RDMA/hns: Init SRQ table for hip08

This patch inits hem resource for SRQ table, includes
SRQWQE and SRQWQE index resource.
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

drivers/infiniband/hw/hns/Makefile

@@ -7,7 +7,7 @@ ccflags-y :=  -Idrivers/net/ethernet/hisilicon/hns3
 obj-$(CONFIG_INFINIBAND_HNS) += hns-roce.o
 hns-roce-objs := hns_roce_main.o hns_roce_cmd.o hns_roce_pd.o \
 	hns_roce_ah.o hns_roce_hem.o hns_roce_mr.o hns_roce_qp.o \
-	hns_roce_cq.o hns_roce_alloc.o hns_roce_db.o
+	hns_roce_cq.o hns_roce_alloc.o hns_roce_db.o hns_roce_srq.o
 obj-$(CONFIG_INFINIBAND_HNS_HIP06) += hns-roce-hw-v1.o
 hns-roce-hw-v1-objs := hns_roce_hw_v1.o
 obj-$(CONFIG_INFINIBAND_HNS_HIP08) += hns-roce-hw-v2.o

drivers/infiniband/hw/hns/hns_roce_alloc.c

@@ -239,6 +239,8 @@ int hns_roce_buf_alloc(struct hns_roce_dev *hr_dev, u32 size, u32 max_direct,
 
 void hns_roce_cleanup_bitmap(struct hns_roce_dev *hr_dev)
 {
+	if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_SRQ)
+		hns_roce_cleanup_srq_table(hr_dev);
 	hns_roce_cleanup_qp_table(hr_dev);
 	hns_roce_cleanup_cq_table(hr_dev);
 	hns_roce_cleanup_mr_table(hr_dev);

drivers/infiniband/hw/hns/hns_roce_device.h

@@ -205,6 +205,8 @@ enum {
 enum hns_roce_mtt_type {
 	MTT_TYPE_WQE,
 	MTT_TYPE_CQE,
+	MTT_TYPE_SRQWQE,
+	MTT_TYPE_IDX
 };
 
 enum {
@@ -340,6 +342,10 @@ struct hns_roce_mr_table {
 	struct hns_roce_hem_table	mtpt_table;
 	struct hns_roce_buddy		mtt_cqe_buddy;
 	struct hns_roce_hem_table	mtt_cqe_table;
+	struct hns_roce_buddy		mtt_srqwqe_buddy;
+	struct hns_roce_hem_table	mtt_srqwqe_table;
+	struct hns_roce_buddy		mtt_idx_buddy;
+	struct hns_roce_hem_table	mtt_idx_table;
 };
 
 struct hns_roce_wq {
@@ -454,6 +460,12 @@ struct hns_roce_cq_table {
 	struct hns_roce_hem_table	table;
 };
 
+struct hns_roce_srq_table {
+	struct hns_roce_bitmap		bitmap;
+	struct xarray			xa;
+	struct hns_roce_hem_table	table;
+};
+
 struct hns_roce_raq_table {
 	struct hns_roce_buf_list	*e_raq_buf;
 };
@@ -680,6 +692,8 @@ struct hns_roce_caps {
 	u32		max_extend_sg;
 	int		num_qps;	/* 256k */
 	int             reserved_qps;
+	u32		max_srq_sg;
+	int		num_srqs;
 	u32		max_wqes;	/* 16k */
 	u32		max_srqs;
 	u32		max_srq_wrs;
@@ -694,12 +708,16 @@ struct hns_roce_caps {
 	int		min_cqes;
 	u32		min_wqes;
 	int		reserved_cqs;
+	int		reserved_srqs;
+	u32		max_srqwqes;
 	int		num_aeq_vectors;	/* 1 */
 	int		num_comp_vectors;
 	int		num_other_vectors;
 	int		num_mtpts;
 	u32		num_mtt_segs;
 	u32		num_cqe_segs;
+	u32		num_srqwqe_segs;
+	u32		num_idx_segs;
 	int		reserved_mrws;
 	int		reserved_uars;
 	int		num_pds;
@@ -713,6 +731,8 @@ struct hns_roce_caps {
 	int		irrl_entry_sz;
 	int		trrl_entry_sz;
 	int		cqc_entry_sz;
+	int		srqc_entry_sz;
+	int		idx_entry_sz;
 	u32		pbl_ba_pg_sz;
 	u32		pbl_buf_pg_sz;
 	u32		pbl_hop_num;
@@ -843,6 +863,7 @@ struct hns_roce_dev {
 	struct hns_roce_uar_table uar_table;
 	struct hns_roce_mr_table  mr_table;
 	struct hns_roce_cq_table  cq_table;
+	struct hns_roce_srq_table srq_table;
 	struct hns_roce_qp_table  qp_table;
 	struct hns_roce_eq_table  eq_table;
 
@@ -955,12 +976,14 @@ int hns_roce_init_mr_table(struct hns_roce_dev *hr_dev);
 int hns_roce_init_eq_table(struct hns_roce_dev *hr_dev);
 int hns_roce_init_cq_table(struct hns_roce_dev *hr_dev);
 int hns_roce_init_qp_table(struct hns_roce_dev *hr_dev);
+int hns_roce_init_srq_table(struct hns_roce_dev *hr_dev);
 
 void hns_roce_cleanup_pd_table(struct hns_roce_dev *hr_dev);
 void hns_roce_cleanup_mr_table(struct hns_roce_dev *hr_dev);
 void hns_roce_cleanup_eq_table(struct hns_roce_dev *hr_dev);
 void hns_roce_cleanup_cq_table(struct hns_roce_dev *hr_dev);
 void hns_roce_cleanup_qp_table(struct hns_roce_dev *hr_dev);
+void hns_roce_cleanup_srq_table(struct hns_roce_dev *hr_dev);
 
 int hns_roce_bitmap_alloc(struct hns_roce_bitmap *bitmap, unsigned long *obj);
 void hns_roce_bitmap_free(struct hns_roce_bitmap *bitmap, unsigned long obj,

drivers/infiniband/hw/hns/hns_roce_hem.c

@@ -1041,6 +1041,15 @@ void hns_roce_cleanup_hem_table(struct hns_roce_dev *hr_dev,
 
 void hns_roce_cleanup_hem(struct hns_roce_dev *hr_dev)
 {
+	if ((hr_dev->caps.num_idx_segs))
+		hns_roce_cleanup_hem_table(hr_dev,
+					   &hr_dev->mr_table.mtt_idx_table);
+	if (hr_dev->caps.num_srqwqe_segs)
+		hns_roce_cleanup_hem_table(hr_dev,
+					   &hr_dev->mr_table.mtt_srqwqe_table);
+	if (hr_dev->caps.srqc_entry_sz)
+		hns_roce_cleanup_hem_table(hr_dev,
+					   &hr_dev->srq_table.table);
 	hns_roce_cleanup_hem_table(hr_dev, &hr_dev->cq_table.table);
 	if (hr_dev->caps.trrl_entry_sz)
 		hns_roce_cleanup_hem_table(hr_dev,

drivers/infiniband/hw/hns/hns_roce_hem.h

@@ -48,6 +48,8 @@ enum {
 	 /* UNMAP HEM */
 	HEM_TYPE_MTT,
 	HEM_TYPE_CQE,
+	HEM_TYPE_SRQWQE,
+	HEM_TYPE_IDX,
 	HEM_TYPE_IRRL,
 	HEM_TYPE_TRRL,
 };

drivers/infiniband/hw/hns/hns_roce_hw_v2.c

@@ -1276,11 +1276,14 @@ static int hns_roce_v2_profile(struct hns_roce_dev *hr_dev)
 	caps->num_qps		= HNS_ROCE_V2_MAX_QP_NUM;
 	caps->max_wqes		= HNS_ROCE_V2_MAX_WQE_NUM;
 	caps->num_cqs		= HNS_ROCE_V2_MAX_CQ_NUM;
+	caps->num_srqs		= HNS_ROCE_V2_MAX_SRQ_NUM;
 	caps->max_cqes		= HNS_ROCE_V2_MAX_CQE_NUM;
+	caps->max_srqwqes	= HNS_ROCE_V2_MAX_SRQWQE_NUM;
 	caps->max_sq_sg		= HNS_ROCE_V2_MAX_SQ_SGE_NUM;
 	caps->max_extend_sg	= HNS_ROCE_V2_MAX_EXTEND_SGE_NUM;
 	caps->max_rq_sg		= HNS_ROCE_V2_MAX_RQ_SGE_NUM;
 	caps->max_sq_inline	= HNS_ROCE_V2_MAX_SQ_INLINE;
+	caps->max_srq_sg	= HNS_ROCE_V2_MAX_SRQ_SGE_NUM;
 	caps->num_uars		= HNS_ROCE_V2_UAR_NUM;
 	caps->phy_num_uars	= HNS_ROCE_V2_PHY_UAR_NUM;
 	caps->num_aeq_vectors	= HNS_ROCE_V2_AEQE_VEC_NUM;
@@ -1289,6 +1292,8 @@ static int hns_roce_v2_profile(struct hns_roce_dev *hr_dev)
 	caps->num_mtpts		= HNS_ROCE_V2_MAX_MTPT_NUM;
 	caps->num_mtt_segs	= HNS_ROCE_V2_MAX_MTT_SEGS;
 	caps->num_cqe_segs	= HNS_ROCE_V2_MAX_CQE_SEGS;
+	caps->num_srqwqe_segs	= HNS_ROCE_V2_MAX_SRQWQE_SEGS;
+	caps->num_idx_segs	= HNS_ROCE_V2_MAX_IDX_SEGS;
 	caps->num_pds		= HNS_ROCE_V2_MAX_PD_NUM;
 	caps->max_qp_init_rdma	= HNS_ROCE_V2_MAX_QP_INIT_RDMA;
 	caps->max_qp_dest_rdma	= HNS_ROCE_V2_MAX_QP_DEST_RDMA;
@@ -1299,8 +1304,10 @@ static int hns_roce_v2_profile(struct hns_roce_dev *hr_dev)
 	caps->irrl_entry_sz	= HNS_ROCE_V2_IRRL_ENTRY_SZ;
 	caps->trrl_entry_sz	= HNS_ROCE_V2_TRRL_ENTRY_SZ;
 	caps->cqc_entry_sz	= HNS_ROCE_V2_CQC_ENTRY_SZ;
+	caps->srqc_entry_sz	= HNS_ROCE_V2_SRQC_ENTRY_SZ;
 	caps->mtpt_entry_sz	= HNS_ROCE_V2_MTPT_ENTRY_SZ;
 	caps->mtt_entry_sz	= HNS_ROCE_V2_MTT_ENTRY_SZ;
+	caps->idx_entry_sz	= 4;
 	caps->cq_entry_sz	= HNS_ROCE_V2_CQE_ENTRY_SIZE;
 	caps->page_size_cap	= HNS_ROCE_V2_PAGE_SIZE_SUPPORTED;
 	caps->reserved_lkey	= 0;
@@ -1308,6 +1315,7 @@ static int hns_roce_v2_profile(struct hns_roce_dev *hr_dev)
 	caps->reserved_mrws	= 1;
 	caps->reserved_uars	= 0;
 	caps->reserved_cqs	= 0;
+	caps->reserved_srqs	= 0;
 	caps->reserved_qps	= HNS_ROCE_V2_RSV_QPS;
 
 	caps->qpc_ba_pg_sz	= 0;

drivers/infiniband/hw/hns/hns_roce_hw_v2.h

@@ -50,9 +50,12 @@
 #define HNS_ROCE_V2_MAX_SRQ_WR			0x8000
 #define HNS_ROCE_V2_MAX_SRQ_SGE			0x100
 #define HNS_ROCE_V2_MAX_CQ_NUM			0x8000
+#define HNS_ROCE_V2_MAX_SRQ_NUM			0x100000
 #define HNS_ROCE_V2_MAX_CQE_NUM			0x10000
+#define HNS_ROCE_V2_MAX_SRQWQE_NUM		0x8000
 #define HNS_ROCE_V2_MAX_RQ_SGE_NUM		0x100
 #define HNS_ROCE_V2_MAX_SQ_SGE_NUM		0xff
+#define HNS_ROCE_V2_MAX_SRQ_SGE_NUM		0x100
 #define HNS_ROCE_V2_MAX_EXTEND_SGE_NUM		0x200000
 #define HNS_ROCE_V2_MAX_SQ_INLINE		0x20
 #define HNS_ROCE_V2_UAR_NUM			256
@@ -64,6 +67,8 @@
 #define HNS_ROCE_V2_MAX_MTPT_NUM		0x8000
 #define HNS_ROCE_V2_MAX_MTT_SEGS		0x1000000
 #define HNS_ROCE_V2_MAX_CQE_SEGS		0x1000000
+#define HNS_ROCE_V2_MAX_SRQWQE_SEGS		0x1000000
+#define HNS_ROCE_V2_MAX_IDX_SEGS		0x1000000
 #define HNS_ROCE_V2_MAX_PD_NUM			0x1000000
 #define HNS_ROCE_V2_MAX_QP_INIT_RDMA		128
 #define HNS_ROCE_V2_MAX_QP_DEST_RDMA		128
@@ -74,6 +79,7 @@
 #define HNS_ROCE_V2_IRRL_ENTRY_SZ		64
 #define HNS_ROCE_V2_TRRL_ENTRY_SZ		48
 #define HNS_ROCE_V2_CQC_ENTRY_SZ		64
+#define HNS_ROCE_V2_SRQC_ENTRY_SZ		64
 #define HNS_ROCE_V2_MTPT_ENTRY_SZ		64
 #define HNS_ROCE_V2_MTT_ENTRY_SZ		64
 #define HNS_ROCE_V2_CQE_ENTRY_SIZE		32

drivers/infiniband/hw/hns/hns_roce_main.c

@@ -651,8 +651,58 @@ static int hns_roce_init_hem(struct hns_roce_dev *hr_dev)
 		goto err_unmap_trrl;
 	}
 
+	if (hr_dev->caps.srqc_entry_sz) {
+		ret = hns_roce_init_hem_table(hr_dev, &hr_dev->srq_table.table,
+					      HEM_TYPE_SRQC,
+					      hr_dev->caps.srqc_entry_sz,
+					      hr_dev->caps.num_srqs, 1);
+		if (ret) {
+			dev_err(dev,
+			      "Failed to init SRQ context memory, aborting.\n");
+			goto err_unmap_cq;
+		}
+	}
+
+	if (hr_dev->caps.num_srqwqe_segs) {
+		ret = hns_roce_init_hem_table(hr_dev,
+					     &hr_dev->mr_table.mtt_srqwqe_table,
+					     HEM_TYPE_SRQWQE,
+					     hr_dev->caps.mtt_entry_sz,
+					     hr_dev->caps.num_srqwqe_segs, 1);
+		if (ret) {
+			dev_err(dev,
+				"Failed to init MTT srqwqe memory, aborting.\n");
+			goto err_unmap_srq;
+		}
+	}
+
+	if (hr_dev->caps.num_idx_segs) {
+		ret = hns_roce_init_hem_table(hr_dev,
+					      &hr_dev->mr_table.mtt_idx_table,
+					      HEM_TYPE_IDX,
+					      hr_dev->caps.idx_entry_sz,
+					      hr_dev->caps.num_idx_segs, 1);
+		if (ret) {
+			dev_err(dev,
+				"Failed to init MTT idx memory, aborting.\n");
+			goto err_unmap_srqwqe;
+		}
+	}
+
 	return 0;
 
+err_unmap_srqwqe:
+	if (hr_dev->caps.num_srqwqe_segs)
+		hns_roce_cleanup_hem_table(hr_dev,
+					   &hr_dev->mr_table.mtt_srqwqe_table);
+
+err_unmap_srq:
+	if (hr_dev->caps.srqc_entry_sz)
+		hns_roce_cleanup_hem_table(hr_dev, &hr_dev->srq_table.table);
+
+err_unmap_cq:
+	hns_roce_cleanup_hem_table(hr_dev, &hr_dev->cq_table.table);
+
 err_unmap_trrl:
 	if (hr_dev->caps.trrl_entry_sz)
 		hns_roce_cleanup_hem_table(hr_dev,
@@ -732,8 +782,21 @@ static int hns_roce_setup_hca(struct hns_roce_dev *hr_dev)
 		goto err_cq_table_free;
 	}
 
+	if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_SRQ) {
+		ret = hns_roce_init_srq_table(hr_dev);
+		if (ret) {
+			dev_err(dev,
+				"Failed to init share receive queue table.\n");
+			goto err_qp_table_free;
+		}
+	}
+
 	return 0;
 
+err_qp_table_free:
+	if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_SRQ)
+		hns_roce_cleanup_qp_table(hr_dev);
+
 err_cq_table_free:
 	hns_roce_cleanup_cq_table(hr_dev);
 

drivers/infiniband/hw/hns/hns_roce_srq.c

+// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
+/*
+ * Copyright (c) 2018 Hisilicon Limited.
+ */
+
+#include <rdma/ib_umem.h>
+#include <rdma/hns-abi.h>
+#include "hns_roce_device.h"
+#include "hns_roce_cmd.h"
+#include "hns_roce_hem.h"
+
+int hns_roce_init_srq_table(struct hns_roce_dev *hr_dev)
+{
+	struct hns_roce_srq_table *srq_table = &hr_dev->srq_table;
+
+	xa_init(&srq_table->xa);
+
+	return hns_roce_bitmap_init(&srq_table->bitmap, hr_dev->caps.num_srqs,
+				    hr_dev->caps.num_srqs - 1,
+				    hr_dev->caps.reserved_srqs, 0);
+}
+
+void hns_roce_cleanup_srq_table(struct hns_roce_dev *hr_dev)
+{
+	hns_roce_bitmap_cleanup(&hr_dev->srq_table.bitmap);
+}