			all_bits = resource_mask(0, CR_I2C1 | CR_I2C2) |
+					resource_mask(1, CR_I2C1 | CR_I2C2);
+		} else {
+			all_bits = resource_mask(0, resource) |
 						resource_mask(1, resource);
+		}
 		my_bit = resource_mask(dd->hfi1_id, resource);
 	} else {
 		/* non-dynamic resources are not split between HFIs */

drivers/staging/rdma/hfi1/hfi.h

@@ -455,9 +455,9 @@ struct rvt_sge_state;
 #define HLS_UP (HLS_UP_INIT | HLS_UP_ARMED | HLS_UP_ACTIVE)
 
 /* use this MTU size if none other is given */
-#define HFI1_DEFAULT_ACTIVE_MTU 8192
+#define HFI1_DEFAULT_ACTIVE_MTU 10240
 /* use this MTU size as the default maximum */
-#define HFI1_DEFAULT_MAX_MTU 8192
+#define HFI1_DEFAULT_MAX_MTU 10240
 /* default partition key */
 #define DEFAULT_PKEY 0xffff
 
@@ -606,7 +606,6 @@ struct hfi1_pportdata {
 	struct work_struct link_vc_work;
 	struct work_struct link_up_work;
 	struct work_struct link_down_work;
-	struct work_struct dc_host_req_work;
 	struct work_struct sma_message_work;
 	struct work_struct freeze_work;
 	struct work_struct link_downgrade_work;
@@ -1258,7 +1257,7 @@ void receive_interrupt_work(struct work_struct *work);
 static inline int hdr2sc(struct hfi1_message_header *hdr, u64 rhf)
 {
 	return ((be16_to_cpu(hdr->lrh[0]) >> 12) & 0xf) |
-	       ((!!(rhf & RHF_DC_INFO_MASK)) << 4);
+	       ((!!(rhf & RHF_DC_INFO_SMASK)) << 4);
 }
 
 static inline u16 generate_jkey(kuid_t uid)
@@ -1333,6 +1332,9 @@ void process_becn(struct hfi1_pportdata *ppd, u8 sl,  u16 rlid, u32 lqpn,
 void return_cnp(struct hfi1_ibport *ibp, struct rvt_qp *qp, u32 remote_qpn,
 		u32 pkey, u32 slid, u32 dlid, u8 sc5,
 		const struct ib_grh *old_grh);
+#define PKEY_CHECK_INVALID -1
+int egress_pkey_check(struct hfi1_pportdata *ppd, __be16 *lrh, __be32 *bth,
+		      u8 sc5, int8_t s_pkey_index);
 
 #define PACKET_EGRESS_TIMEOUT 350
 static inline void pause_for_credit_return(struct hfi1_devdata *dd)
@@ -1776,6 +1778,7 @@ extern struct mutex hfi1_mutex;
 
 #define HFI1_PKT_USER_SC_INTEGRITY					    \
 	(SEND_CTXT_CHECK_ENABLE_DISALLOW_NON_KDETH_PACKETS_SMASK	    \
+	| SEND_CTXT_CHECK_ENABLE_DISALLOW_KDETH_PACKETS_SMASK		\
 	| SEND_CTXT_CHECK_ENABLE_DISALLOW_BYPASS_SMASK		    \
 	| SEND_CTXT_CHECK_ENABLE_DISALLOW_GRH_SMASK)
 

drivers/staging/rdma/hfi1/init.c

@@ -422,9 +422,10 @@ static enum hrtimer_restart cca_timer_fn(struct hrtimer *t)
 	struct cca_timer *cca_timer;
 	struct hfi1_pportdata *ppd;
 	int sl;
-	u16 ccti, ccti_timer, ccti_min;
+	u16 ccti_timer, ccti_min;
 	struct cc_state *cc_state;
 	unsigned long flags;
+	enum hrtimer_restart ret = HRTIMER_NORESTART;
 
 	cca_timer = container_of(t, struct cca_timer, hrtimer);
 	ppd = cca_timer->ppd;
@@ -450,24 +451,21 @@ static enum hrtimer_restart cca_timer_fn(struct hrtimer *t)
 
 	spin_lock_irqsave(&ppd->cca_timer_lock, flags);
 
-	ccti = cca_timer->ccti;
-
-	if (ccti > ccti_min) {
+	if (cca_timer->ccti > ccti_min) {
 		cca_timer->ccti--;
 		set_link_ipg(ppd);
 	}
 
-	spin_unlock_irqrestore(&ppd->cca_timer_lock, flags);
-
-	rcu_read_unlock();
-
-	if (ccti > ccti_min) {
+	if (cca_timer->ccti > ccti_min) {
 		unsigned long nsec = 1024 * ccti_timer;
 		/* ccti_timer is in units of 1.024 usec */
 		hrtimer_forward_now(t, ns_to_ktime(nsec));
-		return HRTIMER_RESTART;
+		ret = HRTIMER_RESTART;
 	}
-	return HRTIMER_NORESTART;
+
+	spin_unlock_irqrestore(&ppd->cca_timer_lock, flags);
+	rcu_read_unlock();
+	return ret;
 }
 
 /*
@@ -496,7 +494,6 @@ void hfi1_init_pportdata(struct pci_dev *pdev, struct hfi1_pportdata *ppd,
 	INIT_WORK(&ppd->link_vc_work, handle_verify_cap);
 	INIT_WORK(&ppd->link_up_work, handle_link_up);
 	INIT_WORK(&ppd->link_down_work, handle_link_down);
-	INIT_WORK(&ppd->dc_host_req_work, handle_8051_request);
 	INIT_WORK(&ppd->freeze_work, handle_freeze);
 	INIT_WORK(&ppd->link_downgrade_work, handle_link_downgrade);
 	INIT_WORK(&ppd->sma_message_work, handle_sma_message);
@@ -1007,7 +1004,7 @@ void hfi1_free_devdata(struct hfi1_devdata *dd)
 	free_percpu(dd->rcv_limit);
 	hfi1_dev_affinity_free(dd);
 	free_percpu(dd->send_schedule);
-	ib_dealloc_device(&dd->verbs_dev.rdi.ibdev);
+	rvt_dealloc_device(&dd->verbs_dev.rdi);
 }
 
 /*
@@ -1110,7 +1107,7 @@ struct hfi1_devdata *hfi1_alloc_devdata(struct pci_dev *pdev, size_t extra)
 bail:
 	if (!list_empty(&dd->list))
 		list_del_init(&dd->list);
-	ib_dealloc_device(&dd->verbs_dev.rdi.ibdev);
+	rvt_dealloc_device(&dd->verbs_dev.rdi);
 	return ERR_PTR(ret);
 }
 

drivers/staging/rdma/hfi1/mad.c

@@ -999,7 +999,21 @@ static int set_port_states(struct hfi1_pportdata *ppd, struct opa_smp *smp,
 			break;
 		}
 
-		set_link_state(ppd, link_state);
+		if ((link_state == HLS_DN_POLL ||
+		     link_state == HLS_DN_DOWNDEF)) {
+			/*
+			 * Going to poll.  No matter what the current state,
+			 * always move offline first, then tune and start the
+			 * link.  This correctly handles a FM link bounce and
+			 * a link enable.  Going offline is a no-op if already
+			 * offline.
+			 */
+			set_link_state(ppd, HLS_DN_OFFLINE);
+			tune_serdes(ppd);
+			start_link(ppd);
+		} else {
+			set_link_state(ppd, link_state);
+		}
 		if (link_state == HLS_DN_DISABLE &&
 		    (ppd->offline_disabled_reason >
 		     HFI1_ODR_MASK(OPA_LINKDOWN_REASON_SMA_DISABLED) ||

drivers/staging/rdma/hfi1/mmu_rb.c

@@ -91,7 +91,7 @@ static unsigned long mmu_node_start(struct mmu_rb_node *node)
 
 static unsigned long mmu_node_last(struct mmu_rb_node *node)
 {
-	return PAGE_ALIGN((node->addr & PAGE_MASK) + node->len) - 1;
+	return PAGE_ALIGN(node->addr + node->len) - 1;
 }
 
 int hfi1_mmu_rb_register(struct rb_root *root, struct mmu_rb_ops *ops)
@@ -126,10 +126,15 @@ void hfi1_mmu_rb_unregister(struct rb_root *root)
 	if (!handler)
 		return;
 
+	/* Unregister first so we don't get any more notifications. */
+	if (current->mm)
+		mmu_notifier_unregister(&handler->mn, current->mm);
+
 	spin_lock_irqsave(&mmu_rb_lock, flags);
 	list_del(&handler->list);
 	spin_unlock_irqrestore(&mmu_rb_lock, flags);
 
+	spin_lock_irqsave(&handler->lock, flags);
 	if (!RB_EMPTY_ROOT(root)) {
 		struct rb_node *node;
 		struct mmu_rb_node *rbnode;
@@ -141,9 +146,8 @@ void hfi1_mmu_rb_unregister(struct rb_root *root)
 				handler->ops->remove(root, rbnode, NULL);
 		}
 	}
+	spin_unlock_irqrestore(&handler->lock, flags);
 
-	if (current->mm)
-		mmu_notifier_unregister(&handler->mn, current->mm);
 	kfree(handler);
 }
 
@@ -235,6 +239,25 @@ struct mmu_rb_node *hfi1_mmu_rb_search(struct rb_root *root, unsigned long addr,
 	return node;
 }
 
+struct mmu_rb_node *hfi1_mmu_rb_extract(struct rb_root *root,
+					unsigned long addr, unsigned long len)
+{
+	struct mmu_rb_handler *handler = find_mmu_handler(root);
+	struct mmu_rb_node *node;
+	unsigned long flags;
+
+	if (!handler)
+		return ERR_PTR(-EINVAL);
+
+	spin_lock_irqsave(&handler->lock, flags);
+	node = __mmu_rb_search(handler, addr, len);
+	if (node)
+		__mmu_int_rb_remove(node, handler->root);
+	spin_unlock_irqrestore(&handler->lock, flags);
+
+	return node;
+}
+
 void hfi1_mmu_rb_remove(struct rb_root *root, struct mmu_rb_node *node)
 {
 	struct mmu_rb_handler *handler = find_mmu_handler(root);
@@ -293,9 +316,9 @@ static void mmu_notifier_mem_invalidate(struct mmu_notifier *mn,
 		hfi1_cdbg(MMU, "Invalidating node addr 0x%llx, len %u",
 			  node->addr, node->len);
 		if (handler->ops->invalidate(root, node)) {
-			spin_unlock_irqrestore(&handler->lock, flags);
-			__mmu_rb_remove(handler, node, mm);
-			spin_lock_irqsave(&handler->lock, flags);
+			__mmu_int_rb_remove(node, root);
+			if (handler->ops->remove)
+				handler->ops->remove(root, node, mm);
 		}
 	}
 	spin_unlock_irqrestore(&handler->lock, flags);

drivers/staging/rdma/hfi1/mmu_rb.h

@@ -70,5 +70,7 @@ int hfi1_mmu_rb_insert(struct rb_root *, struct mmu_rb_node *);
 void hfi1_mmu_rb_remove(struct rb_root *, struct mmu_rb_node *);
 struct mmu_rb_node *hfi1_mmu_rb_search(struct rb_root *, unsigned long,
 				       unsigned long);
+struct mmu_rb_node *hfi1_mmu_rb_extract(struct rb_root *, unsigned long,
+					unsigned long);
 
 #endif /* _HFI1_MMU_RB_H */

drivers/staging/rdma/hfi1/pio.c

@@ -139,23 +139,30 @@ void pio_send_control(struct hfi1_devdata *dd, int op)
 /* Send Context Size (SCS) wildcards */
 #define SCS_POOL_0 -1
 #define SCS_POOL_1 -2
+
 /* Send Context Count (SCC) wildcards */
 #define SCC_PER_VL -1
 #define SCC_PER_CPU  -2
-
 #define SCC_PER_KRCVQ  -3
-#define SCC_ACK_CREDITS  32
+
+/* Send Context Size (SCS) constants */
+#define SCS_ACK_CREDITS  32
+#define SCS_VL15_CREDITS 102	/* 3 pkts of 2048B data + 128B header */
+
+#define PIO_THRESHOLD_CEILING 4096
 
 #define PIO_WAIT_BATCH_SIZE 5
 
 /* default send context sizes */
 static struct sc_config_sizes sc_config_sizes[SC_MAX] = {
 	[SC_KERNEL] = { .size  = SCS_POOL_0,	/* even divide, pool 0 */
-			.count = SCC_PER_VL },/* one per NUMA */
-	[SC_ACK]    = { .size  = SCC_ACK_CREDITS,
+			.count = SCC_PER_VL },	/* one per NUMA */
+	[SC_ACK]    = { .size  = SCS_ACK_CREDITS,
 			.count = SCC_PER_KRCVQ },
 	[SC_USER]   = { .size  = SCS_POOL_0,	/* even divide, pool 0 */
 			.count = SCC_PER_CPU },	/* one per CPU */
+	[SC_VL15]   = { .size  = SCS_VL15_CREDITS,
+			.count = 1 },
 
 };
 
@@ -202,7 +209,8 @@ static int wildcard_to_pool(int wc)
 static const char *sc_type_names[SC_MAX] = {
 	"kernel",
 	"ack",
-	"user"
+	"user",
+	"vl15"
 };
 
 static const char *sc_type_name(int index)
@@ -230,6 +238,22 @@ int init_sc_pools_and_sizes(struct hfi1_devdata *dd)
 	int extra;
 	int i;
 
+	/*
+	 * When SDMA is enabled, kernel context pio packet size is capped by
+	 * "piothreshold". Reduce pio buffer allocation for kernel context by
+	 * setting it to a fixed size. The allocation allows 3-deep buffering
+	 * of the largest pio packets plus up to 128 bytes header, sufficient
+	 * to maintain verbs performance.
+	 *
+	 * When SDMA is disabled, keep the default pooling allocation.
+	 */
+	if (HFI1_CAP_IS_KSET(SDMA)) {
+		u16 max_pkt_size = (piothreshold < PIO_THRESHOLD_CEILING) ?
+					 piothreshold : PIO_THRESHOLD_CEILING;
+		sc_config_sizes[SC_KERNEL].size =
+			3 * (max_pkt_size + 128) / PIO_BLOCK_SIZE;
+	}
+
 	/*
 	 * Step 0:
 	 *	- copy the centipercents/absolute sizes from the pool config
@@ -311,7 +335,7 @@ int init_sc_pools_and_sizes(struct hfi1_devdata *dd)
 		if (i == SC_ACK) {
 			count = dd->n_krcv_queues;
 		} else if (i == SC_KERNEL) {
-			count = (INIT_SC_PER_VL * num_vls) + 1 /* VL15 */;
+			count = INIT_SC_PER_VL * num_vls;
 		} else if (count == SCC_PER_CPU) {
 			count = dd->num_rcv_contexts - dd->n_krcv_queues;
 		} else if (count < 0) {
@@ -596,7 +620,7 @@ u32 sc_mtu_to_threshold(struct send_context *sc, u32 mtu, u32 hdrqentsize)
  * Return value is what to write into the CSR: trigger return when
  * unreturned credits pass this count.
  */
-static u32 sc_percent_to_threshold(struct send_context *sc, u32 percent)
+u32 sc_percent_to_threshold(struct send_context *sc, u32 percent)
 {
 	return (sc->credits * percent) / 100;
 }
@@ -790,7 +814,10 @@ struct send_context *sc_alloc(struct hfi1_devdata *dd, int type,
 	 * For Ack contexts, set a threshold for half the credits.
 	 * For User contexts use the given percentage.  This has been
 	 * sanitized on driver start-up.
-	 * For Kernel contexts, use the default MTU plus a header.
+	 * For Kernel contexts, use the default MTU plus a header
+	 * or half the credits, whichever is smaller. This should
+	 * work for both the 3-deep buffering allocation and the
+	 * pooling allocation.
 	 */
 	if (type == SC_ACK) {
 		thresh = sc_percent_to_threshold(sc, 50);
@@ -798,7 +825,9 @@ struct send_context *sc_alloc(struct hfi1_devdata *dd, int type,
 		thresh = sc_percent_to_threshold(sc,
 						 user_credit_return_threshold);
 	} else { /* kernel */
-		thresh = sc_mtu_to_threshold(sc, hfi1_max_mtu, hdrqentsize);
+		thresh = min(sc_percent_to_threshold(sc, 50),
+			     sc_mtu_to_threshold(sc, hfi1_max_mtu,
+						 hdrqentsize));
 	}
 	reg = thresh << SC(CREDIT_CTRL_THRESHOLD_SHIFT);
 	/* add in early return */
@@ -1531,7 +1560,8 @@ static void sc_piobufavail(struct send_context *sc)
 	unsigned long flags;
 	unsigned i, n = 0;
 
-	if (dd->send_contexts[sc->sw_index].type != SC_KERNEL)
+	if (dd->send_contexts[sc->sw_index].type != SC_KERNEL &&
+	    dd->send_contexts[sc->sw_index].type != SC_VL15)
 		return;
 	list = &sc->piowait;
 	/*
@@ -1900,7 +1930,7 @@ int init_pervl_scs(struct hfi1_devdata *dd)
 	u32 ctxt;
 	struct hfi1_pportdata *ppd = dd->pport;
 
-	dd->vld[15].sc = sc_alloc(dd, SC_KERNEL,
+	dd->vld[15].sc = sc_alloc(dd, SC_VL15,
 				  dd->rcd[0]->rcvhdrqentsize, dd->node);
 	if (!dd->vld[15].sc)
 		goto nomem;

drivers/staging/rdma/hfi1/pio.h

@@ -51,7 +51,8 @@
 #define SC_KERNEL 0
 #define SC_ACK    1
 #define SC_USER   2
-#define SC_MAX    3
+#define SC_VL15   3
+#define SC_MAX    4
 
 /* invalid send context index */
 #define INVALID_SCI 0xff
@@ -293,6 +294,7 @@ void sc_group_release_update(struct hfi1_devdata *dd, u32 hw_context);
 void sc_add_credit_return_intr(struct send_context *sc);
 void sc_del_credit_return_intr(struct send_context *sc);
 void sc_set_cr_threshold(struct send_context *sc, u32 new_threshold);
+u32 sc_percent_to_threshold(struct send_context *sc, u32 percent);
 u32 sc_mtu_to_threshold(struct send_context *sc, u32 mtu, u32 hdrqentsize);
 void hfi1_sc_wantpiobuf_intr(struct send_context *sc, u32 needint);
 void sc_wait(struct hfi1_devdata *dd);

drivers/staging/rdma/hfi1/platform.c

@@ -114,21 +114,11 @@ static int qual_power(struct hfi1_pportdata *ppd)
 	if (ret)
 		return ret;
 
-	if (QSFP_HIGH_PWR(cache[QSFP_MOD_PWR_OFFS]) != 4)
-		cable_power_class = QSFP_HIGH_PWR(cache[QSFP_MOD_PWR_OFFS]);
-	else
-		cable_power_class = QSFP_PWR(cache[QSFP_MOD_PWR_OFFS]);
+	cable_power_class = get_qsfp_power_class(cache[QSFP_MOD_PWR_OFFS]);
 
-	if (cable_power_class <= 3 && cable_power_class > (power_class_max - 1))
+	if (cable_power_class > power_class_max)
 		ppd->offline_disabled_reason =
 			HFI1_ODR_MASK(OPA_LINKDOWN_REASON_POWER_POLICY);
-	else if (cable_power_class > 4 && cable_power_class > (power_class_max))
-		ppd->offline_disabled_reason =
-			HFI1_ODR_MASK(OPA_LINKDOWN_REASON_POWER_POLICY);
-	/*
-	 * cable_power_class will never have value 4 as this simply
-	 * means the high power settings are unused
-	 */
 
 	if (ppd->offline_disabled_reason ==
 			HFI1_ODR_MASK(OPA_LINKDOWN_REASON_POWER_POLICY)) {
@@ -173,12 +163,9 @@ static int set_qsfp_high_power(struct hfi1_pportdata *ppd)
 	u8 *cache = ppd->qsfp_info.cache;
 	int ret;
 
-	if (QSFP_HIGH_PWR(cache[QSFP_MOD_PWR_OFFS]) != 4)
-		cable_power_class = QSFP_HIGH_PWR(cache[QSFP_MOD_PWR_OFFS]);
-	else
-		cable_power_class = QSFP_PWR(cache[QSFP_MOD_PWR_OFFS]);
+	cable_power_class = get_qsfp_power_class(cache[QSFP_MOD_PWR_OFFS]);
 
-	if (cable_power_class) {
+	if (cable_power_class > QSFP_POWER_CLASS_1) {
 		power_ctrl_byte = cache[QSFP_PWR_CTRL_BYTE_OFFS];
 
 		power_ctrl_byte |= 1;
@@ -190,8 +177,7 @@ static int set_qsfp_high_power(struct hfi1_pportdata *ppd)
 		if (ret != 1)
 			return -EIO;
 
-		if (cable_power_class > 3) {
-			/* > power class 4*/
+		if (cable_power_class > QSFP_POWER_CLASS_4) {
 			power_ctrl_byte |= (1 << 2);
 			ret = qsfp_write(ppd, ppd->dd->hfi1_id,
 					 QSFP_PWR_CTRL_BYTE_OFFS,
@@ -212,12 +198,21 @@ static void apply_rx_cdr(struct hfi1_pportdata *ppd,
 {
 	u32 rx_preset;
 	u8 *cache = ppd->qsfp_info.cache;
+	int cable_power_class;
 
 	if (!((cache[QSFP_MOD_PWR_OFFS] & 0x4) &&
 	      (cache[QSFP_CDR_INFO_OFFS] & 0x40)))
 		return;
 
-	/* rx_preset preset to zero to catch error */
+	/* RX CDR present, bypass supported */
+	cable_power_class = get_qsfp_power_class(cache[QSFP_MOD_PWR_OFFS]);
+
+	if (cable_power_class <= QSFP_POWER_CLASS_3) {
+		/* Power class <= 3, ignore config & turn RX CDR on */
+		*cdr_ctrl_byte |= 0xF;
+		return;
+	}
+
 	get_platform_config_field(
 		ppd->dd, PLATFORM_CONFIG_RX_PRESET_TABLE,
 		rx_preset_index, RX_PRESET_TABLE_QSFP_RX_CDR_APPLY,
@@ -250,15 +245,25 @@ static void apply_rx_cdr(struct hfi1_pportdata *ppd,
 
 static void apply_tx_cdr(struct hfi1_pportdata *ppd,
 			 u32 tx_preset_index,
-			 u8 *ctr_ctrl_byte)
+			 u8 *cdr_ctrl_byte)
 {
 	u32 tx_preset;
 	u8 *cache = ppd->qsfp_info.cache;
+	int cable_power_class;
 
 	if (!((cache[QSFP_MOD_PWR_OFFS] & 0x8) &&
 	      (cache[QSFP_CDR_INFO_OFFS] & 0x80)))
 		return;
 
+	/* TX CDR present, bypass supported */
+	cable_power_class = get_qsfp_power_class(cache[QSFP_MOD_PWR_OFFS]);
+
+	if (cable_power_class <= QSFP_POWER_CLASS_3) {
+		/* Power class <= 3, ignore config & turn TX CDR on */
+		*cdr_ctrl_byte |= 0xF0;
+		return;
+	}
+
 	get_platform_config_field(
 		ppd->dd,
 		PLATFORM_CONFIG_TX_PRESET_TABLE, tx_preset_index,
@@ -282,10 +287,10 @@ static void apply_tx_cdr(struct hfi1_pportdata *ppd,
 			(tx_preset << 2) | (tx_preset << 3));
 
 	if (tx_preset)
-		*ctr_ctrl_byte |= (tx_preset << 4);
+		*cdr_ctrl_byte |= (tx_preset << 4);
 	else
 		/* Preserve current/determined RX CDR status */
-		*ctr_ctrl_byte &= ((tx_preset << 4) | 0xF);
+		*cdr_ctrl_byte &= ((tx_preset << 4) | 0xF);
 }
 
 static void apply_cdr_settings(
@@ -598,6 +603,7 @@ static void apply_tunings(
 		       "Applying TX settings");
 }
 
+/* Must be holding the QSFP i2c resource */
 static int tune_active_qsfp(struct hfi1_pportdata *ppd, u32 *ptr_tx_preset,
 			    u32 *ptr_rx_preset, u32 *ptr_total_atten)
 {
@@ -605,26 +611,19 @@ static int tune_active_qsfp(struct hfi1_pportdata *ppd, u32 *ptr_tx_preset,
 	u16 lss = ppd->link_speed_supported, lse = ppd->link_speed_enabled;
 	u8 *cache = ppd->qsfp_info.cache;
 
-	ret = acquire_chip_resource(ppd->dd, qsfp_resource(ppd->dd), QSFP_WAIT);
-	if (ret) {
-		dd_dev_err(ppd->dd, "%s: hfi%d: cannot lock i2c chain\n",
-			   __func__, (int)ppd->dd->hfi1_id);
-		return ret;
-	}
-
 	ppd->qsfp_info.limiting_active = 1;
 
 	ret = set_qsfp_tx(ppd, 0);
 	if (ret)
-		goto bail_unlock;
+		return ret;
 
 	ret = qual_power(ppd);
 	if (ret)
-		goto bail_unlock;
+		return ret;
 
 	ret = qual_bitrate(ppd);
 	if (ret)
-		goto bail_unlock;
+		return ret;
 
 	if (ppd->qsfp_info.reset_needed) {
 		reset_qsfp(ppd);
@@ -636,7 +635,7 @@ static int tune_active_qsfp(struct hfi1_pportdata *ppd, u32 *ptr_tx_preset,
 
 	ret = set_qsfp_high_power(ppd);
 	if (ret)
-		goto bail_unlock;
+		return ret;
 
 	if (cache[QSFP_EQ_INFO_OFFS] & 0x4) {
 		ret = get_platform_config_field(
@@ -646,7 +645,7 @@ static int tune_active_qsfp(struct hfi1_pportdata *ppd, u32 *ptr_tx_preset,
 			ptr_tx_preset, 4);
 		if (ret) {
 			*ptr_tx_preset = OPA_INVALID_INDEX;
-			goto bail_unlock;
+			return ret;
 		}
 	} else {
 		ret = get_platform_config_field(
@@ -656,7 +655,7 @@ static int tune_active_qsfp(struct hfi1_pportdata *ppd, u32 *ptr_tx_preset,
 			ptr_tx_preset, 4);
 		if (ret) {
 			*ptr_tx_preset = OPA_INVALID_INDEX;
-			goto bail_unlock;
+			return ret;
 		}
 	}
 
@@ -665,7 +664,7 @@ static int tune_active_qsfp(struct hfi1_pportdata *ppd, u32 *ptr_tx_preset,
 		PORT_TABLE_RX_PRESET_IDX, ptr_rx_preset, 4);
 	if (ret) {
 		*ptr_rx_preset = OPA_INVALID_INDEX;
-		goto bail_unlock;
+		return ret;
 	}
 
 	if ((lss & OPA_LINK_SPEED_25G) && (lse & OPA_LINK_SPEED_25G))
@@ -685,8 +684,6 @@ static int tune_active_qsfp(struct hfi1_pportdata *ppd, u32 *ptr_tx_preset,
 
 	ret = set_qsfp_tx(ppd, 1);
 
-bail_unlock:
-	release_chip_resource(ppd->dd, qsfp_resource(ppd->dd));
 	return ret;
 }
 
@@ -833,12 +830,22 @@ void tune_serdes(struct hfi1_pportdata *ppd)
 			total_atten = platform_atten + remote_atten;
 
 			tuning_method = OPA_PASSIVE_TUNING;
-		} else
+		} else {
 			ppd->offline_disabled_reason =
 			     HFI1_ODR_MASK(OPA_LINKDOWN_REASON_CHASSIS_CONFIG);
+			goto bail;
+		}
 		break;
 	case PORT_TYPE_QSFP:
 		if (qsfp_mod_present(ppd)) {
+			ret = acquire_chip_resource(ppd->dd,
+						    qsfp_resource(ppd->dd),
+						    QSFP_WAIT);
+			if (ret) {
+				dd_dev_err(ppd->dd, "%s: hfi%d: cannot lock i2c chain\n",
+					   __func__, (int)ppd->dd->hfi1_id);
+				goto bail;
+			}
 			refresh_qsfp_cache(ppd, &ppd->qsfp_info);
 
 			if (ppd->qsfp_info.cache_valid) {
@@ -853,21 +860,23 @@ void tune_serdes(struct hfi1_pportdata *ppd)
 				 * update the cache to reflect the changes
 				 */
 				refresh_qsfp_cache(ppd, &ppd->qsfp_info);
-				if (ret)
-					goto bail;
-
 				limiting_active =
 						ppd->qsfp_info.limiting_active;
 			} else {
 				dd_dev_err(dd,
 					   "%s: Reading QSFP memory failed\n",
 					   __func__);
-				goto bail;
+				ret = -EINVAL; /* a fail indication */
 			}
-		} else
+			release_chip_resource(ppd->dd, qsfp_resource(ppd->dd));
+			if (ret)
+				goto bail;
+		} else {
 			ppd->offline_disabled_reason =
 			   HFI1_ODR_MASK(
 				OPA_LINKDOWN_REASON_LOCAL_MEDIA_NOT_INSTALLED);
+			goto bail;
+		}
 		break;
 	default:
 		dd_dev_info(ppd->dd, "%s: Unknown port type\n", __func__);

drivers/staging/rdma/hfi1/qp.c

@@ -167,8 +167,12 @@ static inline int opa_mtu_enum_to_int(int mtu)
  */
 static inline int verbs_mtu_enum_to_int(struct ib_device *dev, enum ib_mtu mtu)
 {
-	int val = opa_mtu_enum_to_int((int)mtu);
+	int val;
 
+	/* Constraining 10KB packets to 8KB packets */
+	if (mtu == (enum ib_mtu)OPA_MTU_10240)
+		mtu = OPA_MTU_8192;
+	val = opa_mtu_enum_to_int((int)mtu);
 	if (val > 0)
 		return val;
 	return ib_mtu_enum_to_int(mtu);

drivers/staging/rdma/hfi1/qsfp.c

@@ -96,7 +96,7 @@ int i2c_write(struct hfi1_pportdata *ppd, u32 target, int i2c_addr, int offset,
 {
 	int ret;
 
-	if (!check_chip_resource(ppd->dd, qsfp_resource(ppd->dd), __func__))
+	if (!check_chip_resource(ppd->dd, i2c_target(target), __func__))
 		return -EACCES;
 
 	/* make sure the TWSI bus is in a sane state */
@@ -162,7 +162,7 @@ int i2c_read(struct hfi1_pportdata *ppd, u32 target, int i2c_addr, int offset,
 {
 	int ret;
 
-	if (!check_chip_resource(ppd->dd, qsfp_resource(ppd->dd), __func__))
+	if (!check_chip_resource(ppd->dd, i2c_target(target), __func__))
 		return -EACCES;
 
 	/* make sure the TWSI bus is in a sane state */
@@ -192,7 +192,7 @@ int qsfp_write(struct hfi1_pportdata *ppd, u32 target, int addr, void *bp,
 	int ret;
 	u8 page;
 
-	if (!check_chip_resource(ppd->dd, qsfp_resource(ppd->dd), __func__))
+	if (!check_chip_resource(ppd->dd, i2c_target(target), __func__))
 		return -EACCES;
 
 	/* make sure the TWSI bus is in a sane state */
@@ -276,7 +276,7 @@ int qsfp_read(struct hfi1_pportdata *ppd, u32 target, int addr, void *bp,
 	int ret;
 	u8 page;
 
-	if (!check_chip_resource(ppd->dd, qsfp_resource(ppd->dd), __func__))
+	if (!check_chip_resource(ppd->dd, i2c_target(target), __func__))
 		return -EACCES;
 
 	/* make sure the TWSI bus is in a sane state */
@@ -355,6 +355,8 @@ int one_qsfp_read(struct hfi1_pportdata *ppd, u32 target, int addr, void *bp,
  * The calls to qsfp_{read,write} in this function correctly handle the
  * address map difference between this mapping and the mapping implemented
  * by those functions
+ *
+ * The caller must be holding the QSFP i2c chain resource.
  */
 int refresh_qsfp_cache(struct hfi1_pportdata *ppd, struct qsfp_data *cp)
 {
@@ -371,13 +373,9 @@ int refresh_qsfp_cache(struct hfi1_pportdata *ppd, struct qsfp_data *cp)
 
 	if (!qsfp_mod_present(ppd)) {
 		ret = -ENODEV;
-		goto bail_no_release;
+		goto bail;
 	}
 
-	ret = acquire_chip_resource(ppd->dd, qsfp_resource(ppd->dd), QSFP_WAIT);
-	if (ret)
-		goto bail_no_release;
-
 	ret = qsfp_read(ppd, target, 0, cache, QSFP_PAGESIZE);
 	if (ret != QSFP_PAGESIZE) {
 		dd_dev_info(ppd->dd,
@@ -440,8 +438,6 @@ int refresh_qsfp_cache(struct hfi1_pportdata *ppd, struct qsfp_data *cp)
 		}
 	}
 
-	release_chip_resource(ppd->dd, qsfp_resource(ppd->dd));
-
 	spin_lock_irqsave(&ppd->qsfp_info.qsfp_lock, flags);
 	ppd->qsfp_info.cache_valid = 1;
 	ppd->qsfp_info.cache_refresh_required = 0;
@@ -450,8 +446,6 @@ int refresh_qsfp_cache(struct hfi1_pportdata *ppd, struct qsfp_data *cp)
 	return 0;
 
 bail:
-	release_chip_resource(ppd->dd, qsfp_resource(ppd->dd));
-bail_no_release:
 	memset(cache, 0, (QSFP_MAX_NUM_PAGES * 128));
 	return ret;
 }
@@ -466,7 +460,28 @@ const char * const hfi1_qsfp_devtech[16] = {
 #define QSFP_DUMP_CHUNK 16 /* Holds longest string */
 #define QSFP_DEFAULT_HDR_CNT 224
 
-static const char *pwr_codes = "1.5W2.0W2.5W3.5W";
+#define QSFP_PWR(pbyte) (((pbyte) >> 6) & 3)
+#define QSFP_HIGH_PWR(pbyte) ((pbyte) & 3)
+/* For use with QSFP_HIGH_PWR macro */
+#define QSFP_HIGH_PWR_UNUSED	0 /* Bits [1:0] = 00 implies low power module */
+
+/*
+ * Takes power class byte [Page 00 Byte 129] in SFF 8636
+ * Returns power class as integer (1 through 7, per SFF 8636 rev 2.4)
+ */
+int get_qsfp_power_class(u8 power_byte)
+{
+	if (QSFP_HIGH_PWR(power_byte) == QSFP_HIGH_PWR_UNUSED)
+		/* power classes count from 1, their bit encodings from 0 */
+		return (QSFP_PWR(power_byte) + 1);
+	/*
+	 * 00 in the high power classes stands for unused, bringing
+	 * balance to the off-by-1 offset above, we add 4 here to
+	 * account for the difference between the low and high power
+	 * groups
+	 */
+	return (QSFP_HIGH_PWR(power_byte) + 4);
+}
 
 int qsfp_mod_present(struct hfi1_pportdata *ppd)
 {
@@ -537,6 +552,16 @@ int get_cable_info(struct hfi1_devdata *dd, u32 port_num, u32 addr, u32 len,
 	return ret;
 }
 
+static const char *pwr_codes[8] = {"N/AW",
+				  "1.5W",
+				  "2.0W",
+				  "2.5W",
+				  "3.5W",
+				  "4.0W",
+				  "4.5W",
+				  "5.0W"
+				 };
+
 int qsfp_dump(struct hfi1_pportdata *ppd, char *buf, int len)
 {
 	u8 *cache = &ppd->qsfp_info.cache[0];
@@ -546,6 +571,7 @@ int qsfp_dump(struct hfi1_pportdata *ppd, char *buf, int len)
 	int bidx = 0;
 	u8 *atten = &cache[QSFP_ATTEN_OFFS];
 	u8 *vendor_oui = &cache[QSFP_VOUI_OFFS];
+	u8 power_byte = 0;
 
 	sofar = 0;
 	lenstr[0] = ' ';
@@ -555,9 +581,9 @@ int qsfp_dump(struct hfi1_pportdata *ppd, char *buf, int len)
 		if (QSFP_IS_CU(cache[QSFP_MOD_TECH_OFFS]))
 			sprintf(lenstr, "%dM ", cache[QSFP_MOD_LEN_OFFS]);
 
+		power_byte = cache[QSFP_MOD_PWR_OFFS];
 		sofar += scnprintf(buf + sofar, len - sofar, "PWR:%.3sW\n",
-				pwr_codes +
-				(QSFP_PWR(cache[QSFP_MOD_PWR_OFFS]) * 4));
+				pwr_codes[get_qsfp_power_class(power_byte)]);
 
 		sofar += scnprintf(buf + sofar, len - sofar, "TECH:%s%s\n",
 				lenstr,

drivers/staging/rdma/hfi1/qsfp.h

@@ -82,8 +82,9 @@
 /* Byte 128 is Identifier: must be 0x0c for QSFP, or 0x0d for QSFP+ */
 #define QSFP_MOD_ID_OFFS 128
 /*
- * Byte 129 is "Extended Identifier". We only care about D7,D6: Power class
- *  0:1.5W, 1:2.0W, 2:2.5W, 3:3.5W
+ * Byte 129 is "Extended Identifier".
+ * For bits [7:6]: 0:1.5W, 1:2.0W, 2:2.5W, 3:3.5W
+ * For bits [1:0]: 0:Unused, 1:4W, 2:4.5W, 3:5W
  */
 #define QSFP_MOD_PWR_OFFS 129
 /* Byte 130 is Connector type. Not Intel req'd */
@@ -190,6 +191,9 @@ extern const char *const hfi1_qsfp_devtech[16];
 #define QSFP_HIGH_BIAS_WARNING		0x22
 #define QSFP_LOW_BIAS_WARNING		0x11
 
+#define QSFP_ATTEN_SDR(attenarray) (attenarray[0])
+#define QSFP_ATTEN_DDR(attenarray) (attenarray[1])
+
 /*
  * struct qsfp_data encapsulates state of QSFP device for one port.
  * it will be part of port-specific data if a board supports QSFP.
@@ -201,12 +205,6 @@ extern const char *const hfi1_qsfp_devtech[16];
  * and let the qsfp_lock arbitrate access to common resources.
  *
  */
-
-#define QSFP_PWR(pbyte) (((pbyte) >> 6) & 3)
-#define QSFP_HIGH_PWR(pbyte) (((pbyte) & 3) | 4)
-#define QSFP_ATTEN_SDR(attenarray) (attenarray[0])
-#define QSFP_ATTEN_DDR(attenarray) (attenarray[1])
-
 struct qsfp_data {
 	/* Helps to find our way */
 	struct hfi1_pportdata *ppd;
@@ -223,6 +221,7 @@ struct qsfp_data {
 
 int refresh_qsfp_cache(struct hfi1_pportdata *ppd,
 		       struct qsfp_data *cp);
+int get_qsfp_power_class(u8 power_byte);
 int qsfp_mod_present(struct hfi1_pportdata *ppd);
 int get_cable_info(struct hfi1_devdata *dd, u32 port_num, u32 addr,
 		   u32 len, u8 *data);

drivers/staging/rdma/hfi1/rc.c

@@ -1497,7 +1497,7 @@ static int do_rc_ack(struct rvt_qp *qp, u32 aeth, u32 psn, int opcode,
 		/* Ignore reserved NAK codes. */
 		goto bail_stop;
 	}
-	return ret;
+	/* cannot be reached  */
 bail_stop:
 	hfi1_stop_rc_timers(qp);
 	return ret;
@@ -2021,8 +2021,6 @@ void process_becn(struct hfi1_pportdata *ppd, u8 sl, u16 rlid, u32 lqpn,
 	if (sl >= OPA_MAX_SLS)
 		return;
 
-	cca_timer = &ppd->cca_timer[sl];
-
 	cc_state = get_cc_state(ppd);
 
 	if (!cc_state)
@@ -2041,6 +2039,7 @@ void process_becn(struct hfi1_pportdata *ppd, u8 sl, u16 rlid, u32 lqpn,
 
 	spin_lock_irqsave(&ppd->cca_timer_lock, flags);
 
+	cca_timer = &ppd->cca_timer[sl];
 	if (cca_timer->ccti < ccti_limit) {
 		if (cca_timer->ccti + ccti_incr <= ccti_limit)
 			cca_timer->ccti += ccti_incr;
@@ -2049,8 +2048,6 @@ void process_becn(struct hfi1_pportdata *ppd, u8 sl, u16 rlid, u32 lqpn,
 		set_link_ipg(ppd);
 	}
 
-	spin_unlock_irqrestore(&ppd->cca_timer_lock, flags);
-
 	ccti = cca_timer->ccti;
 
 	if (!hrtimer_active(&cca_timer->hrtimer)) {
@@ -2061,6 +2058,8 @@ void process_becn(struct hfi1_pportdata *ppd, u8 sl, u16 rlid, u32 lqpn,
 			      HRTIMER_MODE_REL);
 	}
 
+	spin_unlock_irqrestore(&ppd->cca_timer_lock, flags);
+
 	if ((trigger_threshold != 0) && (ccti >= trigger_threshold))
 		log_cca_event(ppd, sl, rlid, lqpn, rqpn, svc_type);
 }

drivers/staging/rdma/hfi1/ruc.c

@@ -831,7 +831,6 @@ void hfi1_do_send(struct rvt_qp *qp)
 	struct hfi1_pkt_state ps;
 	struct hfi1_qp_priv *priv = qp->priv;
 	int (*make_req)(struct rvt_qp *qp, struct hfi1_pkt_state *ps);
-	unsigned long flags;
 	unsigned long timeout;
 	unsigned long timeout_int;
 	int cpu;
@@ -866,11 +865,11 @@ void hfi1_do_send(struct rvt_qp *qp)
 		timeout_int = SEND_RESCHED_TIMEOUT;
 	}
 
-	spin_lock_irqsave(&qp->s_lock, flags);
+	spin_lock_irqsave(&qp->s_lock, ps.flags);
 
 	/* Return if we are already busy processing a work request. */
 	if (!hfi1_send_ok(qp)) {
-		spin_unlock_irqrestore(&qp->s_lock, flags);
+		spin_unlock_irqrestore(&qp->s_lock, ps.flags);
 		return;
 	}
 
@@ -884,7 +883,7 @@ void hfi1_do_send(struct rvt_qp *qp)
 	do {
 		/* Check for a constructed packet to be sent. */
 		if (qp->s_hdrwords != 0) {
-			spin_unlock_irqrestore(&qp->s_lock, flags);
+			spin_unlock_irqrestore(&qp->s_lock, ps.flags);
 			/*
 			 * If the packet cannot be sent now, return and
 			 * the send tasklet will be woken up later.
@@ -897,11 +896,14 @@ void hfi1_do_send(struct rvt_qp *qp)
 			if (unlikely(time_after(jiffies, timeout))) {
 				if (workqueue_congested(cpu,
 							ps.ppd->hfi1_wq)) {
-					spin_lock_irqsave(&qp->s_lock, flags);
+					spin_lock_irqsave(
+						&qp->s_lock,
+						ps.flags);
 					qp->s_flags &= ~RVT_S_BUSY;
 					hfi1_schedule_send(qp);
-					spin_unlock_irqrestore(&qp->s_lock,
-							       flags);
+					spin_unlock_irqrestore(
+						&qp->s_lock,
+						ps.flags);
 					this_cpu_inc(
 						*ps.ppd->dd->send_schedule);
 					return;
@@ -913,11 +915,11 @@ void hfi1_do_send(struct rvt_qp *qp)
 				}
 				timeout = jiffies + (timeout_int) / 8;
 			}
-			spin_lock_irqsave(&qp->s_lock, flags);
+			spin_lock_irqsave(&qp->s_lock, ps.flags);
 		}
 	} while (make_req(qp, &ps));
 
-	spin_unlock_irqrestore(&qp->s_lock, flags);
+	spin_unlock_irqrestore(&qp->s_lock, ps.flags);
 }
 
 /*

drivers/staging/rdma/hfi1/sysfs.c

@@ -84,7 +84,7 @@ static ssize_t read_cc_table_bin(struct file *filp, struct kobject *kobj,
 		rcu_read_unlock();
 		return -EINVAL;
 	}
-	memcpy(buf, &cc_state->cct, count);
+	memcpy(buf, (void *)&cc_state->cct + pos, count);
 	rcu_read_unlock();
 
 	return count;
@@ -131,7 +131,7 @@ static ssize_t read_cc_setting_bin(struct file *filp, struct kobject *kobj,
 		rcu_read_unlock();
 		return -EINVAL;
 	}
-	memcpy(buf, &cc_state->cong_setting, count);
+	memcpy(buf, (void *)&cc_state->cong_setting + pos, count);
 	rcu_read_unlock();
 
 	return count;

drivers/staging/rdma/hfi1/ud.c

@@ -322,7 +322,7 @@ int hfi1_make_ud_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps)
 			     (lid == ppd->lid ||
 			      (lid == be16_to_cpu(IB_LID_PERMISSIVE) &&
 			      qp->ibqp.qp_type == IB_QPT_GSI)))) {
-			unsigned long flags;
+			unsigned long tflags = ps->flags;
 			/*
 			 * If DMAs are in progress, we can't generate
 			 * a completion for the loopback packet since
@@ -335,10 +335,10 @@ int hfi1_make_ud_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps)
 				goto bail;
 			}
 			qp->s_cur = next_cur;
-			local_irq_save(flags);
-			spin_unlock_irqrestore(&qp->s_lock, flags);
+			spin_unlock_irqrestore(&qp->s_lock, tflags);
 			ud_loopback(qp, wqe);
-			spin_lock_irqsave(&qp->s_lock, flags);
+			spin_lock_irqsave(&qp->s_lock, tflags);
+			ps->flags = tflags;
 			hfi1_send_complete(qp, wqe, IB_WC_SUCCESS);
 			goto done_free_tx;
 		}

drivers/staging/rdma/hfi1/user_exp_rcv.c

@@ -399,8 +399,11 @@ int hfi1_user_exp_rcv_setup(struct file *fp, struct hfi1_tid_info *tinfo)
 	 * pages, accept the amount pinned so far and program only that.
 	 * User space knows how to deal with partially programmed buffers.
 	 */
-	if (!hfi1_can_pin_pages(dd, fd->tid_n_pinned, npages))
-		return -ENOMEM;
+	if (!hfi1_can_pin_pages(dd, fd->tid_n_pinned, npages)) {
+		ret = -ENOMEM;
+		goto bail;
+	}
+
 	pinned = hfi1_acquire_user_pages(vaddr, npages, true, pages);
 	if (pinned <= 0) {
 		ret = pinned;

drivers/staging/rdma/hfi1/user_sdma.c

@@ -180,6 +180,8 @@ struct user_sdma_iovec {
 	u64 offset;
 };
 
+#define SDMA_CACHE_NODE_EVICT BIT(0)
+
 struct sdma_mmu_node {
 	struct mmu_rb_node rb;
 	struct list_head list;
@@ -187,6 +189,7 @@ struct sdma_mmu_node {
 	atomic_t refcount;
 	struct page **pages;
 	unsigned npages;
+	unsigned long flags;
 };
 
 struct user_sdma_request {
@@ -597,6 +600,13 @@ int hfi1_user_sdma_process_request(struct file *fp, struct iovec *iovec,
 		goto free_req;
 	}
 
+	/* Checking P_KEY for requests from user-space */
+	if (egress_pkey_check(dd->pport, req->hdr.lrh, req->hdr.bth, sc,
+			      PKEY_CHECK_INVALID)) {
+		ret = -EINVAL;
+		goto free_req;
+	}
+
 	/*
 	 * Also should check the BTH.lnh. If it says the next header is GRH then
 	 * the RXE parsing will be off and will land in the middle of the KDETH
@@ -1030,27 +1040,29 @@ static inline int num_user_pages(const struct iovec *iov)
 	return 1 + ((epage - spage) >> PAGE_SHIFT);
 }
 
-/* Caller must hold pq->evict_lock */
 static u32 sdma_cache_evict(struct hfi1_user_sdma_pkt_q *pq, u32 npages)
 {
 	u32 cleared = 0;
 	struct sdma_mmu_node *node, *ptr;
+	struct list_head to_evict = LIST_HEAD_INIT(to_evict);
 
+	spin_lock(&pq->evict_lock);
 	list_for_each_entry_safe_reverse(node, ptr, &pq->evict, list) {
 		/* Make sure that no one is still using the node. */
 		if (!atomic_read(&node->refcount)) {
-			/*
-			 * Need to use the page count now as the remove callback
-			 * will free the node.
-			 */
+			set_bit(SDMA_CACHE_NODE_EVICT, &node->flags);
+			list_del_init(&node->list);
+			list_add(&node->list, &to_evict);
 			cleared += node->npages;
-			spin_unlock(&pq->evict_lock);
-			hfi1_mmu_rb_remove(&pq->sdma_rb_root, &node->rb);
-			spin_lock(&pq->evict_lock);
 			if (cleared >= npages)
 				break;
 		}
 	}
+	spin_unlock(&pq->evict_lock);
+
+	list_for_each_entry_safe(node, ptr, &to_evict, list)
+		hfi1_mmu_rb_remove(&pq->sdma_rb_root, &node->rb);
+
 	return cleared;
 }
 
@@ -1062,9 +1074,9 @@ static int pin_vector_pages(struct user_sdma_request *req,
 	struct sdma_mmu_node *node = NULL;
 	struct mmu_rb_node *rb_node;
 
-	rb_node = hfi1_mmu_rb_search(&pq->sdma_rb_root,
-				     (unsigned long)iovec->iov.iov_base,
-				     iovec->iov.iov_len);
+	rb_node = hfi1_mmu_rb_extract(&pq->sdma_rb_root,
+				      (unsigned long)iovec->iov.iov_base,
+				      iovec->iov.iov_len);
 	if (rb_node && !IS_ERR(rb_node))
 		node = container_of(rb_node, struct sdma_mmu_node, rb);
 	else
@@ -1076,7 +1088,6 @@ static int pin_vector_pages(struct user_sdma_request *req,
 			return -ENOMEM;
 
 		node->rb.addr = (unsigned long)iovec->iov.iov_base;
-		node->rb.len = iovec->iov.iov_len;
 		node->pq = pq;
 		atomic_set(&node->refcount, 0);
 		INIT_LIST_HEAD(&node->list);
@@ -1093,11 +1104,25 @@ static int pin_vector_pages(struct user_sdma_request *req,
 		memcpy(pages, node->pages, node->npages * sizeof(*pages));
 
 		npages -= node->npages;
+
+		/*
+		 * If rb_node is NULL, it means that this is brand new node
+		 * and, therefore not on the eviction list.
+		 * If, however, the rb_node is non-NULL, it means that the
+		 * node is already in RB tree and, therefore on the eviction
+		 * list (nodes are unconditionally inserted in the eviction
+		 * list). In that case, we have to remove the node prior to
+		 * calling the eviction function in order to prevent it from
+		 * freeing this node.
+		 */
+		if (rb_node) {
+			spin_lock(&pq->evict_lock);
+			list_del_init(&node->list);
+			spin_unlock(&pq->evict_lock);
+		}
 retry:
 		if (!hfi1_can_pin_pages(pq->dd, pq->n_locked, npages)) {
-			spin_lock(&pq->evict_lock);
 			cleared = sdma_cache_evict(pq, npages);
-			spin_unlock(&pq->evict_lock);
 			if (cleared >= npages)
 				goto retry;
 		}
@@ -1117,37 +1142,32 @@ static int pin_vector_pages(struct user_sdma_request *req,
 			goto bail;
 		}
 		kfree(node->pages);
+		node->rb.len = iovec->iov.iov_len;
 		node->pages = pages;
 		node->npages += pinned;
 		npages = node->npages;
 		spin_lock(&pq->evict_lock);
-		if (!rb_node)
-			list_add(&node->list, &pq->evict);
-		else
-			list_move(&node->list, &pq->evict);
+		list_add(&node->list, &pq->evict);
 		pq->n_locked += pinned;
 		spin_unlock(&pq->evict_lock);
 	}
 	iovec->pages = node->pages;
 	iovec->npages = npages;
 
-	if (!rb_node) {
-		ret = hfi1_mmu_rb_insert(&req->pq->sdma_rb_root, &node->rb);
-		if (ret) {
-			spin_lock(&pq->evict_lock);
+	ret = hfi1_mmu_rb_insert(&req->pq->sdma_rb_root, &node->rb);
+	if (ret) {
+		spin_lock(&pq->evict_lock);
+		if (!list_empty(&node->list))
 			list_del(&node->list);
-			pq->n_locked -= node->npages;
-			spin_unlock(&pq->evict_lock);
-			ret = 0;
-			goto bail;
-		}
-	} else {
-		atomic_inc(&node->refcount);
+		pq->n_locked -= node->npages;
+		spin_unlock(&pq->evict_lock);
+		goto bail;
 	}
 	return 0;
 bail:
-	if (!rb_node)
-		kfree(node);
+	if (rb_node)
+		unpin_vector_pages(current->mm, node->pages, 0, node->npages);
+	kfree(node);
 	return ret;
 }
 
@@ -1558,7 +1578,20 @@ static void sdma_rb_remove(struct rb_root *root, struct mmu_rb_node *mnode,
 		container_of(mnode, struct sdma_mmu_node, rb);
 
 	spin_lock(&node->pq->evict_lock);
-	list_del(&node->list);
+	/*
+	 * We've been called by the MMU notifier but this node has been
+	 * scheduled for eviction. The eviction function will take care
+	 * of freeing this node.
+	 * We have to take the above lock first because we are racing
+	 * against the setting of the bit in the eviction function.
+	 */
+	if (mm && test_bit(SDMA_CACHE_NODE_EVICT, &node->flags)) {
+		spin_unlock(&node->pq->evict_lock);
+		return;
+	}
+
+	if (!list_empty(&node->list))
+		list_del(&node->list);
 	node->pq->n_locked -= node->npages;
 	spin_unlock(&node->pq->evict_lock);
 

drivers/staging/rdma/hfi1/verbs.c

@@ -545,7 +545,7 @@ static inline int qp_ok(int opcode, struct hfi1_packet *packet)
 
 	if (!(ib_rvt_state_ops[packet->qp->state] & RVT_PROCESS_RECV_OK))
 		goto dropit;
-	if (((opcode & OPCODE_QP_MASK) == packet->qp->allowed_ops) ||
+	if (((opcode & RVT_OPCODE_QP_MASK) == packet->qp->allowed_ops) ||
 	    (opcode == IB_OPCODE_CNP))
 		return 1;
 dropit:
@@ -1089,16 +1089,16 @@ int hfi1_verbs_send_pio(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
 
 /*
  * egress_pkey_matches_entry - return 1 if the pkey matches ent (ent
- * being an entry from the ingress partition key table), return 0
+ * being an entry from the partition key table), return 0
  * otherwise. Use the matching criteria for egress partition keys
  * specified in the OPAv1 spec., section 9.1l.7.
  */
 static inline int egress_pkey_matches_entry(u16 pkey, u16 ent)
 {
 	u16 mkey = pkey & PKEY_LOW_15_MASK;
-	u16 ment = ent & PKEY_LOW_15_MASK;
+	u16 mentry = ent & PKEY_LOW_15_MASK;
 
-	if (mkey == ment) {
+	if (mkey == mentry) {
 		/*
 		 * If pkey[15] is set (full partition member),
 		 * is bit 15 in the corresponding table element
@@ -1111,32 +1111,32 @@ static inline int egress_pkey_matches_entry(u16 pkey, u16 ent)
 	return 0;
 }
 
-/*
- * egress_pkey_check - return 0 if hdr's pkey matches according to the
- * criteria in the OPAv1 spec., section 9.11.7.
+/**
+ * egress_pkey_check - check P_KEY of a packet
+ * @ppd:    Physical IB port data
+ * @lrh: Local route header
+ * @bth: Base transport header
+ * @sc5:    SC for packet
+ * @s_pkey_index: It will be used for look up optimization for kernel contexts
+ * only. If it is negative value, then it means user contexts is calling this
+ * function.
+ *
+ * It checks if hdr's pkey is valid.
+ *
+ * Return: 0 on success, otherwise, 1
  */
-static inline int egress_pkey_check(struct hfi1_pportdata *ppd,
-				    struct hfi1_ib_header *hdr,
-				    struct rvt_qp *qp)
+int egress_pkey_check(struct hfi1_pportdata *ppd, __be16 *lrh, __be32 *bth,
+		      u8 sc5, int8_t s_pkey_index)
 {
-	struct hfi1_qp_priv *priv = qp->priv;
-	struct hfi1_other_headers *ohdr;
 	struct hfi1_devdata *dd;
-	int i = 0;
+	int i;
 	u16 pkey;
-	u8 lnh, sc5 = priv->s_sc;
+	int is_user_ctxt_mechanism = (s_pkey_index < 0);
 
 	if (!(ppd->part_enforce & HFI1_PART_ENFORCE_OUT))
 		return 0;
 
-	/* locate the pkey within the headers */
-	lnh = be16_to_cpu(hdr->lrh[0]) & 3;
-	if (lnh == HFI1_LRH_GRH)
-		ohdr = &hdr->u.l.oth;
-	else
-		ohdr = &hdr->u.oth;
-
-	pkey = (u16)be32_to_cpu(ohdr->bth[0]);
+	pkey = (u16)be32_to_cpu(bth[0]);
 
 	/* If SC15, pkey[0:14] must be 0x7fff */
 	if ((sc5 == 0xf) && ((pkey & PKEY_LOW_15_MASK) != PKEY_LOW_15_MASK))
@@ -1146,28 +1146,37 @@ static inline int egress_pkey_check(struct hfi1_pportdata *ppd,
 	if ((pkey & PKEY_LOW_15_MASK) == 0)
 		goto bad;
 
-	/* The most likely matching pkey has index qp->s_pkey_index */
-	if (unlikely(!egress_pkey_matches_entry(pkey,
-						ppd->pkeys
-						[qp->s_pkey_index]))) {
-		/* no match - try the entire table */
-		for (; i < MAX_PKEY_VALUES; i++) {
-			if (egress_pkey_matches_entry(pkey, ppd->pkeys[i]))
-				break;
-		}
+	/*
+	 * For the kernel contexts only, if a qp is passed into the function,
+	 * the most likely matching pkey has index qp->s_pkey_index
+	 */
+	if (!is_user_ctxt_mechanism &&
+	    egress_pkey_matches_entry(pkey, ppd->pkeys[s_pkey_index])) {
+		return 0;
 	}
 
-	if (i < MAX_PKEY_VALUES)
-		return 0;
+	for (i = 0; i < MAX_PKEY_VALUES; i++) {
+		if (egress_pkey_matches_entry(pkey, ppd->pkeys[i]))
+			return 0;
+	}
 bad:
-	incr_cntr64(&ppd->port_xmit_constraint_errors);
-	dd = ppd->dd;
-	if (!(dd->err_info_xmit_constraint.status & OPA_EI_STATUS_SMASK)) {
-		u16 slid = be16_to_cpu(hdr->lrh[3]);
+	/*
+	 * For the user-context mechanism, the P_KEY check would only happen
+	 * once per SDMA request, not once per packet.  Therefore, there's no
+	 * need to increment the counter for the user-context mechanism.
+	 */
+	if (!is_user_ctxt_mechanism) {
+		incr_cntr64(&ppd->port_xmit_constraint_errors);
+		dd = ppd->dd;
+		if (!(dd->err_info_xmit_constraint.status &
+		      OPA_EI_STATUS_SMASK)) {
+			u16 slid = be16_to_cpu(lrh[3]);
 
-		dd->err_info_xmit_constraint.status |= OPA_EI_STATUS_SMASK;
-		dd->err_info_xmit_constraint.slid = slid;
-		dd->err_info_xmit_constraint.pkey = pkey;
+			dd->err_info_xmit_constraint.status |=
+				OPA_EI_STATUS_SMASK;
+			dd->err_info_xmit_constraint.slid = slid;
+			dd->err_info_xmit_constraint.pkey = pkey;
+		}
 	}
 	return 1;
 }
@@ -1227,11 +1236,26 @@ int hfi1_verbs_send(struct rvt_qp *qp, struct hfi1_pkt_state *ps)
 {
 	struct hfi1_devdata *dd = dd_from_ibdev(qp->ibqp.device);
 	struct hfi1_qp_priv *priv = qp->priv;
+	struct hfi1_other_headers *ohdr;
+	struct hfi1_ib_header *hdr;
 	send_routine sr;
 	int ret;
+	u8 lnh;
+
+	hdr = &ps->s_txreq->phdr.hdr;
+	/* locate the pkey within the headers */
+	lnh = be16_to_cpu(hdr->lrh[0]) & 3;
+	if (lnh == HFI1_LRH_GRH)
+		ohdr = &hdr->u.l.oth;
+	else
+		ohdr = &hdr->u.oth;
 
 	sr = get_send_routine(qp, ps->s_txreq);
-	ret = egress_pkey_check(dd->pport, &ps->s_txreq->phdr.hdr, qp);
+	ret = egress_pkey_check(dd->pport,
+				hdr->lrh,
+				ohdr->bth,
+				priv->s_sc,
+				qp->s_pkey_index);
 	if (unlikely(ret)) {
 		/*
 		 * The value we are returning here does not get propagated to

drivers/staging/rdma/hfi1/verbs.h

@@ -215,6 +215,7 @@ struct hfi1_pkt_state {
 	struct hfi1_ibport *ibp;
 	struct hfi1_pportdata *ppd;
 	struct verbs_txreq *s_txreq;
+	unsigned long flags;
 };
 
 #define HFI1_PSN_CREDIT  16
@@ -334,9 +335,6 @@ int hfi1_process_mad(struct ib_device *ibdev, int mad_flags, u8 port,
 #endif
 #define PSN_MODIFY_MASK 0xFFFFFF
 
-/* Number of bits to pay attention to in the opcode for checking qp type */
-#define OPCODE_QP_MASK 0xE0
-
 /*
  * Compare the lower 24 bits of the msn values.
  * Returns an integer <, ==, or > than zero.

drivers/target/target_core_transport.c

@@ -2195,7 +2195,7 @@ static void target_complete_ok_work(struct work_struct *work)
 	transport_handle_queue_full(cmd, cmd->se_dev);
 }
 
-static inline void transport_free_sgl(struct scatterlist *sgl, int nents)
+void target_free_sgl(struct scatterlist *sgl, int nents)
 {
 	struct scatterlist *sg;
 	int count;
@@ -2205,6 +2205,7 @@ static inline void transport_free_sgl(struct scatterlist *sgl, int nents)
 
 	kfree(sgl);
 }
+EXPORT_SYMBOL(target_free_sgl);
 
 static inline void transport_reset_sgl_orig(struct se_cmd *cmd)
 {
@@ -2225,7 +2226,7 @@ static inline void transport_reset_sgl_orig(struct se_cmd *cmd)
 static inline void transport_free_pages(struct se_cmd *cmd)
 {
 	if (!(cmd->se_cmd_flags & SCF_PASSTHROUGH_PROT_SG_TO_MEM_NOALLOC)) {
-		transport_free_sgl(cmd->t_prot_sg, cmd->t_prot_nents);
+		target_free_sgl(cmd->t_prot_sg, cmd->t_prot_nents);
 		cmd->t_prot_sg = NULL;
 		cmd->t_prot_nents = 0;
 	}
@@ -2236,7 +2237,7 @@ static inline void transport_free_pages(struct se_cmd *cmd)
 		 * SG_TO_MEM_NOALLOC to function with COMPARE_AND_WRITE
 		 */
 		if (cmd->se_cmd_flags & SCF_COMPARE_AND_WRITE) {
-			transport_free_sgl(cmd->t_bidi_data_sg,
+			target_free_sgl(cmd->t_bidi_data_sg,
 					   cmd->t_bidi_data_nents);
 			cmd->t_bidi_data_sg = NULL;
 			cmd->t_bidi_data_nents = 0;
@@ -2246,11 +2247,11 @@ static inline void transport_free_pages(struct se_cmd *cmd)
 	}
 	transport_reset_sgl_orig(cmd);
 
-	transport_free_sgl(cmd->t_data_sg, cmd->t_data_nents);
+	target_free_sgl(cmd->t_data_sg, cmd->t_data_nents);
 	cmd->t_data_sg = NULL;
 	cmd->t_data_nents = 0;
 
-	transport_free_sgl(cmd->t_bidi_data_sg, cmd->t_bidi_data_nents);
+	target_free_sgl(cmd->t_bidi_data_sg, cmd->t_bidi_data_nents);
 	cmd->t_bidi_data_sg = NULL;
 	cmd->t_bidi_data_nents = 0;
 }
@@ -2324,20 +2325,22 @@ EXPORT_SYMBOL(transport_kunmap_data_sg);
 
 int
 target_alloc_sgl(struct scatterlist **sgl, unsigned int *nents, u32 length,
-		 bool zero_page)
+		 bool zero_page, bool chainable)
 {
 	struct scatterlist *sg;
 	struct page *page;
 	gfp_t zero_flag = (zero_page) ? __GFP_ZERO : 0;
-	unsigned int nent;
+	unsigned int nalloc, nent;
 	int i = 0;
 
-	nent = DIV_ROUND_UP(length, PAGE_SIZE);
-	sg = kmalloc(sizeof(struct scatterlist) * nent, GFP_KERNEL);
+	nalloc = nent = DIV_ROUND_UP(length, PAGE_SIZE);
+	if (chainable)
+		nalloc++;
+	sg = kmalloc_array(nalloc, sizeof(struct scatterlist), GFP_KERNEL);
 	if (!sg)
 		return -ENOMEM;
 
-	sg_init_table(sg, nent);
+	sg_init_table(sg, nalloc);
 
 	while (length) {
 		u32 page_len = min_t(u32, length, PAGE_SIZE);
@@ -2361,6 +2364,7 @@ target_alloc_sgl(struct scatterlist **sgl, unsigned int *nents, u32 length,
 	kfree(sg);
 	return -ENOMEM;
 }
+EXPORT_SYMBOL(target_alloc_sgl);
 
 /*
  * Allocate any required resources to execute the command.  For writes we
@@ -2376,7 +2380,7 @@ transport_generic_new_cmd(struct se_cmd *cmd)
 	if (cmd->prot_op != TARGET_PROT_NORMAL &&
 	    !(cmd->se_cmd_flags & SCF_PASSTHROUGH_PROT_SG_TO_MEM_NOALLOC)) {
 		ret = target_alloc_sgl(&cmd->t_prot_sg, &cmd->t_prot_nents,
-				       cmd->prot_length, true);
+				       cmd->prot_length, true, false);
 		if (ret < 0)
 			return TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
 	}
@@ -2401,13 +2405,13 @@ transport_generic_new_cmd(struct se_cmd *cmd)
 
 			ret = target_alloc_sgl(&cmd->t_bidi_data_sg,
 					       &cmd->t_bidi_data_nents,
-					       bidi_length, zero_flag);
+					       bidi_length, zero_flag, false);
 			if (ret < 0)
 				return TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
 		}
 
 		ret = target_alloc_sgl(&cmd->t_data_sg, &cmd->t_data_nents,
-				       cmd->data_length, zero_flag);
+				       cmd->data_length, zero_flag, false);
 		if (ret < 0)
 			return TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
 	} else if ((cmd->se_cmd_flags & SCF_COMPARE_AND_WRITE) &&
@@ -2421,7 +2425,7 @@ transport_generic_new_cmd(struct se_cmd *cmd)
 
 		ret = target_alloc_sgl(&cmd->t_bidi_data_sg,
 				       &cmd->t_bidi_data_nents,
-				       caw_length, zero_flag);
+				       caw_length, zero_flag, false);
 		if (ret < 0)
 			return TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
 	}

drivers/target/target_core_xcopy.c

@@ -563,7 +563,7 @@ static int target_xcopy_setup_pt_cmd(
 
 	if (alloc_mem) {
 		rc = target_alloc_sgl(&cmd->t_data_sg, &cmd->t_data_nents,
-				      cmd->data_length, false);
+				      cmd->data_length, false, false);
 		if (rc < 0) {
 			ret = rc;
 			goto out;

include/linux/mlx5/cq.h

@@ -53,6 +53,11 @@ struct mlx5_core_cq {
 	unsigned		arm_sn;
 	struct mlx5_rsc_debug	*dbg;
 	int			pid;
+	struct {
+		struct list_head list;
+		void (*comp)(struct mlx5_core_cq *);
+		void		*priv;
+	} tasklet_ctx;
 };
 
 

include/linux/mlx5/driver.h

@@ -42,6 +42,7 @@
 #include <linux/vmalloc.h>
 #include <linux/radix-tree.h>
 #include <linux/workqueue.h>
+#include <linux/interrupt.h>
 
 #include <linux/mlx5/device.h>
 #include <linux/mlx5/doorbell.h>
@@ -312,6 +313,14 @@ struct mlx5_buf {
 	u8			page_shift;
 };
 
+struct mlx5_eq_tasklet {
+	struct list_head list;
+	struct list_head process_list;
+	struct tasklet_struct task;
+	/* lock on completion tasklet list */
+	spinlock_t lock;
+};
+
 struct mlx5_eq {
 	struct mlx5_core_dev   *dev;
 	__be32 __iomem	       *doorbell;
@@ -325,6 +334,7 @@ struct mlx5_eq {
 	struct list_head	list;
 	int			index;
 	struct mlx5_rsc_debug	*dbg;
+	struct mlx5_eq_tasklet	tasklet_ctx;
 };
 
 struct mlx5_core_psv {