[PATCH rdma-core 0/6] libhns: Add support for Dynamic Context Attachment

[PATCH rdma-core 0/6] libhns: Add support for Dynamic Context Attachment

[Weihang Li]

The HIP09 introduces the DCA(Dynamic Context Attachment) feature which
supports many RC QPs to share the WQE buffer in a memory pool. If a QP
enables DCA feature, the WQE's buffer will not be allocated when creating
but when the users start to post WRs. This will reduce the memory
consumption when there are too many QPs are inactive.

For more detailed information, please refer to the man pages provided by
the last patch of this series.

This series is associated with the kernel one "RDMA/hns: Add support for
Dynamic Context Attachment", and two RFC versions of this series has been
sent before.

No changes since RFC v2.
* Link: https://patchwork.kernel.org/project/linux-rdma/cover/1614847759-33139-1-git-send-email-liweihang@huawei.com/

Changes since RFC v1:
* Add direct verbs to set the parameters about size that used to
  configuring DCA. 
* Add man pages to explain what is DCA, how does it works and how to use
  it.
* Link: https://patchwork.kernel.org/project/linux-rdma/cover/1612667574-56673-1-git-send-email-liweihang@huawei.com/

Weihang Li (1):
  Update kernel headers

Xi Wang (5):
  libhns: Introduce DCA for RC QP
  libhns: Add support for shrinking DCA memory pool
  libhns: Add support for attaching QP's WQE buffer
  libhns: Add direct verbs support to config DCA
  libhns: Add man pages to introduce DCA feature

 CMakeLists.txt                             |   1 +
 debian/control                             |   2 +-
 debian/ibverbs-providers.install           |   1 +
 debian/ibverbs-providers.lintian-overrides |   4 +-
 debian/ibverbs-providers.symbols           |   6 +
 debian/libibverbs-dev.install              |   6 +
 kernel-headers/rdma/hns-abi.h              |  64 +++++
 providers/hns/CMakeLists.txt               |   9 +-
 providers/hns/hns_roce_u.c                 | 100 ++++++++
 providers/hns/hns_roce_u.h                 |  44 ++++
 providers/hns/hns_roce_u_abi.h             |   1 +
 providers/hns/hns_roce_u_buf.c             | 387 +++++++++++++++++++++++++++++
 providers/hns/hns_roce_u_hw_v2.c           | 130 ++++++++--
 providers/hns/hns_roce_u_hw_v2.h           |   7 +
 providers/hns/hns_roce_u_verbs.c           |  67 ++++-
 providers/hns/hnsdv.h                      |  61 +++++
 providers/hns/libhns.map                   |   9 +
 providers/hns/man/CMakeLists.txt           |   7 +
 providers/hns/man/hns_dca.7.md             |  35 +++
 providers/hns/man/hnsdv.7.md               |  34 +++
 providers/hns/man/hnsdv_create_qp.3.md     |  69 +++++
 providers/hns/man/hnsdv_is_supported.3.md  |  39 +++
 providers/hns/man/hnsdv_open_device.3.md   |  70 ++++++
 redhat/rdma-core.spec                      |   7 +-
 suse/rdma-core.spec                        |  21 +-
 25 files changed, 1149 insertions(+), 32 deletions(-)
 create mode 100644 providers/hns/hnsdv.h
 create mode 100644 providers/hns/libhns.map
 create mode 100644 providers/hns/man/CMakeLists.txt
 create mode 100644 providers/hns/man/hns_dca.7.md
 create mode 100644 providers/hns/man/hnsdv.7.md
 create mode 100644 providers/hns/man/hnsdv_create_qp.3.md
 create mode 100644 providers/hns/man/hnsdv_is_supported.3.md
 create mode 100644 providers/hns/man/hnsdv_open_device.3.md

-- 
2.7.4

[PATCH rdma-core 1/6] Update kernel headers

[Weihang Li]

To commit ?? ("RDMA/hns: Add method to query WQE buffer's address").

Signed-off-by: Weihang Li <liweihang@huawei.com>
---
 kernel-headers/rdma/hns-abi.h | 64 +++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 64 insertions(+)

diff --git a/kernel-headers/rdma/hns-abi.h b/kernel-headers/rdma/hns-abi.h
index 42b1776..edcecc1 100644
--- a/kernel-headers/rdma/hns-abi.h
+++ b/kernel-headers/rdma/hns-abi.h
@@ -77,21 +77,85 @@ enum hns_roce_qp_cap_flags {
 	HNS_ROCE_QP_CAP_RQ_RECORD_DB = 1 << 0,
 	HNS_ROCE_QP_CAP_SQ_RECORD_DB = 1 << 1,
 	HNS_ROCE_QP_CAP_OWNER_DB = 1 << 2,
+	HNS_ROCE_QP_CAP_DCA = 1 << 4,
 };
 
 struct hns_roce_ib_create_qp_resp {
 	__aligned_u64 cap_flags;
 };
 
+enum {
+	HNS_ROCE_CAP_FLAG_DCA_MODE = 1 << 15,
+};
+
 struct hns_roce_ib_alloc_ucontext_resp {
 	__u32	qp_tab_size;
 	__u32	cqe_size;
 	__u32	srq_tab_size;
 	__u32	reserved;
+	__aligned_u64 cap_flags;
 };
 
 struct hns_roce_ib_alloc_pd_resp {
 	__u32 pdn;
 };
 
+#define UVERBS_ID_NS_MASK 0xF000
+#define UVERBS_ID_NS_SHIFT 12
+
+enum hns_ib_objects {
+	HNS_IB_OBJECT_DCA_MEM = (1U << UVERBS_ID_NS_SHIFT),
+};
+
+enum hns_ib_dca_mem_methods {
+	HNS_IB_METHOD_DCA_MEM_REG = (1U << UVERBS_ID_NS_SHIFT),
+	HNS_IB_METHOD_DCA_MEM_DEREG,
+	HNS_IB_METHOD_DCA_MEM_SHRINK,
+	HNS_IB_METHOD_DCA_MEM_ATTACH,
+	HNS_IB_METHOD_DCA_MEM_DETACH,
+	HNS_IB_METHOD_DCA_MEM_QUERY,
+};
+
+enum hns_ib_dca_mem_reg_attrs {
+	HNS_IB_ATTR_DCA_MEM_REG_HANDLE = (1U << UVERBS_ID_NS_SHIFT),
+	HNS_IB_ATTR_DCA_MEM_REG_LEN,
+	HNS_IB_ATTR_DCA_MEM_REG_ADDR,
+	HNS_IB_ATTR_DCA_MEM_REG_KEY,
+};
+
+enum hns_ib_dca_mem_dereg_attrs {
+	HNS_IB_ATTR_DCA_MEM_DEREG_HANDLE = (1U << UVERBS_ID_NS_SHIFT),
+};
+
+enum hns_ib_dca_mem_shrink_attrs {
+	HNS_IB_ATTR_DCA_MEM_SHRINK_HANDLE = (1U << UVERBS_ID_NS_SHIFT),
+	HNS_IB_ATTR_DCA_MEM_SHRINK_RESERVED_SIZE,
+	HNS_IB_ATTR_DCA_MEM_SHRINK_OUT_FREE_KEY,
+	HNS_IB_ATTR_DCA_MEM_SHRINK_OUT_FREE_MEMS,
+};
+
+#define HNS_IB_ATTACH_FLAGS_NEW_BUFFER 1U
+
+enum hns_ib_dca_mem_attach_attrs {
+	HNS_IB_ATTR_DCA_MEM_ATTACH_HANDLE = (1U << UVERBS_ID_NS_SHIFT),
+	HNS_IB_ATTR_DCA_MEM_ATTACH_SQ_OFFSET,
+	HNS_IB_ATTR_DCA_MEM_ATTACH_SGE_OFFSET,
+	HNS_IB_ATTR_DCA_MEM_ATTACH_RQ_OFFSET,
+	HNS_IB_ATTR_DCA_MEM_ATTACH_OUT_ALLOC_FLAGS,
+	HNS_IB_ATTR_DCA_MEM_ATTACH_OUT_ALLOC_PAGES,
+};
+
+enum hns_ib_dca_mem_detach_attrs {
+	HNS_IB_ATTR_DCA_MEM_DETACH_HANDLE = (1U << UVERBS_ID_NS_SHIFT),
+	HNS_IB_ATTR_DCA_MEM_DETACH_SQ_INDEX,
+};
+
+enum hns_ib_dca_mem_query_attrs {
+	HNS_IB_ATTR_DCA_MEM_QUERY_HANDLE = (1U << UVERBS_ID_NS_SHIFT),
+	HNS_IB_ATTR_DCA_MEM_QUERY_PAGE_INDEX,
+	HNS_IB_ATTR_DCA_MEM_QUERY_OUT_KEY,
+	HNS_IB_ATTR_DCA_MEM_QUERY_OUT_OFFSET,
+	HNS_IB_ATTR_DCA_MEM_QUERY_OUT_PAGE_COUNT,
+};
+
 #endif /* HNS_ABI_USER_H */
-- 
2.7.4

[PATCH rdma-core 2/6] libhns: Introduce DCA for RC QP

[Weihang Li]

From: Xi Wang <wangxi11@huawei.com>

The HIP09 introduces the DCA(Dynamic context attachment) feature which
supports many RC QPs to share the WQE buffer in a memory pool, this will
reduce the memory consumption when there are too many QPs inactive.

This patch wraps two functions for adding buffers to memory pool and
removing buffers from memory pool by calling ib cmd implemented in hns
kernel driver.

If a QP enables DCA feature, the WQE's buffer will be attached to the
memory pool when the users start to post WRs and be detached when all CQEs
has been polled.

Signed-off-by: Xi Wang <wangxi11@huawei.com>
Signed-off-by: Weihang Li <liweihang@huawei.com>
---
 providers/hns/hns_roce_u.c     |  45 ++++++++++++++
 providers/hns/hns_roce_u.h     |  18 ++++++
 providers/hns/hns_roce_u_buf.c | 137 +++++++++++++++++++++++++++++++++++++++++
 3 files changed, 200 insertions(+)

diff --git a/providers/hns/hns_roce_u.c b/providers/hns/hns_roce_u.c
index 3b31ad3..a4e0997 100644
--- a/providers/hns/hns_roce_u.c
+++ b/providers/hns/hns_roce_u.c
@@ -95,6 +95,40 @@ static const struct verbs_context_ops hns_common_ops = {
 	.get_srq_num = hns_roce_u_get_srq_num,
 };
 
+static int init_dca_context(struct hns_roce_context *ctx, int page_size)
+{
+	struct hns_roce_dca_ctx *dca_ctx = &ctx->dca_ctx;
+	int ret;
+
+	if (!(ctx->cap_flags & HNS_ROCE_CAP_FLAG_DCA_MODE))
+		return 0;
+
+	list_head_init(&dca_ctx->mem_list);
+	ret = pthread_spin_init(&dca_ctx->lock, PTHREAD_PROCESS_PRIVATE);
+	if (ret)
+		return ret;
+
+	dca_ctx->unit_size = page_size * HNS_DCA_DEFAULT_UNIT_PAGES;
+	dca_ctx->max_size = HNS_DCA_MAX_MEM_SIZE;
+	dca_ctx->mem_cnt = 0;
+
+	return 0;
+}
+
+static void uninit_dca_context(struct hns_roce_context *ctx)
+{
+	struct hns_roce_dca_ctx *dca_ctx = &ctx->dca_ctx;
+
+	if (!(ctx->cap_flags & HNS_ROCE_CAP_FLAG_DCA_MODE))
+		return;
+
+	pthread_spin_lock(&dca_ctx->lock);
+	hns_roce_cleanup_dca_mem(ctx);
+	pthread_spin_unlock(&dca_ctx->lock);
+
+	pthread_spin_destroy(&dca_ctx->lock);
+}
+
 static struct verbs_context *hns_roce_alloc_context(struct ibv_device *ibdev,
 						    int cmd_fd,
 						    void *private_data)
@@ -123,6 +157,8 @@ static struct verbs_context *hns_roce_alloc_context(struct ibv_device *ibdev,
 	else
 		context->cqe_size = HNS_ROCE_V3_CQE_SIZE;
 
+	context->cap_flags = resp.cap_flags;
+
 	context->num_qps = resp.qp_tab_size;
 	context->num_srqs = resp.srq_tab_size;
 
@@ -178,8 +214,15 @@ static struct verbs_context *hns_roce_alloc_context(struct ibv_device *ibdev,
 	verbs_set_ops(&context->ibv_ctx, &hns_common_ops);
 	verbs_set_ops(&context->ibv_ctx, &hr_dev->u_hw->hw_ops);
 
+	if (init_dca_context(context, hr_dev->page_size))
+		goto tptr_free;
+
 	return &context->ibv_ctx;
 
+tptr_free:
+	if (hr_dev->hw_version == HNS_ROCE_HW_VER1)
+		munmap(context->cq_tptr_base, HNS_ROCE_CQ_DB_BUF_SIZE);
+
 db_free:
 	munmap(context->uar, hr_dev->page_size);
 	context->uar = NULL;
@@ -199,6 +242,8 @@ static void hns_roce_free_context(struct ibv_context *ibctx)
 	if (hr_dev->hw_version == HNS_ROCE_HW_VER1)
 		munmap(context->cq_tptr_base, HNS_ROCE_CQ_DB_BUF_SIZE);
 
+	uninit_dca_context(context);
+
 	verbs_uninit_context(&context->ibv_ctx);
 	free(context);
 }
diff --git a/providers/hns/hns_roce_u.h b/providers/hns/hns_roce_u.h
index 8f805dd..d26ef9c 100644
--- a/providers/hns/hns_roce_u.h
+++ b/providers/hns/hns_roce_u.h
@@ -145,8 +145,21 @@ struct hns_roce_db_page {
 	bitmap			*bitmap;
 };
 
+#define HNS_DCA_MAX_MEM_SIZE ~0UL
+#define HNS_DCA_DEFAULT_UNIT_PAGES 16
+
+struct hns_roce_dca_ctx {
+	struct list_head mem_list;
+	pthread_spinlock_t lock;
+	uint64_t max_size;
+	uint64_t curr_size;
+	int mem_cnt;
+	unsigned int unit_size;
+};
+
 struct hns_roce_context {
 	struct verbs_context		ibv_ctx;
+	uint32_t			cap_flags;
 	void				*uar;
 	pthread_spinlock_t		uar_lock;
 
@@ -179,6 +192,8 @@ struct hns_roce_context {
 	unsigned int			max_srq_sge;
 	int				max_cqe;
 	unsigned int			cqe_size;
+
+	struct hns_roce_dca_ctx		dca_ctx;
 };
 
 struct hns_roce_pd {
@@ -424,6 +439,9 @@ void hns_roce_free_buf(struct hns_roce_buf *buf);
 
 void hns_roce_free_qp_buf(struct hns_roce_qp *qp, struct hns_roce_context *ctx);
 
+void hns_roce_cleanup_dca_mem(struct hns_roce_context *ctx);
+int hns_roce_add_dca_mem(struct hns_roce_context *ctx, uint32_t size);
+
 void hns_roce_init_qp_indices(struct hns_roce_qp *qp);
 
 extern const struct hns_roce_u_hw hns_roce_u_hw_v1;
diff --git a/providers/hns/hns_roce_u_buf.c b/providers/hns/hns_roce_u_buf.c
index 471dd9c..82a8849 100644
--- a/providers/hns/hns_roce_u_buf.c
+++ b/providers/hns/hns_roce_u_buf.c
@@ -60,3 +60,140 @@ void hns_roce_free_buf(struct hns_roce_buf *buf)
 
 	munmap(buf->buf, buf->length);
 }
+
+struct hns_roce_dca_mem {
+	uint32_t handle;
+	struct list_node entry;
+	struct hns_roce_buf buf;
+	struct hns_roce_context *ctx;
+};
+
+static void free_dca_mem(struct hns_roce_context *ctx,
+			 struct hns_roce_dca_mem *mem)
+{
+	hns_roce_free_buf(&mem->buf);
+	free(mem);
+}
+
+static struct hns_roce_dca_mem *alloc_dca_mem(uint32_t size)
+{
+	struct hns_roce_dca_mem *mem = NULL;
+	int ret;
+
+	mem = malloc(sizeof(struct hns_roce_dca_mem));
+	if (!mem) {
+		errno = ENOMEM;
+		return NULL;
+	}
+
+	ret = hns_roce_alloc_buf(&mem->buf, size, HNS_HW_PAGE_SIZE);
+	if (ret) {
+		errno = ENOMEM;
+		free(mem);
+		return NULL;
+	}
+
+	return mem;
+}
+
+static inline uint64_t dca_mem_to_key(struct hns_roce_dca_mem *dca_mem)
+{
+	return (uintptr_t)dca_mem;
+}
+
+static inline void *dca_mem_addr(struct hns_roce_dca_mem *dca_mem, int offset)
+{
+	return dca_mem->buf.buf + offset;
+}
+
+static int register_dca_mem(struct hns_roce_context *ctx, uint64_t key,
+			    void *addr, uint32_t size, uint32_t *handle)
+{
+	struct ib_uverbs_attr *attr;
+	int ret;
+
+	DECLARE_COMMAND_BUFFER(cmd, HNS_IB_OBJECT_DCA_MEM,
+			       HNS_IB_METHOD_DCA_MEM_REG, 4);
+	fill_attr_in_uint32(cmd, HNS_IB_ATTR_DCA_MEM_REG_LEN, size);
+	fill_attr_in_uint64(cmd, HNS_IB_ATTR_DCA_MEM_REG_ADDR,
+			    ioctl_ptr_to_u64(addr));
+	fill_attr_in_uint64(cmd, HNS_IB_ATTR_DCA_MEM_REG_KEY, key);
+	attr = fill_attr_out_obj(cmd, HNS_IB_ATTR_DCA_MEM_REG_HANDLE);
+
+	ret = execute_ioctl(&ctx->ibv_ctx.context, cmd);
+	if (ret)
+		return ret;
+
+	*handle = read_attr_obj(HNS_IB_ATTR_DCA_MEM_REG_HANDLE, attr);
+
+	return ret;
+}
+
+static void deregister_dca_mem(struct hns_roce_context *ctx, uint32_t handle)
+{
+	DECLARE_COMMAND_BUFFER(cmd, HNS_IB_OBJECT_DCA_MEM,
+			       HNS_IB_METHOD_DCA_MEM_DEREG, 1);
+	fill_attr_in_obj(cmd, HNS_IB_ATTR_DCA_MEM_DEREG_HANDLE, handle);
+	execute_ioctl(&ctx->ibv_ctx.context, cmd);
+}
+
+void hns_roce_cleanup_dca_mem(struct hns_roce_context *ctx)
+{
+	struct hns_roce_dca_ctx *dca_ctx = &ctx->dca_ctx;
+	struct hns_roce_dca_mem *mem;
+	struct hns_roce_dca_mem *tmp;
+
+	list_for_each_safe(&dca_ctx->mem_list, mem, tmp, entry)
+		deregister_dca_mem(ctx, mem->handle);
+}
+
+static bool add_dca_mem_enabled(struct hns_roce_dca_ctx *ctx,
+				uint32_t alloc_size)
+{
+	bool enable;
+
+	pthread_spin_lock(&ctx->lock);
+
+	if (ctx->unit_size == 0) /* Pool size can't be increased */
+		enable = false;
+	else if (ctx->max_size == HNS_DCA_MAX_MEM_SIZE) /* Pool size no limit */
+		enable = true;
+	else /* Pool size doesn't exceed max size */
+		enable = (ctx->curr_size + alloc_size) < ctx->max_size;
+
+	pthread_spin_unlock(&ctx->lock);
+
+	return enable;
+}
+
+int hns_roce_add_dca_mem(struct hns_roce_context *ctx, uint32_t size)
+{
+	struct hns_roce_dca_ctx *dca_ctx = &ctx->dca_ctx;
+	struct hns_roce_dca_mem *mem;
+	int ret;
+
+	if (!add_dca_mem_enabled(&ctx->dca_ctx, size))
+		return -ENOMEM;
+
+	/* Step 1: Alloc DCA mem address */
+	mem = alloc_dca_mem(DIV_ROUND_UP(size, dca_ctx->unit_size));
+	if (!mem)
+		return -ENOMEM;
+
+	/* Step 2: Register DCA mem uobject to pin user address */
+	ret = register_dca_mem(ctx, dca_mem_to_key(mem), dca_mem_addr(mem, 0),
+			       mem->buf.length, &mem->handle);
+	if (ret) {
+		free_dca_mem(ctx, mem);
+		return ret;
+	}
+
+	/* Step 3: Add DCA mem node to pool */
+	pthread_spin_lock(&dca_ctx->lock);
+	list_add_tail(&dca_ctx->mem_list, &mem->entry);
+	dca_ctx->mem_cnt++;
+	dca_ctx->curr_size += mem->buf.length;
+	pthread_spin_unlock(&dca_ctx->lock);
+
+	return 0;
+}
-- 
2.7.4

[PATCH rdma-core 3/6] libhns: Add support for shrinking DCA memory pool

[Weihang Li]

From: Xi Wang <wangxi11@huawei.com>

The QP's WQE buffer may be detached after QP is modified or CQE is polled,
and the state of DCA mem object may be changed as clean for no QP is using
it. So shrink the clean DCA mem from the memory pool and destroy the DCA
mem's buffer to reduce the memory consumption.

Signed-off-by: Xi Wang <wangxi11@huawei.com>
Signed-off-by: Weihang Li <liweihang@huawei.com>
---
 providers/hns/hns_roce_u.h       |  2 +
 providers/hns/hns_roce_u_buf.c   | 96 ++++++++++++++++++++++++++++++++++++++++
 providers/hns/hns_roce_u_hw_v2.c |  7 +++
 3 files changed, 105 insertions(+)

diff --git a/providers/hns/hns_roce_u.h b/providers/hns/hns_roce_u.h
index d26ef9c..a8a007e 100644
--- a/providers/hns/hns_roce_u.h
+++ b/providers/hns/hns_roce_u.h
@@ -152,6 +152,7 @@ struct hns_roce_dca_ctx {
 	struct list_head mem_list;
 	pthread_spinlock_t lock;
 	uint64_t max_size;
+	uint64_t min_size;
 	uint64_t curr_size;
 	int mem_cnt;
 	unsigned int unit_size;
@@ -439,6 +440,7 @@ void hns_roce_free_buf(struct hns_roce_buf *buf);
 
 void hns_roce_free_qp_buf(struct hns_roce_qp *qp, struct hns_roce_context *ctx);
 
+void hns_roce_shrink_dca_mem(struct hns_roce_context *ctx);
 void hns_roce_cleanup_dca_mem(struct hns_roce_context *ctx);
 int hns_roce_add_dca_mem(struct hns_roce_context *ctx, uint32_t size);
 
diff --git a/providers/hns/hns_roce_u_buf.c b/providers/hns/hns_roce_u_buf.c
index 82a8849..4970eac 100644
--- a/providers/hns/hns_roce_u_buf.c
+++ b/providers/hns/hns_roce_u_buf.c
@@ -101,6 +101,20 @@ static inline uint64_t dca_mem_to_key(struct hns_roce_dca_mem *dca_mem)
 	return (uintptr_t)dca_mem;
 }
 
+static struct hns_roce_dca_mem *key_to_dca_mem(struct hns_roce_dca_ctx *ctx,
+					       uint64_t key)
+{
+	struct hns_roce_dca_mem *mem;
+	struct hns_roce_dca_mem *tmp;
+
+	list_for_each_safe(&ctx->mem_list, mem, tmp, entry) {
+		if (dca_mem_to_key(mem) == key)
+			return mem;
+	}
+
+	return NULL;
+}
+
 static inline void *dca_mem_addr(struct hns_roce_dca_mem *dca_mem, int offset)
 {
 	return dca_mem->buf.buf + offset;
@@ -147,6 +161,25 @@ void hns_roce_cleanup_dca_mem(struct hns_roce_context *ctx)
 		deregister_dca_mem(ctx, mem->handle);
 }
 
+struct hns_dca_mem_shrink_resp {
+	uint32_t free_mems;
+	uint64_t free_key;
+};
+
+static int shrink_dca_mem(struct hns_roce_context *ctx, uint32_t handle,
+			  uint64_t size, struct hns_dca_mem_shrink_resp *resp)
+{
+	DECLARE_COMMAND_BUFFER(cmd, HNS_IB_OBJECT_DCA_MEM,
+			       HNS_IB_METHOD_DCA_MEM_SHRINK, 4);
+	fill_attr_in_obj(cmd, HNS_IB_ATTR_DCA_MEM_SHRINK_HANDLE, handle);
+	fill_attr_in_uint64(cmd, HNS_IB_ATTR_DCA_MEM_SHRINK_RESERVED_SIZE, size);
+	fill_attr_out(cmd, HNS_IB_ATTR_DCA_MEM_SHRINK_OUT_FREE_KEY,
+		      &resp->free_key, sizeof(resp->free_key));
+	fill_attr_out(cmd, HNS_IB_ATTR_DCA_MEM_SHRINK_OUT_FREE_MEMS,
+		      &resp->free_mems, sizeof(resp->free_mems));
+
+	return execute_ioctl(&ctx->ibv_ctx.context, cmd);
+}
 static bool add_dca_mem_enabled(struct hns_roce_dca_ctx *ctx,
 				uint32_t alloc_size)
 {
@@ -197,3 +230,66 @@ int hns_roce_add_dca_mem(struct hns_roce_context *ctx, uint32_t size)
 
 	return 0;
 }
+
+static bool shrink_dca_mem_enabled(struct hns_roce_dca_ctx *ctx)
+{
+	bool enable;
+
+	pthread_spin_lock(&ctx->lock);
+	enable = ctx->mem_cnt > 0 && ctx->min_size < ctx->max_size;
+	pthread_spin_unlock(&ctx->lock);
+
+	return enable;
+}
+
+void hns_roce_shrink_dca_mem(struct hns_roce_context *ctx)
+{
+	struct hns_roce_dca_ctx *dca_ctx = &ctx->dca_ctx;
+	struct hns_dca_mem_shrink_resp resp = {};
+	struct hns_roce_dca_mem *mem;
+	int dca_mem_cnt;
+	uint32_t handle;
+	int ret;
+
+	pthread_spin_lock(&dca_ctx->lock);
+	dca_mem_cnt = ctx->dca_ctx.mem_cnt;
+	pthread_spin_unlock(&dca_ctx->lock);
+	while (dca_mem_cnt > 0 && shrink_dca_mem_enabled(dca_ctx)) {
+		resp.free_mems = 0;
+		/* Step 1: Use any DCA mem uobject to shrink pool */
+		pthread_spin_lock(&dca_ctx->lock);
+		mem = list_tail(&dca_ctx->mem_list,
+				struct hns_roce_dca_mem, entry);
+		handle = mem ? mem->handle : 0;
+		pthread_spin_unlock(&dca_ctx->lock);
+		if (!mem)
+			break;
+
+		ret = shrink_dca_mem(ctx, handle, dca_ctx->min_size, &resp);
+		if (ret || likely(resp.free_mems < 1))
+			break;
+
+		/* Step 2: Remove shrunk DCA mem node from pool */
+		pthread_spin_lock(&dca_ctx->lock);
+		mem = key_to_dca_mem(dca_ctx, resp.free_key);
+		if (mem) {
+			list_del(&mem->entry);
+			dca_ctx->mem_cnt--;
+			dca_ctx->curr_size -= mem->buf.length;
+		}
+
+		handle = mem ? mem->handle : 0;
+		pthread_spin_unlock(&dca_ctx->lock);
+		if (!mem)
+			break;
+
+		/* Step 3: Destroy DCA mem uobject */
+		deregister_dca_mem(ctx, handle);
+		free_dca_mem(ctx, mem);
+		/* No any free memory after deregister 1 DCA mem */
+		if (resp.free_mems <= 1)
+			break;
+
+		dca_mem_cnt--;
+	}
+}
diff --git a/providers/hns/hns_roce_u_hw_v2.c b/providers/hns/hns_roce_u_hw_v2.c
index 4988943..16f033c 100644
--- a/providers/hns/hns_roce_u_hw_v2.c
+++ b/providers/hns/hns_roce_u_hw_v2.c
@@ -654,6 +654,10 @@ static int hns_roce_u_v2_poll_cq(struct ibv_cq *ibvcq, int ne,
 
 	pthread_spin_unlock(&cq->lock);
 
+	/* Try to shrink the DCA mem */
+	if (ctx->dca_ctx.mem_cnt > 0)
+		hns_roce_shrink_dca_mem(ctx);
+
 	return err == V2_CQ_POLL_ERR ? err : npolled;
 }
 
@@ -1524,6 +1528,9 @@ static int hns_roce_u_v2_destroy_qp(struct ibv_qp *ibqp)
 
 	free(qp);
 
+	if (ctx->dca_ctx.mem_cnt > 0)
+		hns_roce_shrink_dca_mem(ctx);
+
 	return ret;
 }
 
-- 
2.7.4

[PATCH rdma-core 4/6] libhns: Add support for attaching QP's WQE buffer

[Weihang Li]

From: Xi Wang <wangxi11@huawei.com>

If a uQP works in DCA mode, the WQE's buffer will be split as many blocks
and be stored into a list. The blocks are allocated from the DCA's memory
pool before posting WRs and are dropped when the QP's CI is equal to PI
after polling CQ.

Signed-off-by: Xi Wang <wangxi11@huawei.com>
Signed-off-by: Weihang Li <liweihang@huawei.com>
---
 providers/hns/hns_roce_u.h       |  24 +++++-
 providers/hns/hns_roce_u_buf.c   | 156 ++++++++++++++++++++++++++++++++++++++-
 providers/hns/hns_roce_u_hw_v2.c | 123 ++++++++++++++++++++++++++----
 providers/hns/hns_roce_u_hw_v2.h |   7 ++
 providers/hns/hns_roce_u_verbs.c |  32 ++++++--
 5 files changed, 318 insertions(+), 24 deletions(-)

diff --git a/providers/hns/hns_roce_u.h b/providers/hns/hns_roce_u.h
index a8a007e..a488694 100644
--- a/providers/hns/hns_roce_u.h
+++ b/providers/hns/hns_roce_u.h
@@ -122,6 +122,12 @@ struct hns_roce_device {
 	int				hw_version;
 };
 
+struct hns_roce_buf_list {
+	void **bufs;
+	unsigned int max_cnt;
+	unsigned int shift;
+};
+
 struct hns_roce_buf {
 	void				*buf;
 	unsigned int			length;
@@ -282,9 +288,20 @@ struct hns_roce_rinl_buf {
 	unsigned int			wqe_cnt;
 };
 
+struct hns_roce_dca_attach_attr {
+	uint32_t sq_offset;
+	uint32_t sge_offset;
+	uint32_t rq_offset;
+};
+
+struct hns_roce_dca_detach_attr {
+	uint32_t sq_index;
+};
+
 struct hns_roce_qp {
 	struct verbs_qp			verbs_qp;
 	struct hns_roce_buf		buf;
+	struct hns_roce_buf_list	page_list;
 	int				max_inline_data;
 	int				buf_size;
 	unsigned int			sq_signal_bits;
@@ -331,6 +348,7 @@ struct hns_roce_u_hw {
  * minimum page size.
  */
 #define hr_hw_page_align(x) align(x, HNS_HW_PAGE_SIZE)
+#define hr_hw_page_count(x) (hr_hw_page_align(x) / HNS_HW_PAGE_SIZE)
 
 static inline unsigned int to_hr_hem_entries_size(int count, int buf_shift)
 {
@@ -440,9 +458,13 @@ void hns_roce_free_buf(struct hns_roce_buf *buf);
 
 void hns_roce_free_qp_buf(struct hns_roce_qp *qp, struct hns_roce_context *ctx);
 
+int hns_roce_attach_dca_mem(struct hns_roce_context *ctx, uint32_t handle,
+			    struct hns_roce_dca_attach_attr *attr,
+			    uint32_t size, struct hns_roce_buf_list *buf_list);
+void hns_roce_detach_dca_mem(struct hns_roce_context *ctx, uint32_t handle,
+			     struct hns_roce_dca_detach_attr *attr);
 void hns_roce_shrink_dca_mem(struct hns_roce_context *ctx);
 void hns_roce_cleanup_dca_mem(struct hns_roce_context *ctx);
-int hns_roce_add_dca_mem(struct hns_roce_context *ctx, uint32_t size);
 
 void hns_roce_init_qp_indices(struct hns_roce_qp *qp);
 
diff --git a/providers/hns/hns_roce_u_buf.c b/providers/hns/hns_roce_u_buf.c
index 4970eac..d3e90d2 100644
--- a/providers/hns/hns_roce_u_buf.c
+++ b/providers/hns/hns_roce_u_buf.c
@@ -180,6 +180,66 @@ static int shrink_dca_mem(struct hns_roce_context *ctx, uint32_t handle,
 
 	return execute_ioctl(&ctx->ibv_ctx.context, cmd);
 }
+
+struct hns_dca_mem_query_resp {
+	uint64_t key;
+	uint32_t offset;
+	uint32_t page_count;
+};
+
+static int query_dca_mem(struct hns_roce_context *ctx, uint32_t handle,
+			 uint32_t index, struct hns_dca_mem_query_resp *resp)
+{
+	DECLARE_COMMAND_BUFFER(cmd, HNS_IB_OBJECT_DCA_MEM,
+			       HNS_IB_METHOD_DCA_MEM_QUERY, 5);
+	fill_attr_in_obj(cmd, HNS_IB_ATTR_DCA_MEM_QUERY_HANDLE, handle);
+	fill_attr_in_uint32(cmd, HNS_IB_ATTR_DCA_MEM_QUERY_PAGE_INDEX, index);
+	fill_attr_out(cmd, HNS_IB_ATTR_DCA_MEM_QUERY_OUT_KEY,
+		      &resp->key, sizeof(resp->key));
+	fill_attr_out(cmd, HNS_IB_ATTR_DCA_MEM_QUERY_OUT_OFFSET,
+		      &resp->offset, sizeof(resp->offset));
+	fill_attr_out(cmd, HNS_IB_ATTR_DCA_MEM_QUERY_OUT_PAGE_COUNT,
+		      &resp->page_count, sizeof(resp->page_count));
+	return execute_ioctl(&ctx->ibv_ctx.context, cmd);
+}
+
+static int detach_dca_mem(struct hns_roce_context *ctx, uint32_t handle,
+			  struct hns_roce_dca_detach_attr *attr)
+{
+	DECLARE_COMMAND_BUFFER(cmd, HNS_IB_OBJECT_DCA_MEM,
+			       HNS_IB_METHOD_DCA_MEM_DETACH, 4);
+	fill_attr_in_obj(cmd, HNS_IB_ATTR_DCA_MEM_DETACH_HANDLE, handle);
+	fill_attr_in_uint32(cmd, HNS_IB_ATTR_DCA_MEM_DETACH_SQ_INDEX,
+			    attr->sq_index);
+	return execute_ioctl(&ctx->ibv_ctx.context, cmd);
+}
+
+struct hns_dca_mem_attach_resp {
+#define HNS_DCA_ATTACH_OUT_FLAGS_NEW_BUFFER BIT(0)
+	uint32_t alloc_flags;
+	uint32_t alloc_pages;
+};
+
+static int attach_dca_mem(struct hns_roce_context *ctx, uint32_t handle,
+			  struct hns_roce_dca_attach_attr *attr,
+			  struct hns_dca_mem_attach_resp *resp)
+{
+	DECLARE_COMMAND_BUFFER(cmd, HNS_IB_OBJECT_DCA_MEM,
+			       HNS_IB_METHOD_DCA_MEM_ATTACH, 6);
+	fill_attr_in_obj(cmd, HNS_IB_ATTR_DCA_MEM_ATTACH_HANDLE, handle);
+	fill_attr_in_uint32(cmd, HNS_IB_ATTR_DCA_MEM_ATTACH_SQ_OFFSET,
+			    attr->sq_offset);
+	fill_attr_in_uint32(cmd, HNS_IB_ATTR_DCA_MEM_ATTACH_SGE_OFFSET,
+			    attr->sge_offset);
+	fill_attr_in_uint32(cmd, HNS_IB_ATTR_DCA_MEM_ATTACH_RQ_OFFSET,
+			    attr->rq_offset);
+	fill_attr_out(cmd, HNS_IB_ATTR_DCA_MEM_ATTACH_OUT_ALLOC_FLAGS,
+		      &resp->alloc_flags, sizeof(resp->alloc_flags));
+	fill_attr_out(cmd, HNS_IB_ATTR_DCA_MEM_ATTACH_OUT_ALLOC_PAGES,
+		      &resp->alloc_pages, sizeof(resp->alloc_pages));
+	return execute_ioctl(&ctx->ibv_ctx.context, cmd);
+}
+
 static bool add_dca_mem_enabled(struct hns_roce_dca_ctx *ctx,
 				uint32_t alloc_size)
 {
@@ -199,7 +259,7 @@ static bool add_dca_mem_enabled(struct hns_roce_dca_ctx *ctx,
 	return enable;
 }
 
-int hns_roce_add_dca_mem(struct hns_roce_context *ctx, uint32_t size)
+static int add_dca_mem(struct hns_roce_context *ctx, uint32_t size)
 {
 	struct hns_roce_dca_ctx *dca_ctx = &ctx->dca_ctx;
 	struct hns_roce_dca_mem *mem;
@@ -293,3 +353,97 @@ void hns_roce_shrink_dca_mem(struct hns_roce_context *ctx)
 		dca_mem_cnt--;
 	}
 }
+
+static void config_page_list(void *addr, struct hns_roce_buf_list *page_list,
+			     uint32_t page_index, int page_count)
+{
+	void **bufs = &page_list->bufs[page_index];
+	int page_size = 1 << page_list->shift;
+	int i;
+
+	for (i = 0; i < page_count; i++) {
+		bufs[i] = addr;
+		addr += page_size;
+	}
+}
+
+static int setup_dca_buf_list(struct hns_roce_context *ctx, uint32_t handle,
+			      struct hns_roce_buf_list *buf_list,
+			      uint32_t page_count)
+{
+	struct hns_roce_dca_ctx *dca_ctx = &ctx->dca_ctx;
+	struct hns_dca_mem_query_resp resp = {};
+	struct hns_roce_dca_mem *mem;
+	uint32_t idx = 0;
+	int ret;
+
+	while (idx < page_count && idx < buf_list->max_cnt) {
+		resp.page_count = 0;
+		ret = query_dca_mem(ctx, handle, idx, &resp);
+		if (ret)
+			return -ENOMEM;
+
+		if (resp.page_count < 1)
+			break;
+
+		pthread_spin_lock(&dca_ctx->lock);
+		mem = key_to_dca_mem(dca_ctx, resp.key);
+		if (mem && resp.offset < mem->buf.length) {
+			config_page_list(dca_mem_addr(mem, resp.offset),
+					 buf_list, idx, resp.page_count);
+		} else {
+			pthread_spin_unlock(&dca_ctx->lock);
+			break;
+		}
+
+		pthread_spin_unlock(&dca_ctx->lock);
+
+		idx += resp.page_count;
+	}
+
+	return (idx >= page_count) ? 0 : ENOMEM;
+}
+
+#define DCA_EXPAND_MEM_TRY_TIMES 3
+int hns_roce_attach_dca_mem(struct hns_roce_context *ctx, uint32_t handle,
+			    struct hns_roce_dca_attach_attr *attr,
+			    uint32_t size, struct hns_roce_buf_list *buf_list)
+{
+	uint32_t buf_pages = size >> buf_list->shift;
+	struct hns_dca_mem_attach_resp resp = {};
+	bool is_new_buf = true;
+	int try_times = 0;
+	int ret;
+
+	do {
+		resp.alloc_pages = 0;
+		ret = attach_dca_mem(ctx, handle, attr, &resp);
+		if (ret)
+			break;
+
+		if (resp.alloc_pages >= buf_pages) {
+			is_new_buf = !!(resp.alloc_flags &
+				     HNS_DCA_ATTACH_OUT_FLAGS_NEW_BUFFER);
+			break;
+		}
+
+		ret = add_dca_mem(ctx, size);
+		if (ret)
+			break;
+	} while (try_times++ < DCA_EXPAND_MEM_TRY_TIMES);
+
+	if (ret || resp.alloc_pages < buf_pages)
+		return -ENOMEM;
+
+	/* DCA config not changed */
+	if (!is_new_buf && buf_list->bufs[0])
+		return 0;
+
+	return setup_dca_buf_list(ctx, handle, buf_list, buf_pages);
+}
+
+void hns_roce_detach_dca_mem(struct hns_roce_context *ctx, uint32_t handle,
+			     struct hns_roce_dca_detach_attr *attr)
+{
+	detach_dca_mem(ctx, handle, attr);
+}
diff --git a/providers/hns/hns_roce_u_hw_v2.c b/providers/hns/hns_roce_u_hw_v2.c
index 16f033c..d80c83d 100644
--- a/providers/hns/hns_roce_u_hw_v2.c
+++ b/providers/hns/hns_roce_u_hw_v2.c
@@ -227,19 +227,28 @@ static struct hns_roce_v2_cqe *next_cqe_sw_v2(struct hns_roce_cq *cq)
 	return get_sw_cqe_v2(cq, cq->cons_index);
 }
 
+static inline void *get_wqe(struct hns_roce_qp *qp, unsigned int offset)
+{
+	if (qp->page_list.bufs)
+		return qp->page_list.bufs[offset >> qp->page_list.shift] +
+			(offset & ((1 << qp->page_list.shift) - 1));
+	else
+		return qp->buf.buf + offset;
+}
+
 static void *get_recv_wqe_v2(struct hns_roce_qp *qp, unsigned int n)
 {
-	return qp->buf.buf + qp->rq.offset + (n << qp->rq.wqe_shift);
+	return get_wqe(qp, qp->rq.offset + (n << qp->rq.wqe_shift));
 }
 
 static void *get_send_wqe(struct hns_roce_qp *qp, unsigned int n)
 {
-	return qp->buf.buf + qp->sq.offset + (n << qp->sq.wqe_shift);
+	return get_wqe(qp, qp->sq.offset + (n << qp->sq.wqe_shift));
 }
 
 static void *get_send_sge_ex(struct hns_roce_qp *qp, unsigned int n)
 {
-	return qp->buf.buf + qp->ex_sge.offset + (n << qp->ex_sge.sge_shift);
+	return get_wqe(qp, qp->ex_sge.offset + (n << qp->ex_sge.sge_shift));
 }
 
 static void *get_srq_wqe(struct hns_roce_srq *srq, int n)
@@ -502,6 +511,62 @@ static int hns_roce_handle_recv_inl_wqe(struct hns_roce_v2_cqe *cqe,
 	return V2_CQ_OK;
 }
 
+static inline bool check_qp_dca_enable(struct hns_roce_qp *qp)
+{
+	return !!qp->page_list.bufs && (qp->flags & HNS_ROCE_QP_CAP_DCA);
+}
+
+static int dca_attach_qp_buf(struct hns_roce_context *ctx,
+			     struct hns_roce_qp *qp)
+{
+	struct hns_roce_dca_attach_attr attr = {};
+	uint32_t idx;
+
+	pthread_spin_lock(&qp->sq.lock);
+	pthread_spin_lock(&qp->rq.lock);
+
+	if (qp->sq.wqe_cnt > 0) {
+		idx = qp->sq.head & (qp->sq.wqe_cnt - 1);
+		attr.sq_offset = idx << qp->sq.wqe_shift;
+	}
+
+	if (qp->ex_sge.sge_cnt > 0) {
+		idx = qp->next_sge & (qp->ex_sge.sge_cnt - 1);
+		attr.sge_offset = idx << qp->ex_sge.sge_shift;
+	}
+
+	if (qp->rq.wqe_cnt > 0) {
+		idx = qp->rq.head & (qp->rq.wqe_cnt - 1);
+		attr.rq_offset = idx << qp->rq.wqe_shift;
+	}
+
+	pthread_spin_unlock(&qp->rq.lock);
+	pthread_spin_unlock(&qp->sq.lock);
+
+	return hns_roce_attach_dca_mem(ctx, qp->verbs_qp.qp.handle, &attr,
+				       qp->buf_size, &qp->page_list);
+}
+
+static void dca_detach_qp_buf(struct hns_roce_context *ctx,
+			      struct hns_roce_qp *qp)
+{
+	struct hns_roce_dca_detach_attr attr;
+	bool is_empty;
+
+	pthread_spin_lock(&qp->sq.lock);
+	pthread_spin_lock(&qp->rq.lock);
+
+	is_empty = qp->sq.head == qp->sq.tail && qp->rq.head == qp->rq.tail;
+	if (is_empty && qp->sq.wqe_cnt > 0)
+		attr.sq_index = qp->sq.head & (qp->sq.wqe_cnt - 1);
+
+	pthread_spin_unlock(&qp->rq.lock);
+	pthread_spin_unlock(&qp->sq.lock);
+
+	if (is_empty)
+		hns_roce_detach_dca_mem(ctx, qp->verbs_qp.qp.handle, &attr);
+}
+
 static int hns_roce_v2_poll_one(struct hns_roce_cq *cq,
 				struct hns_roce_qp **cur_qp, struct ibv_wc *wc)
 {
@@ -641,6 +706,9 @@ static int hns_roce_u_v2_poll_cq(struct ibv_cq *ibvcq, int ne,
 
 	for (npolled = 0; npolled < ne; ++npolled) {
 		err = hns_roce_v2_poll_one(cq, &qp, wc + npolled);
+		if (qp && check_qp_dca_enable(qp))
+			dca_detach_qp_buf(ctx, qp);
+
 		if (err != V2_CQ_OK)
 			break;
 	}
@@ -690,18 +758,23 @@ static int hns_roce_u_v2_arm_cq(struct ibv_cq *ibvcq, int solicited)
 	return 0;
 }
 
-static int check_qp_send(struct ibv_qp *qp, struct hns_roce_context *ctx)
+static int check_qp_send(struct hns_roce_qp *qp, struct hns_roce_context *ctx)
 {
-	if (unlikely(qp->qp_type != IBV_QPT_RC &&
-		     qp->qp_type != IBV_QPT_UD) &&
-		     qp->qp_type != IBV_QPT_XRC_SEND)
+	struct ibv_qp *ibvqp = &qp->verbs_qp.qp;
+
+	if (unlikely(ibvqp->qp_type != IBV_QPT_RC &&
+		     ibvqp->qp_type != IBV_QPT_UD) &&
+		     ibvqp->qp_type != IBV_QPT_XRC_SEND)
 		return -EINVAL;
 
-	if (unlikely(qp->state == IBV_QPS_RESET ||
-		     qp->state == IBV_QPS_INIT ||
-		     qp->state == IBV_QPS_RTR))
+	if (unlikely(ibvqp->state == IBV_QPS_RESET ||
+		     ibvqp->state == IBV_QPS_INIT ||
+		     ibvqp->state == IBV_QPS_RTR))
 		return -EINVAL;
 
+	if (check_qp_dca_enable(qp))
+		return dca_attach_qp_buf(ctx, qp);
+
 	return 0;
 }
 
@@ -1019,6 +1092,16 @@ static int set_rc_inl(struct hns_roce_qp *qp, const struct ibv_send_wr *wr,
 	return 0;
 }
 
+static inline void fill_rc_dca_fields(uint32_t qp_num,
+				      struct hns_roce_rc_sq_wqe *wqe)
+{
+	roce_set_field(wqe->byte_4, RC_SQ_WQE_BYTE_4_SQPN_L_M,
+		       RC_SQ_WQE_BYTE_4_SQPN_L_S, qp_num);
+	roce_set_field(wqe->byte_4, RC_SQ_WQE_BYTE_4_SQPN_H_M,
+		       RC_SQ_WQE_BYTE_4_SQPN_H_S,
+		       qp_num >> RC_SQ_WQE_BYTE_4_SQPN_L_W);
+}
+
 static void set_bind_mw_seg(struct hns_roce_rc_sq_wqe *wqe,
 			    const struct ibv_send_wr *wr)
 {
@@ -1134,6 +1217,9 @@ static int set_rc_wqe(void *wqe, struct hns_roce_qp *qp, struct ibv_send_wr *wr,
 		return ret;
 
 wqe_valid:
+	if (check_qp_dca_enable(qp))
+		fill_rc_dca_fields(qp->verbs_qp.qp.qp_num, rc_sq_wqe);
+
 	/*
 	 * The pipeline can sequentially post all valid WQEs into WQ buffer,
 	 * including new WQEs waiting for the doorbell to update the PI again.
@@ -1160,7 +1246,7 @@ int hns_roce_u_v2_post_send(struct ibv_qp *ibvqp, struct ibv_send_wr *wr,
 	struct ibv_qp_attr attr;
 	int ret;
 
-	ret = check_qp_send(ibvqp, ctx);
+	ret = check_qp_send(qp, ctx);
 	if (unlikely(ret)) {
 		*bad_wr = wr;
 		return ret;
@@ -1235,15 +1321,20 @@ out:
 	return ret;
 }
 
-static int check_qp_recv(struct ibv_qp *qp, struct hns_roce_context *ctx)
+static int check_qp_recv(struct hns_roce_qp *qp, struct hns_roce_context *ctx)
 {
-	if (unlikely(qp->qp_type != IBV_QPT_RC &&
-		     qp->qp_type != IBV_QPT_UD))
+	struct ibv_qp *ibvqp = &qp->verbs_qp.qp;
+
+	if (unlikely(ibvqp->qp_type != IBV_QPT_RC &&
+		     ibvqp->qp_type != IBV_QPT_UD))
 		return -EINVAL;
 
-	if (qp->state == IBV_QPS_RESET || qp->srq)
+	if (ibvqp->state == IBV_QPS_RESET || ibvqp->srq)
 		return -EINVAL;
 
+	if (check_qp_dca_enable(qp))
+		return dca_attach_qp_buf(ctx, qp);
+
 	return 0;
 }
 
@@ -1286,7 +1377,7 @@ static int hns_roce_u_v2_post_recv(struct ibv_qp *ibvqp, struct ibv_recv_wr *wr,
 	struct ibv_qp_attr attr;
 	int ret;
 
-	ret = check_qp_recv(ibvqp, ctx);
+	ret = check_qp_recv(qp, ctx);
 	if (unlikely(ret)) {
 		*bad_wr = wr;
 		return ret;
diff --git a/providers/hns/hns_roce_u_hw_v2.h b/providers/hns/hns_roce_u_hw_v2.h
index c13d82e..be6be73 100644
--- a/providers/hns/hns_roce_u_hw_v2.h
+++ b/providers/hns/hns_roce_u_hw_v2.h
@@ -239,6 +239,13 @@ struct hns_roce_rc_sq_wqe {
 
 #define RC_SQ_WQE_BYTE_4_RDMA_WRITE_S 22
 
+#define RC_SQ_WQE_BYTE_4_SQPN_L_W 2
+#define RC_SQ_WQE_BYTE_4_SQPN_L_S 5
+#define RC_SQ_WQE_BYTE_4_SQPN_L_M GENMASK(6, 5)
+
+#define RC_SQ_WQE_BYTE_4_SQPN_H_S 13
+#define RC_SQ_WQE_BYTE_4_SQPN_H_M GENMASK(30, 13)
+
 #define RC_SQ_WQE_BYTE_16_XRC_SRQN_S 0
 #define RC_SQ_WQE_BYTE_16_XRC_SRQN_M \
 	(((1UL << 24) - 1) << RC_SQ_WQE_BYTE_16_XRC_SRQN_S)
diff --git a/providers/hns/hns_roce_u_verbs.c b/providers/hns/hns_roce_u_verbs.c
index abff092..21e295a 100644
--- a/providers/hns/hns_roce_u_verbs.c
+++ b/providers/hns/hns_roce_u_verbs.c
@@ -870,6 +870,14 @@ static int calc_qp_buff_size(struct hns_roce_device *hr_dev,
 	return 0;
 }
 
+static inline bool check_qp_support_dca(bool pool_en, enum ibv_qp_type qp_type)
+{
+	if (pool_en && (qp_type == IBV_QPT_RC || qp_type == IBV_QPT_XRC_SEND))
+		return true;
+
+	return false;
+}
+
 static void qp_free_wqe(struct hns_roce_qp *qp)
 {
 	qp_free_recv_inl_buf(qp);
@@ -881,8 +889,8 @@ static void qp_free_wqe(struct hns_roce_qp *qp)
 	hns_roce_free_buf(&qp->buf);
 }
 
-static int qp_alloc_wqe(struct ibv_qp_cap *cap, struct hns_roce_qp *qp,
-			struct hns_roce_context *ctx)
+static int qp_alloc_wqe(struct ibv_qp_init_attr_ex *attr,
+			struct hns_roce_qp *qp, struct hns_roce_context *ctx)
 {
 	struct hns_roce_device *hr_dev = to_hr_dev(ctx->ibv_ctx.context.device);
 
@@ -900,12 +908,24 @@ static int qp_alloc_wqe(struct ibv_qp_cap *cap, struct hns_roce_qp *qp,
 	}
 
 	if (qp->rq_rinl_buf.wqe_cnt) {
-		if (qp_alloc_recv_inl_buf(cap, qp))
+		if (qp_alloc_recv_inl_buf(&attr->cap, qp))
 			goto err_alloc;
 	}
 
-	if (hns_roce_alloc_buf(&qp->buf, qp->buf_size, HNS_HW_PAGE_SIZE))
-		goto err_alloc;
+	if (check_qp_support_dca(ctx->dca_ctx.max_size != 0, attr->qp_type)) {
+		/* when DCA is enabled, use a buffer list to store page addr */
+		qp->buf.buf = NULL;
+		qp->page_list.max_cnt = hr_hw_page_count(qp->buf_size);
+		qp->page_list.shift = HNS_HW_PAGE_SHIFT;
+		qp->page_list.bufs = calloc(qp->page_list.max_cnt,
+					    sizeof(void *));
+		if (!qp->page_list.bufs)
+			goto err_alloc;
+	} else {
+		if (hns_roce_alloc_buf(&qp->buf, qp->buf_size,
+				       HNS_HW_PAGE_SIZE))
+			goto err_alloc;
+	}
 
 	return 0;
 
@@ -1123,7 +1143,7 @@ static int hns_roce_alloc_qp_buf(struct ibv_qp_init_attr_ex *attr,
 	    pthread_spin_init(&qp->rq.lock, PTHREAD_PROCESS_PRIVATE))
 		return -ENOMEM;
 
-	ret = qp_alloc_wqe(&attr->cap, qp, ctx);
+	ret = qp_alloc_wqe(attr, qp, ctx);
 	if (ret)
 		return ret;
 
-- 
2.7.4

[PATCH rdma-core 5/6] libhns: Add direct verbs support to config DCA

[Weihang Li]

From: Xi Wang <wangxi11@huawei.com>

Add two direct verbs to config DCA:
1. hnsdv_open_device() is used to config DCA memory pool.
2. hnsdv_create_qp() is used to create a DCA QP.

Signed-off-by: Xi Wang <wangxi11@huawei.com>
Signed-off-by: Weihang Li <liweihang@huawei.com>
---
 debian/control                             |  2 +-
 debian/ibverbs-providers.install           |  1 +
 debian/ibverbs-providers.lintian-overrides |  4 +-
 debian/ibverbs-providers.symbols           |  6 +++
 debian/libibverbs-dev.install              |  4 ++
 providers/hns/CMakeLists.txt               |  9 +++-
 providers/hns/hns_roce_u.c                 | 71 ++++++++++++++++++++++++++----
 providers/hns/hns_roce_u.h                 |  4 +-
 providers/hns/hns_roce_u_abi.h             |  1 +
 providers/hns/hns_roce_u_verbs.c           | 43 ++++++++++++++----
 providers/hns/hnsdv.h                      | 61 +++++++++++++++++++++++++
 providers/hns/libhns.map                   |  9 ++++
 redhat/rdma-core.spec                      |  5 ++-
 suse/rdma-core.spec                        | 21 ++++++++-
 14 files changed, 218 insertions(+), 23 deletions(-)
 create mode 100644 providers/hns/hnsdv.h
 create mode 100644 providers/hns/libhns.map

diff --git a/debian/control b/debian/control
index 8cbab0b..1ccb7f4 100644
--- a/debian/control
+++ b/debian/control
@@ -94,7 +94,7 @@ Description: User space provider drivers for libibverbs
   - cxgb4: Chelsio T4 iWARP HCAs
   - efa: Amazon Elastic Fabric Adapter
   - hfi1verbs: Intel Omni-Path HFI
-  - hns: HiSilicon Hip06 SoC
+  - hns: HiSilicon Hip06+ SoC
   - i40iw: Intel Ethernet Connection X722 RDMA
   - ipathverbs: QLogic InfiniPath HCAs
   - mlx4: Mellanox ConnectX-3 InfiniBand HCAs
diff --git a/debian/ibverbs-providers.install b/debian/ibverbs-providers.install
index 4f971fb..c6ecbbc 100644
--- a/debian/ibverbs-providers.install
+++ b/debian/ibverbs-providers.install
@@ -1,5 +1,6 @@
 etc/libibverbs.d/
 usr/lib/*/libefa.so.*
 usr/lib/*/libibverbs/lib*-rdmav*.so
+usr/lib/*/libhns.so.*
 usr/lib/*/libmlx4.so.*
 usr/lib/*/libmlx5.so.*
diff --git a/debian/ibverbs-providers.lintian-overrides b/debian/ibverbs-providers.lintian-overrides
index 8a44d54..f6afb70 100644
--- a/debian/ibverbs-providers.lintian-overrides
+++ b/debian/ibverbs-providers.lintian-overrides
@@ -1,2 +1,2 @@
-# libefa, libmlx4 and libmlx5 are ibverbs provider that provides more functions.
-ibverbs-providers: package-name-doesnt-match-sonames libefa1 libmlx4-1 libmlx5-1
+# libefa, libhns, libmlx4 and libmlx5 are ibverbs provider that provides more functions.
+ibverbs-providers: package-name-doesnt-match-sonames libefa1 libhns-1 libmlx4-1 libmlx5-1
diff --git a/debian/ibverbs-providers.symbols b/debian/ibverbs-providers.symbols
index 3c75ecc..34bc9d5 100644
--- a/debian/ibverbs-providers.symbols
+++ b/debian/ibverbs-providers.symbols
@@ -136,3 +136,9 @@ libefa.so.1 ibverbs-providers #MINVER#
  efadv_create_qp_ex@EFA_1.1 26
  efadv_query_device@EFA_1.1 26
  efadv_query_ah@EFA_1.1 26
+libhns.so.1 ibverbs-providers #MINVER#
+* Build-Depends-Package: libibverbs-dev
+ HNS_1.0@HNS_1.0 34
+ hnsdv_is_supported@HNS_1.0 34
+ hnsdv_open_device@HNS_1.0 34
+ hnsdv_create_qp@HNS_1.0 34
diff --git a/debian/libibverbs-dev.install b/debian/libibverbs-dev.install
index bc8caa5..7d6e6a2 100644
--- a/debian/libibverbs-dev.install
+++ b/debian/libibverbs-dev.install
@@ -1,5 +1,6 @@
 usr/include/infiniband/arch.h
 usr/include/infiniband/efadv.h
+usr/include/infiniband/hnsdv.h
 usr/include/infiniband/ib_user_ioctl_verbs.h
 usr/include/infiniband/mlx4dv.h
 usr/include/infiniband/mlx5_api.h
@@ -14,6 +15,8 @@ usr/include/infiniband/verbs_api.h
 usr/lib/*/lib*-rdmav*.a
 usr/lib/*/libefa.a
 usr/lib/*/libefa.so
+usr/lib/*/libhns.a
+usr/lib/*/libhns.so
 usr/lib/*/libibverbs*.so
 usr/lib/*/libibverbs.a
 usr/lib/*/libmlx4.a
@@ -21,6 +24,7 @@ usr/lib/*/libmlx4.so
 usr/lib/*/libmlx5.a
 usr/lib/*/libmlx5.so
 usr/lib/*/pkgconfig/libefa.pc
+usr/lib/*/pkgconfig/libhns.pc
 usr/lib/*/pkgconfig/libibverbs.pc
 usr/lib/*/pkgconfig/libmlx4.pc
 usr/lib/*/pkgconfig/libmlx5.pc
diff --git a/providers/hns/CMakeLists.txt b/providers/hns/CMakeLists.txt
index 697dbd7..6e602f6 100644
--- a/providers/hns/CMakeLists.txt
+++ b/providers/hns/CMakeLists.txt
@@ -1,4 +1,5 @@
-rdma_provider(hns
+rdma_shared_provider(hns libhns.map
+  1 1.0.${PACKAGE_VERSION}
   hns_roce_u.c
   hns_roce_u_buf.c
   hns_roce_u_db.c
@@ -6,3 +7,9 @@ rdma_provider(hns
   hns_roce_u_hw_v2.c
   hns_roce_u_verbs.c
 )
+
+publish_headers(infiniband
+	hnsdv.h
+)
+
+rdma_pkg_config("hns" "libibverbs" "${CMAKE_THREAD_LIBS_INIT}")
diff --git a/providers/hns/hns_roce_u.c b/providers/hns/hns_roce_u.c
index a4e0997..230befe 100644
--- a/providers/hns/hns_roce_u.c
+++ b/providers/hns/hns_roce_u.c
@@ -95,22 +95,69 @@ static const struct verbs_context_ops hns_common_ops = {
 	.get_srq_num = hns_roce_u_get_srq_num,
 };
 
-static int init_dca_context(struct hns_roce_context *ctx, int page_size)
+bool hnsdv_is_supported(struct ibv_device *device)
+{
+	return is_hns_dev(device);
+}
+
+struct ibv_context *hnsdv_open_device(struct ibv_device *device,
+				      struct hnsdv_context_attr *attr)
+{
+	if (!is_hns_dev(device)) {
+		errno = EOPNOTSUPP;
+		return NULL;
+	}
+
+	return verbs_open_device(device, attr);
+}
+
+static void set_dca_pool_param(struct hnsdv_context_attr *attr, int page_size,
+			       struct hns_roce_dca_ctx *ctx)
+{
+	if (attr->comp_mask & HNSDV_CONTEXT_MASK_DCA_UNIT_SIZE)
+		ctx->unit_size = align(attr->dca_unit_size, page_size);
+	else
+		ctx->unit_size = page_size * HNS_DCA_DEFAULT_UNIT_PAGES;
+
+	/* The memory pool cannot be expanded, only init the DCA context. */
+	if (ctx->unit_size == 0)
+		return;
+
+	/* If not set, the memory pool can be expanded unlimitedly. */
+	if (attr->comp_mask & HNSDV_CONTEXT_MASK_DCA_MAX_SIZE)
+		ctx->max_size = DIV_ROUND_UP(attr->dca_max_size,
+					     ctx->unit_size);
+	else
+		ctx->max_size = HNS_DCA_MAX_MEM_SIZE;
+
+	/* If not set, the memory pool cannot be shrunk. */
+	if (attr->comp_mask & HNSDV_CONTEXT_MASK_DCA_MIN_SIZE)
+		ctx->min_size = DIV_ROUND_UP(attr->dca_min_size,
+					     ctx->unit_size);
+	else
+		ctx->min_size = HNS_DCA_MAX_MEM_SIZE;
+}
+
+static int init_dca_context(struct hns_roce_context *ctx, int page_size,
+			    struct hnsdv_context_attr *attr)
 {
 	struct hns_roce_dca_ctx *dca_ctx = &ctx->dca_ctx;
 	int ret;
 
-	if (!(ctx->cap_flags & HNS_ROCE_CAP_FLAG_DCA_MODE))
-		return 0;
-
+	dca_ctx->unit_size = 0;
+	dca_ctx->mem_cnt = 0;
 	list_head_init(&dca_ctx->mem_list);
 	ret = pthread_spin_init(&dca_ctx->lock, PTHREAD_PROCESS_PRIVATE);
 	if (ret)
 		return ret;
 
-	dca_ctx->unit_size = page_size * HNS_DCA_DEFAULT_UNIT_PAGES;
-	dca_ctx->max_size = HNS_DCA_MAX_MEM_SIZE;
-	dca_ctx->mem_cnt = 0;
+	if (!attr)
+		return 0;
+
+	if (!(attr->flags & HNSDV_CONTEXT_FLAGS_DCA))
+		return 0;
+
+	set_dca_pool_param(attr, page_size, dca_ctx);
 
 	return 0;
 }
@@ -133,6 +180,7 @@ static struct verbs_context *hns_roce_alloc_context(struct ibv_device *ibdev,
 						    int cmd_fd,
 						    void *private_data)
 {
+	struct hnsdv_context_attr *ctx_attr = private_data;
 	struct hns_roce_device *hr_dev = to_hr_dev(ibdev);
 	struct hns_roce_alloc_ucontext_resp resp = {};
 	struct ibv_device_attr dev_attrs;
@@ -214,7 +262,7 @@ static struct verbs_context *hns_roce_alloc_context(struct ibv_device *ibdev,
 	verbs_set_ops(&context->ibv_ctx, &hns_common_ops);
 	verbs_set_ops(&context->ibv_ctx, &hr_dev->u_hw->hw_ops);
 
-	if (init_dca_context(context, hr_dev->page_size))
+	if (init_dca_context(context, hr_dev->page_size, ctx_attr))
 		goto tptr_free;
 
 	return &context->ibv_ctx;
@@ -278,4 +326,11 @@ static const struct verbs_device_ops hns_roce_dev_ops = {
 	.uninit_device = hns_uninit_device,
 	.alloc_context = hns_roce_alloc_context,
 };
+
+bool is_hns_dev(struct ibv_device *device)
+{
+	struct verbs_device *verbs_device = verbs_get_device(device);
+
+	return verbs_device->ops == &hns_roce_dev_ops;
+}
 PROVIDER_DRIVER(hns, hns_roce_dev_ops);
diff --git a/providers/hns/hns_roce_u.h b/providers/hns/hns_roce_u.h
index a488694..5c8427a 100644
--- a/providers/hns/hns_roce_u.h
+++ b/providers/hns/hns_roce_u.h
@@ -157,11 +157,11 @@ struct hns_roce_db_page {
 struct hns_roce_dca_ctx {
 	struct list_head mem_list;
 	pthread_spinlock_t lock;
+	uint32_t unit_size;
 	uint64_t max_size;
 	uint64_t min_size;
 	uint64_t curr_size;
 	int mem_cnt;
-	unsigned int unit_size;
 };
 
 struct hns_roce_context {
@@ -391,6 +391,8 @@ static inline struct hns_roce_ah *to_hr_ah(struct ibv_ah *ibv_ah)
 	return container_of(ibv_ah, struct hns_roce_ah, ibv_ah);
 }
 
+bool is_hns_dev(struct ibv_device *device);
+
 int hns_roce_u_query_device(struct ibv_context *context,
 			    const struct ibv_query_device_ex_input *input,
 			    struct ibv_device_attr_ex *attr, size_t attr_size);
diff --git a/providers/hns/hns_roce_u_abi.h b/providers/hns/hns_roce_u_abi.h
index e56f9d3..92404bc 100644
--- a/providers/hns/hns_roce_u_abi.h
+++ b/providers/hns/hns_roce_u_abi.h
@@ -36,6 +36,7 @@
 #include <infiniband/kern-abi.h>
 #include <rdma/hns-abi.h>
 #include <kernel-abi/hns-abi.h>
+#include "hnsdv.h"
 
 DECLARE_DRV_CMD(hns_roce_alloc_pd, IB_USER_VERBS_CMD_ALLOC_PD,
 		empty, hns_roce_ib_alloc_pd_resp);
diff --git a/providers/hns/hns_roce_u_verbs.c b/providers/hns/hns_roce_u_verbs.c
index 21e295a..350a6d2 100644
--- a/providers/hns/hns_roce_u_verbs.c
+++ b/providers/hns/hns_roce_u_verbs.c
@@ -870,9 +870,21 @@ static int calc_qp_buff_size(struct hns_roce_device *hr_dev,
 	return 0;
 }
 
-static inline bool check_qp_support_dca(bool pool_en, enum ibv_qp_type qp_type)
+static inline bool check_qp_support_dca(struct hns_roce_dca_ctx *dca_ctx,
+					struct ibv_qp_init_attr_ex *attr,
+					struct hnsdv_qp_init_attr *hns_attr)
 {
-	if (pool_en && (qp_type == IBV_QPT_RC || qp_type == IBV_QPT_XRC_SEND))
+	/* DCA pool disable */
+	if (!dca_ctx->unit_size)
+		return false;
+
+	/* Unsupport type */
+	if (attr->qp_type != IBV_QPT_RC && attr->qp_type != IBV_QPT_XRC_SEND)
+		return false;
+
+	if (hns_attr &&
+	    (hns_attr->comp_mask & HNSDV_QP_INIT_ATTR_MASK_QP_CREATE_FLAGS) &&
+	    (hns_attr->create_flags & HNSDV_QP_CREATE_DYNAMIC_CONTEXT_ATTACH))
 		return true;
 
 	return false;
@@ -890,6 +902,7 @@ static void qp_free_wqe(struct hns_roce_qp *qp)
 }
 
 static int qp_alloc_wqe(struct ibv_qp_init_attr_ex *attr,
+			struct hnsdv_qp_init_attr *hns_attr,
 			struct hns_roce_qp *qp, struct hns_roce_context *ctx)
 {
 	struct hns_roce_device *hr_dev = to_hr_dev(ctx->ibv_ctx.context.device);
@@ -912,7 +925,7 @@ static int qp_alloc_wqe(struct ibv_qp_init_attr_ex *attr,
 			goto err_alloc;
 	}
 
-	if (check_qp_support_dca(ctx->dca_ctx.max_size != 0, attr->qp_type)) {
+	if (check_qp_support_dca(&ctx->dca_ctx, attr, hns_attr)) {
 		/* when DCA is enabled, use a buffer list to store page addr */
 		qp->buf.buf = NULL;
 		qp->page_list.max_cnt = hr_hw_page_count(qp->buf_size);
@@ -1134,6 +1147,7 @@ void hns_roce_free_qp_buf(struct hns_roce_qp *qp, struct hns_roce_context *ctx)
 }
 
 static int hns_roce_alloc_qp_buf(struct ibv_qp_init_attr_ex *attr,
+				 struct hnsdv_qp_init_attr *hns_attr,
 				 struct hns_roce_qp *qp,
 				 struct hns_roce_context *ctx)
 {
@@ -1143,7 +1157,7 @@ static int hns_roce_alloc_qp_buf(struct ibv_qp_init_attr_ex *attr,
 	    pthread_spin_init(&qp->rq.lock, PTHREAD_PROCESS_PRIVATE))
 		return -ENOMEM;
 
-	ret = qp_alloc_wqe(attr, qp, ctx);
+	ret = qp_alloc_wqe(attr, hns_attr, qp, ctx);
 	if (ret)
 		return ret;
 
@@ -1155,7 +1169,8 @@ static int hns_roce_alloc_qp_buf(struct ibv_qp_init_attr_ex *attr,
 }
 
 static struct ibv_qp *create_qp(struct ibv_context *ibv_ctx,
-				struct ibv_qp_init_attr_ex *attr)
+				struct ibv_qp_init_attr_ex *attr,
+				struct hnsdv_qp_init_attr *hns_attr)
 {
 	struct hns_roce_context *context = to_hr_ctx(ibv_ctx);
 	struct hns_roce_qp *qp;
@@ -1173,7 +1188,7 @@ static struct ibv_qp *create_qp(struct ibv_context *ibv_ctx,
 
 	hns_roce_set_qp_params(attr, qp, context);
 
-	ret = hns_roce_alloc_qp_buf(attr, qp, context);
+	ret = hns_roce_alloc_qp_buf(attr, hns_attr, qp, context);
 	if (ret)
 		goto err_buf;
 
@@ -1213,7 +1228,7 @@ struct ibv_qp *hns_roce_u_create_qp(struct ibv_pd *pd,
 	attrx.comp_mask = IBV_QP_INIT_ATTR_PD;
 	attrx.pd = pd;
 
-	qp = create_qp(pd->context, &attrx);
+	qp = create_qp(pd->context, &attrx, NULL);
 	if (qp)
 		memcpy(attr, &attrx, sizeof(*attr));
 
@@ -1223,7 +1238,19 @@ struct ibv_qp *hns_roce_u_create_qp(struct ibv_pd *pd,
 struct ibv_qp *hns_roce_u_create_qp_ex(struct ibv_context *context,
 				       struct ibv_qp_init_attr_ex *attr)
 {
-	return create_qp(context, attr);
+	return create_qp(context, attr, NULL);
+}
+
+struct ibv_qp *hnsdv_create_qp(struct ibv_context *context,
+			       struct ibv_qp_init_attr_ex *qp_attr,
+			       struct hnsdv_qp_init_attr *hns_attr)
+{
+	if (!is_hns_dev(context->device)) {
+		errno = EOPNOTSUPP;
+		return NULL;
+	}
+
+	return create_qp(context, qp_attr, hns_attr);
 }
 
 struct ibv_qp *hns_roce_u_open_qp(struct ibv_context *context,
diff --git a/providers/hns/hnsdv.h b/providers/hns/hnsdv.h
new file mode 100644
index 0000000..876183b
--- /dev/null
+++ b/providers/hns/hnsdv.h
@@ -0,0 +1,61 @@
+/* SPDX-License-Identifier: GPL-2.0 OR BSD-2-Clause */
+/*
+ * Copyright (c) 2021 Hisilicon Limited.
+ */
+
+#ifndef __HNSDV_H__
+#define __HNSDV_H__
+
+#include <stdio.h>
+#include <sys/types.h>
+
+#include <infiniband/verbs.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+enum hnsdv_context_attr_flags {
+	HNSDV_CONTEXT_FLAGS_DCA = 1 << 0,
+};
+
+enum hnsdv_context_comp_mask {
+	HNSDV_CONTEXT_MASK_DCA_UNIT_SIZE = 1 << 0,
+	HNSDV_CONTEXT_MASK_DCA_MAX_SIZE = 1 << 1,
+	HNSDV_CONTEXT_MASK_DCA_MIN_SIZE = 1 << 2,
+};
+
+struct hnsdv_context_attr {
+	uint32_t flags; /* Use enum hnsdv_context_attr_flags */
+	uint64_t comp_mask; /* Use enum hnsdv_context_comp_mask */
+	uint32_t dca_unit_size;
+	uint64_t dca_max_size;
+	uint64_t dca_min_size;
+};
+
+bool hnsdv_is_supported(struct ibv_device *device);
+struct ibv_context *hnsdv_open_device(struct ibv_device *device,
+				      struct hnsdv_context_attr *attr);
+
+enum hnsdv_qp_create_flags {
+	HNSDV_QP_CREATE_DYNAMIC_CONTEXT_ATTACH = 1 << 0,
+};
+
+enum hnsdv_qp_init_attr_mask {
+	HNSDV_QP_INIT_ATTR_MASK_QP_CREATE_FLAGS	= 1 << 0,
+};
+
+struct hnsdv_qp_init_attr {
+	uint64_t comp_mask;	/* Use enum hnsdv_qp_init_attr_mask */
+	uint32_t create_flags;	/* Use enum hnsdv_qp_create_flags */
+};
+
+struct ibv_qp *hnsdv_create_qp(struct ibv_context *context,
+			       struct ibv_qp_init_attr_ex *qp_attr,
+			       struct hnsdv_qp_init_attr *hns_qp_attr);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* __HNSDV_H__ */
diff --git a/providers/hns/libhns.map b/providers/hns/libhns.map
new file mode 100644
index 0000000..aed491c
--- /dev/null
+++ b/providers/hns/libhns.map
@@ -0,0 +1,9 @@
+/* Export symbols should be added below according to
+   Documentation/versioning.md document. */
+HNS_1.0 {
+	global:
+		hnsdv_is_supported;
+		hnsdv_open_device;
+		hnsdv_create_qp;
+	local: *;
+};
diff --git a/redhat/rdma-core.spec b/redhat/rdma-core.spec
index 207859d..e1dda8f 100644
--- a/redhat/rdma-core.spec
+++ b/redhat/rdma-core.spec
@@ -151,6 +151,8 @@ Provides: libefa = %{version}-%{release}
 Obsoletes: libefa < %{version}-%{release}
 Provides: libhfi1 = %{version}-%{release}
 Obsoletes: libhfi1 < %{version}-%{release}
+Provides: libhns = %{version}-%{release}
+Obsoletes: libhns < %{version}-%{release}
 Provides: libi40iw = %{version}-%{release}
 Obsoletes: libi40iw < %{version}-%{release}
 Provides: libipathverbs = %{version}-%{release}
@@ -178,7 +180,7 @@ Device-specific plug-in ibverbs userspace drivers are included:
 - libcxgb4: Chelsio T4 iWARP HCA
 - libefa: Amazon Elastic Fabric Adapter
 - libhfi1: Intel Omni-Path HFI
-- libhns: HiSilicon Hip06 SoC
+- libhns: HiSilicon Hip06+ SoC
 - libi40iw: Intel Ethernet Connection X722 RDMA
 - libipathverbs: QLogic InfiniPath HCA
 - libmlx4: Mellanox ConnectX-3 InfiniBand HCA
@@ -563,6 +565,7 @@ fi
 %dir %{_sysconfdir}/libibverbs.d
 %dir %{_libdir}/libibverbs
 %{_libdir}/libefa.so.*
+%{_libdir}/libhns.so.*
 %{_libdir}/libibverbs*.so.*
 %{_libdir}/libibverbs/*.so
 %{_libdir}/libmlx5.so.*
diff --git a/suse/rdma-core.spec b/suse/rdma-core.spec
index db6a361..1c14773 100644
--- a/suse/rdma-core.spec
+++ b/suse/rdma-core.spec
@@ -30,6 +30,7 @@ License:        GPL-2.0-only OR BSD-2-Clause
 Group:          Productivity/Networking/Other
 
 %define efa_so_major    1
+%define hns_so_major    1
 %define verbs_so_major  1
 %define rdmacm_so_major 1
 %define umad_so_major   3
@@ -39,6 +40,7 @@ Group:          Productivity/Networking/Other
 %define mad_major       5
 
 %define  efa_lname    libefa%{efa_so_major}
+%define  hns_lname    libhns%{hns_so_major}
 %define  verbs_lname  libibverbs%{verbs_so_major}
 %define  rdmacm_lname librdmacm%{rdmacm_so_major}
 %define  umad_lname   libibumad%{umad_so_major}
@@ -145,6 +147,7 @@ Requires:       %{umad_lname} = %{version}-%{release}
 Requires:       %{verbs_lname} = %{version}-%{release}
 %if 0%{?dma_coherent}
 Requires:       %{efa_lname} = %{version}-%{release}
+Requires:       %{hns_lname} = %{version}-%{release}
 Requires:       %{mlx4_lname} = %{version}-%{release}
 Requires:       %{mlx5_lname} = %{version}-%{release}
 %endif
@@ -185,6 +188,7 @@ Requires:       %{name}%{?_isa} = %{version}-%{release}
 Obsoletes:      libcxgb4-rdmav2 < %{version}-%{release}
 Obsoletes:      libefa-rdmav2 < %{version}-%{release}
 Obsoletes:      libhfi1verbs-rdmav2 < %{version}-%{release}
+Obsoletes:      libhns-rdmav2 < %{version}-%{release}
 Obsoletes:      libi40iw-rdmav2 < %{version}-%{release}
 Obsoletes:      libipathverbs-rdmav2 < %{version}-%{release}
 Obsoletes:      libmlx4-rdmav2 < %{version}-%{release}
@@ -194,6 +198,7 @@ Obsoletes:      libocrdma-rdmav2 < %{version}-%{release}
 Obsoletes:      librxe-rdmav2 < %{version}-%{release}
 %if 0%{?dma_coherent}
 Requires:       %{efa_lname} = %{version}-%{release}
+Requires:       %{hns_lname} = %{version}-%{release}
 Requires:       %{mlx4_lname} = %{version}-%{release}
 Requires:       %{mlx5_lname} = %{version}-%{release}
 %endif
@@ -212,7 +217,7 @@ Device-specific plug-in ibverbs userspace drivers are included:
 - libcxgb4: Chelsio T4 iWARP HCA
 - libefa: Amazon Elastic Fabric Adapter
 - libhfi1: Intel Omni-Path HFI
-- libhns: HiSilicon Hip06 SoC
+- libhns: HiSilicon Hip06+ SoC
 - libi40iw: Intel Ethernet Connection X722 RDMA
 - libipathverbs: QLogic InfiniPath HCA
 - libmlx4: Mellanox ConnectX-3 InfiniBand HCA
@@ -239,6 +244,13 @@ Group:          System/Libraries
 %description -n %efa_lname
 This package contains the efa runtime library.
 
+%package -n %hns_lname
+Summary:        HNS runtime library
+Group:          System/Libraries
+
+%description -n %hns_lname
+This package contains the hns runtime library.
+
 %package -n %mlx4_lname
 Summary:        MLX4 runtime library
 Group:          System/Libraries
@@ -482,6 +494,9 @@ rm -rf %{buildroot}/%{_sbindir}/srp_daemon.sh
 %post -n %efa_lname -p /sbin/ldconfig
 %postun -n %efa_lname -p /sbin/ldconfig
 
+%post -n %hns_lname -p /sbin/ldconfig
+%postun -n %hns_lname -p /sbin/ldconfig
+
 %post -n %mlx4_lname -p /sbin/ldconfig
 %postun -n %mlx4_lname -p /sbin/ldconfig
 
@@ -664,6 +679,10 @@ rm -rf %{buildroot}/%{_sbindir}/srp_daemon.sh
 %defattr(-,root,root)
 %{_libdir}/libefa*.so.*
 
+%files -n %hns_lname
+%defattr(-,root,root)
+%{_libdir}/libhns*.so.*
+
 %files -n %mlx4_lname
 %defattr(-,root,root)
 %{_libdir}/libmlx4*.so.*
-- 
2.7.4

[PATCH rdma-core 6/6] libhns: Add man pages to introduce DCA feature

[Weihang Li]

From: Xi Wang <wangxi11@huawei.com>

Document hns DCA feature and related direct verbs.

Signed-off-by: Xi Wang <wangxi11@huawei.com>
Signed-off-by: Weihang Li <liweihang@huawei.com>
---
 CMakeLists.txt                            |  1 +
 debian/ibverbs-providers.install          |  2 +-
 debian/libibverbs-dev.install             |  2 +
 providers/hns/man/CMakeLists.txt          |  7 ++++
 providers/hns/man/hns_dca.7.md            | 35 ++++++++++++++++
 providers/hns/man/hnsdv.7.md              | 34 +++++++++++++++
 providers/hns/man/hnsdv_create_qp.3.md    | 69 ++++++++++++++++++++++++++++++
 providers/hns/man/hnsdv_is_supported.3.md | 39 +++++++++++++++++
 providers/hns/man/hnsdv_open_device.3.md  | 70 +++++++++++++++++++++++++++++++
 redhat/rdma-core.spec                     |  2 +
 10 files changed, 260 insertions(+), 1 deletion(-)
 create mode 100644 providers/hns/man/CMakeLists.txt
 create mode 100644 providers/hns/man/hns_dca.7.md
 create mode 100644 providers/hns/man/hnsdv.7.md
 create mode 100644 providers/hns/man/hnsdv_create_qp.3.md
 create mode 100644 providers/hns/man/hnsdv_is_supported.3.md
 create mode 100644 providers/hns/man/hnsdv_open_device.3.md

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 74293bf..744179d 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -669,6 +669,7 @@ add_subdirectory(providers/cxgb4) # NO SPARSE
 add_subdirectory(providers/efa)
 add_subdirectory(providers/efa/man)
 add_subdirectory(providers/hns)
+add_subdirectory(providers/hns/man)
 add_subdirectory(providers/i40iw) # NO SPARSE
 add_subdirectory(providers/mlx4)
 add_subdirectory(providers/mlx4/man)
diff --git a/debian/ibverbs-providers.install b/debian/ibverbs-providers.install
index c6ecbbc..c4c4c11 100644
--- a/debian/ibverbs-providers.install
+++ b/debian/ibverbs-providers.install
@@ -1,6 +1,6 @@
 etc/libibverbs.d/
 usr/lib/*/libefa.so.*
-usr/lib/*/libibverbs/lib*-rdmav*.so
 usr/lib/*/libhns.so.*
+usr/lib/*/libibverbs/lib*-rdmav*.so
 usr/lib/*/libmlx4.so.*
 usr/lib/*/libmlx5.so.*
diff --git a/debian/libibverbs-dev.install b/debian/libibverbs-dev.install
index 7d6e6a2..89a02a8 100644
--- a/debian/libibverbs-dev.install
+++ b/debian/libibverbs-dev.install
@@ -29,11 +29,13 @@ usr/lib/*/pkgconfig/libibverbs.pc
 usr/lib/*/pkgconfig/libmlx4.pc
 usr/lib/*/pkgconfig/libmlx5.pc
 usr/share/man/man3/efadv_*.3
+usr/share/man/man3/hns*.3
 usr/share/man/man3/ibv_*
 usr/share/man/man3/mbps_to_ibv_rate.3
 usr/share/man/man3/mlx4dv_*.3
 usr/share/man/man3/mlx5dv_*.3
 usr/share/man/man3/mult_to_ibv_rate.3
 usr/share/man/man7/efadv.7
+usr/share/man/man3/hns*.7
 usr/share/man/man7/mlx4dv.7
 usr/share/man/man7/mlx5dv.7
diff --git a/providers/hns/man/CMakeLists.txt b/providers/hns/man/CMakeLists.txt
new file mode 100644
index 0000000..b375a65
--- /dev/null
+++ b/providers/hns/man/CMakeLists.txt
@@ -0,0 +1,7 @@
+rdma_man_pages(
+  hnsdv_is_supported.3.md
+  hnsdv_open_device.3.md
+  hnsdv_create_qp.3.md
+  hnsdv.7
+  hns_dca.7
+)
diff --git a/providers/hns/man/hns_dca.7.md b/providers/hns/man/hns_dca.7.md
new file mode 100644
index 0000000..de26d07
--- /dev/null
+++ b/providers/hns/man/hns_dca.7.md
@@ -0,0 +1,35 @@
+---
+layout: page
+title: DCA
+section: 7
+tagline: DCA
+date: 2021-03-03
+header: "HNS DCA Manual"
+footer: hns
+---
+
+# NAME
+
+DCA - Dynamic Context Attachment
+
+This allows all WQEs to share a memory pool that belongs to the user context.
+
+# DESCRIPTION
+
+The DCA feature aims to reduce memory consumption by sharing WQE memory for QPs working in sparse traffic scenarios.
+
+The DCA memory pool consists of multiple umem objects. Each umem object is a buffer allocated in user driver and register into kernel driver. The ULP need to setup the memory pool's parameter by calling hnsdv_open_device() and the driver will expand or shrink the memory pool based on this parameter.
+
+When a QP's DCA was enabled by setting create flags through ibv_create_qp_ex(), the WQE buffer will not be allocated directly until the ULP invokes the ibv_post_xxx(). If the memory in the pool is insufficient and the capacity expansion conditions are met, the driver will add new umem objects to the pool.
+
+When all WQEs of a QP are not used by the ROCEE after ibv_poll_cq() or ibv_modify_qp() are invoked, the WQE buffer will be reclaimed to the DCA memory pool. If the free memory in the pool meets the shrink conditions, the driver will delete the unused umem object.
+
+# SEE ALSO
+
+*hnsdv_open_device(3)*, *hnsdv_create_qp(3)*
+
+# AUTHORS
+
+Xi Wang <wangxi11@huawei.com>
+
+Weihang Li <liweihang@huawei.com>
diff --git a/providers/hns/man/hnsdv.7.md b/providers/hns/man/hnsdv.7.md
new file mode 100644
index 0000000..ada73ec
--- /dev/null
+++ b/providers/hns/man/hnsdv.7.md
@@ -0,0 +1,34 @@
+---
+layout: page
+title: HNSDV
+section: 7
+tagline: Verbs
+date: 2021-03-03
+header: "HNS Direct Verbs Manual"
+footer: hns
+---
+
+# NAME
+
+hnsdv - Direct verbs for hns devices
+
+This provides low level access to hns devices to perform direct operations,
+without general branching performed by libibverbs.
+
+# DESCRIPTION
+
+The libibverbs API is an abstract one. It is agnostic to any underlying provider specific implementation. While this abstraction has the advantage of user applications portability, it has a performance penalty. For some applications optimizing performance is more important than portability.
+
+The hns direct verbs API is intended for such applications. It exposes hns specific low level operations, allowing the application to bypass the libibverbs API.
+
+The direct include of hnsdv.h together with linkage to hns library will allow usage of this new interface.
+
+# SEE ALSO
+
+**verbs**(7)
+
+# AUTHORS
+
+Xi Wang <wangxi11@huawei.com>
+
+Weihang Li <liweihang@huawei.com>
diff --git a/providers/hns/man/hnsdv_create_qp.3.md b/providers/hns/man/hnsdv_create_qp.3.md
new file mode 100644
index 0000000..57446e9
--- /dev/null
+++ b/providers/hns/man/hnsdv_create_qp.3.md
@@ -0,0 +1,69 @@
+---
+layout: page
+title: hnsdv_create_qp
+section: 3
+tagline: Verbs
+date: 2021-3-15
+header: "hns Programmer's Manual"
+footer: hns
+---
+
+# NAME
+
+hnsdv_create_qp - creates a queue pair (QP)
+
+# SYNOPSIS
+
+```c
+#include <infiniband/hnsdv.h>
+
+struct ibv_qp *hnsdv_create_qp(struct ibv_context *context,
+			       struct ibv_qp_init_attr_ex *attr,
+			       struct hnsdv_qp_init_attr *hns_attr)
+```
+
+
+# DESCRIPTION
+
+**hnsdv_create_qp()** creates a queue pair (QP) with specific driver properties.
+
+# ARGUMENTS
+
+Please see *ibv_create_qp_ex(3)* man page for *context* and *attr*.
+
+## hns_attr
+
+```c
+struct hnsdv_qp_init_attr {
+	uint64_t comp_mask;
+	uint32_t create_flags;
+};
+```
+
+*comp_mask*
+:	Bitmask specifying what fields in the structure are valid:
+	HNSDV_QP_INIT_ATTR_MASK_QP_CREATE_FLAGS:
+		valid values in *create_flags*
+
+*create_flags*
+:	A bitwise OR of the various values described below.
+
+	HNSDV_QP_CREATE_DYNAMIC_CONTEXT_ATTACH :
+		Enable DCA feature for QP, the WQE buffer will allocate
+		from DCA memory pool when calling ibv_post_send() or
+		ibv_post_recv().
+
+# RETURN VALUE
+
+**hnsdv_create_qp()**
+returns a pointer to the created QP, on error NULL will be returned and errno will be set.
+
+# SEE ALSO
+
+**ibv_create_qp_ex**(3),
+
+# AUTHOR
+
+Xi Wang <wangxi11@huawei.com>
+
+Weihang Li <liweihang@huawei.com>
diff --git a/providers/hns/man/hnsdv_is_supported.3.md b/providers/hns/man/hnsdv_is_supported.3.md
new file mode 100644
index 0000000..b5f00bd
--- /dev/null
+++ b/providers/hns/man/hnsdv_is_supported.3.md
@@ -0,0 +1,39 @@
+---
+layout: page
+title: hnsdv_is_supported
+section: 3
+tagline: Verbs
+---
+
+# NAME
+
+hnsdv_is_supported - Check whether an RDMA device implemented by the hns provider
+
+# SYNOPSIS
+
+```c
+#include <infiniband/hnsdv.h>
+
+bool hnsdv_is_supported(struct ibv_device *device);
+```
+
+# DESCRIPTION
+
+hnsdv functions may be called only if this function returns true for the RDMA device.
+
+# ARGUMENTS
+
+*device*
+:	RDMA device to check.
+
+# RETURN VALUE
+
+Returns true if device is implemented by hns provider.
+
+# SEE ALSO
+
+*hnsdv(7)*
+
+# AUTHOR
+
+Xi Wang <wangxi11@huawei.com>
diff --git a/providers/hns/man/hnsdv_open_device.3.md b/providers/hns/man/hnsdv_open_device.3.md
new file mode 100644
index 0000000..c05ce5d
--- /dev/null
+++ b/providers/hns/man/hnsdv_open_device.3.md
@@ -0,0 +1,70 @@
+---
+layout: page
+title: hnsdv_open_device
+section: 3
+tagline: Verbs
+---
+
+# NAME
+
+hnsdv_open_device - Open an RDMA device context for the hns provider
+
+# SYNOPSIS
+
+```c
+#include <infiniband/hnsdv.h>
+
+struct ibv_context *
+hnsdv_open_device(struct ibv_device *device, struct hnsdv_context_attr *attr);
+```
+
+# DESCRIPTION
+
+Open an RDMA device context with specific hns provider attributes.
+
+# ARGUMENTS
+
+*device*
+:	RDMA device to open.
+
+## *attr* argument
+
+```c
+struct hnsdv_context_attr {
+        uint32_t flags;
+        uint64_t comp_mask;
+	uint32_t dca_unit_size;
+	uint64_t dca_max_size;
+	uint64_t dca_min_size;
+};
+```
+
+*flags*
+:       A bitwise OR of the various values described below.
+
+        *HNSDV_CONTEXT_FLAGS_DCA*:
+        Create a DCA memory pool to support all QPs share it.
+
+*comp_mask*
+:       Bitmask specifying what fields in the structure are valid
+
+*dca_unit_size*
+:       The unit size when adding a new buffer to DCA memory pool.
+
+*dca_max_size*
+:       The DCA pool will be expanded when the total size is smaller than maximal size.
+
+*dca_min_size*
+:       The DCA pool will be shrunk when the free size is bigger than minimal size.
+
+# RETURN VALUE
+
+Returns a pointer to the allocated device context, or NULL if the request fails.
+
+# SEE ALSO
+
+*hnsdv_create_qp(3)*, *hns_dca(7)*
+
+# AUTHOR
+
+Xi Wang <wangxi11@huawei.com>
diff --git a/redhat/rdma-core.spec b/redhat/rdma-core.spec
index e1dda8f..fcbec50 100644
--- a/redhat/rdma-core.spec
+++ b/redhat/rdma-core.spec
@@ -440,6 +440,7 @@ fi
 %{_libdir}/lib*.so
 %{_libdir}/pkgconfig/*.pc
 %{_mandir}/man3/efadv*
+%{_mandir}/man3/hns*
 %{_mandir}/man3/ibv_*
 %{_mandir}/man3/rdma*
 %{_mandir}/man3/umad*
@@ -448,6 +449,7 @@ fi
 %{_mandir}/man3/mlx5dv*
 %{_mandir}/man3/mlx4dv*
 %{_mandir}/man7/efadv*
+%{_mandir}/man7/hns*
 %{_mandir}/man7/mlx5dv*
 %{_mandir}/man7/mlx4dv*
 %{_mandir}/man3/ibnd_*
-- 
2.7.4

Re: [PATCH rdma-core 0/6] libhns: Add support for Dynamic Context Attachment

[Jason Gunthorpe]

On Mon, May 10, 2021 at 09:12:58PM +0800, Weihang Li wrote:
> The HIP09 introduces the DCA(Dynamic Context Attachment) feature which
> supports many RC QPs to share the WQE buffer in a memory pool. If a QP
> enables DCA feature, the WQE's buffer will not be allocated when creating
> but when the users start to post WRs. This will reduce the memory
> consumption when there are too many QPs are inactive.
> 
> For more detailed information, please refer to the man pages provided by
> the last patch of this series.
> 
> This series is associated with the kernel one "RDMA/hns: Add support for
> Dynamic Context Attachment", and two RFC versions of this series has been
> sent before.
> 
> No changes since RFC v2.
> * Link: https://patchwork.kernel.org/project/linux-rdma/cover/1614847759-33139-1-git-send-email-liweihang@huawei.com/
> 
> Changes since RFC v1:
> * Add direct verbs to set the parameters about size that used to
>   configuring DCA. 
> * Add man pages to explain what is DCA, how does it works and how to use
>   it.
> * Link: https://patchwork.kernel.org/project/linux-rdma/cover/1612667574-56673-1-git-send-email-liweihang@huawei.com/
> 
> Weihang Li (1):
>   Update kernel headers
> 
> Xi Wang (5):
>   libhns: Introduce DCA for RC QP
>   libhns: Add support for shrinking DCA memory pool
>   libhns: Add support for attaching QP's WQE buffer
>   libhns: Add direct verbs support to config DCA
>   libhns: Add man pages to introduce DCA feature

This should be sent on GitHub

Jason

Re: [PATCH rdma-core 0/6] libhns: Add support for Dynamic Context Attachment

[liweihang]

On 2021/6/4 22:40, Jason Gunthorpe wrote:
> On Mon, May 10, 2021 at 09:12:58PM +0800, Weihang Li wrote:
>> The HIP09 introduces the DCA(Dynamic Context Attachment) feature which
>> supports many RC QPs to share the WQE buffer in a memory pool. If a QP
>> enables DCA feature, the WQE's buffer will not be allocated when creating
>> but when the users start to post WRs. This will reduce the memory
>> consumption when there are too many QPs are inactive.
>>
>> For more detailed information, please refer to the man pages provided by
>> the last patch of this series.
>>
>> This series is associated with the kernel one "RDMA/hns: Add support for
>> Dynamic Context Attachment", and two RFC versions of this series has been
>> sent before.
>>
>> No changes since RFC v2.
>> * Link: https://patchwork.kernel.org/project/linux-rdma/cover/1614847759-33139-1-git-send-email-liweihang@huawei.com/
>>
>> Changes since RFC v1:
>> * Add direct verbs to set the parameters about size that used to
>>   configuring DCA. 
>> * Add man pages to explain what is DCA, how does it works and how to use
>>   it.
>> * Link: https://patchwork.kernel.org/project/linux-rdma/cover/1612667574-56673-1-git-send-email-liweihang@huawei.com/
>>
>> Weihang Li (1):
>>   Update kernel headers
>>
>> Xi Wang (5):
>>   libhns: Introduce DCA for RC QP
>>   libhns: Add support for shrinking DCA memory pool
>>   libhns: Add support for attaching QP's WQE buffer
>>   libhns: Add direct verbs support to config DCA
>>   libhns: Add man pages to introduce DCA feature
> 
> This should be sent on GitHub
> 
> Jason
> 

The implementation needs some modification, so I haven't sent it to github yet.
The next version will be sent to github at the same time, thank you.

Weihang