* [Patch v2 1/3] RDMA/mana_ib: register RDMA device with GDMA
2023-12-04 23:02 [Patch v2 0/3] Register with RDMA SOC interface and support for CQ longli
@ 2023-12-04 23:02 ` longli
2023-12-04 23:02 ` [Patch v2 2/3] RDMA/mana_ib: query device capabilities longli
2023-12-04 23:02 ` [Patch v2 3/3] RDMA/mana_ib: Add CQ interrupt support for RAW QP longli
2 siblings, 0 replies; 7+ messages in thread
From: longli @ 2023-12-04 23:02 UTC (permalink / raw)
To: Jason Gunthorpe, Leon Romanovsky, Ajay Sharma, Dexuan Cui,
K. Y. Srinivasan, Haiyang Zhang, Wei Liu, David S. Miller,
Eric Dumazet, Jakub Kicinski, Paolo Abeni
Cc: linux-rdma, linux-hyperv, netdev, linux-kernel, Long Li
From: Long Li <longli@microsoft.com>
Software client needs to register with the RDMA management interface on
the SoC to access more features, including querying device capabilities
and RC queue pair.
Signed-off-by: Long Li <longli@microsoft.com>
---
drivers/infiniband/hw/mana/device.c | 24 +++++++++++++++----
drivers/infiniband/hw/mana/main.c | 4 ++--
drivers/infiniband/hw/mana/qp.c | 15 ++++++------
.../net/ethernet/microsoft/mana/gdma_main.c | 5 ++++
include/net/mana/gdma.h | 4 ++++
5 files changed, 38 insertions(+), 14 deletions(-)
diff --git a/drivers/infiniband/hw/mana/device.c b/drivers/infiniband/hw/mana/device.c
index d4541b8707e4..fe025e13a45c 100644
--- a/drivers/infiniband/hw/mana/device.c
+++ b/drivers/infiniband/hw/mana/device.c
@@ -68,7 +68,6 @@ static int mana_ib_probe(struct auxiliary_device *adev,
ibdev_dbg(&dev->ib_dev, "mdev=%p id=%d num_ports=%d\n", mdev,
mdev->dev_id.as_uint32, dev->ib_dev.phys_port_cnt);
- dev->gdma_dev = mdev;
dev->ib_dev.node_type = RDMA_NODE_IB_CA;
/*
@@ -78,16 +77,28 @@ static int mana_ib_probe(struct auxiliary_device *adev,
dev->ib_dev.num_comp_vectors = 1;
dev->ib_dev.dev.parent = mdev->gdma_context->dev;
- ret = ib_register_device(&dev->ib_dev, "mana_%d",
- mdev->gdma_context->dev);
+ ret = mana_gd_register_device(&mdev->gdma_context->mana_ib);
if (ret) {
- ib_dealloc_device(&dev->ib_dev);
- return ret;
+ ibdev_err(&dev->ib_dev, "Failed to register device, ret %d",
+ ret);
+ goto free_ib_device;
}
+ dev->gdma_dev = &mdev->gdma_context->mana_ib;
+
+ ret = ib_register_device(&dev->ib_dev, "mana_%d",
+ mdev->gdma_context->dev);
+ if (ret)
+ goto deregister_device;
dev_set_drvdata(&adev->dev, dev);
return 0;
+
+deregister_device:
+ mana_gd_deregister_device(dev->gdma_dev);
+free_ib_device:
+ ib_dealloc_device(&dev->ib_dev);
+ return ret;
}
static void mana_ib_remove(struct auxiliary_device *adev)
@@ -95,6 +106,9 @@ static void mana_ib_remove(struct auxiliary_device *adev)
struct mana_ib_dev *dev = dev_get_drvdata(&adev->dev);
ib_unregister_device(&dev->ib_dev);
+
+ mana_gd_deregister_device(dev->gdma_dev);
+
ib_dealloc_device(&dev->ib_dev);
}
diff --git a/drivers/infiniband/hw/mana/main.c b/drivers/infiniband/hw/mana/main.c
index 7be4c3adb4e2..53730306ed9b 100644
--- a/drivers/infiniband/hw/mana/main.c
+++ b/drivers/infiniband/hw/mana/main.c
@@ -8,7 +8,7 @@
void mana_ib_uncfg_vport(struct mana_ib_dev *dev, struct mana_ib_pd *pd,
u32 port)
{
- struct gdma_dev *gd = dev->gdma_dev;
+ struct gdma_dev *gd = &dev->gdma_dev->gdma_context->mana;
struct mana_port_context *mpc;
struct net_device *ndev;
struct mana_context *mc;
@@ -31,7 +31,7 @@ void mana_ib_uncfg_vport(struct mana_ib_dev *dev, struct mana_ib_pd *pd,
int mana_ib_cfg_vport(struct mana_ib_dev *dev, u32 port, struct mana_ib_pd *pd,
u32 doorbell_id)
{
- struct gdma_dev *mdev = dev->gdma_dev;
+ struct gdma_dev *mdev = &dev->gdma_dev->gdma_context->mana;
struct mana_port_context *mpc;
struct mana_context *mc;
struct net_device *ndev;
diff --git a/drivers/infiniband/hw/mana/qp.c b/drivers/infiniband/hw/mana/qp.c
index 4b3b5b274e84..ae45d28eef5e 100644
--- a/drivers/infiniband/hw/mana/qp.c
+++ b/drivers/infiniband/hw/mana/qp.c
@@ -21,8 +21,8 @@ static int mana_ib_cfg_vport_steering(struct mana_ib_dev *dev,
u32 req_buf_size;
int i, err;
- mdev = dev->gdma_dev;
- gc = mdev->gdma_context;
+ gc = dev->gdma_dev->gdma_context;
+ mdev = &gc->mana;
req_buf_size =
sizeof(*req) + sizeof(mana_handle_t) * MANA_INDIRECT_TABLE_SIZE;
@@ -102,20 +102,21 @@ static int mana_ib_create_qp_rss(struct ib_qp *ibqp, struct ib_pd *pd,
struct ib_rwq_ind_table *ind_tbl = attr->rwq_ind_tbl;
struct mana_ib_create_qp_rss_resp resp = {};
struct mana_ib_create_qp_rss ucmd = {};
- struct gdma_dev *gd = mdev->gdma_dev;
mana_handle_t *mana_ind_table;
struct mana_port_context *mpc;
+ unsigned int ind_tbl_size;
struct mana_context *mc;
struct net_device *ndev;
struct mana_ib_cq *cq;
struct mana_ib_wq *wq;
- unsigned int ind_tbl_size;
+ struct gdma_dev *gd;
struct ib_cq *ibcq;
struct ib_wq *ibwq;
int i = 0;
u32 port;
int ret;
+ gd = &mdev->gdma_dev->gdma_context->mana;
mc = gd->driver_data;
if (!udata || udata->inlen < sizeof(ucmd))
@@ -266,8 +267,8 @@ static int mana_ib_create_qp_raw(struct ib_qp *ibqp, struct ib_pd *ibpd,
struct mana_ib_ucontext *mana_ucontext =
rdma_udata_to_drv_context(udata, struct mana_ib_ucontext,
ibucontext);
+ struct gdma_dev *gd = &mdev->gdma_dev->gdma_context->mana;
struct mana_ib_create_qp_resp resp = {};
- struct gdma_dev *gd = mdev->gdma_dev;
struct mana_ib_create_qp ucmd = {};
struct mana_obj_spec wq_spec = {};
struct mana_obj_spec cq_spec = {};
@@ -437,7 +438,7 @@ static int mana_ib_destroy_qp_rss(struct mana_ib_qp *qp,
{
struct mana_ib_dev *mdev =
container_of(qp->ibqp.device, struct mana_ib_dev, ib_dev);
- struct gdma_dev *gd = mdev->gdma_dev;
+ struct gdma_dev *gd = &mdev->gdma_dev->gdma_context->mana;
struct mana_port_context *mpc;
struct mana_context *mc;
struct net_device *ndev;
@@ -464,7 +465,7 @@ static int mana_ib_destroy_qp_raw(struct mana_ib_qp *qp, struct ib_udata *udata)
{
struct mana_ib_dev *mdev =
container_of(qp->ibqp.device, struct mana_ib_dev, ib_dev);
- struct gdma_dev *gd = mdev->gdma_dev;
+ struct gdma_dev *gd = &mdev->gdma_dev->gdma_context->mana;
struct ib_pd *ibpd = qp->ibqp.pd;
struct mana_port_context *mpc;
struct mana_context *mc;
diff --git a/drivers/net/ethernet/microsoft/mana/gdma_main.c b/drivers/net/ethernet/microsoft/mana/gdma_main.c
index 6367de0c2c2e..e6e71e3c357c 100644
--- a/drivers/net/ethernet/microsoft/mana/gdma_main.c
+++ b/drivers/net/ethernet/microsoft/mana/gdma_main.c
@@ -158,6 +158,9 @@ static int mana_gd_detect_devices(struct pci_dev *pdev)
if (dev_type == GDMA_DEVICE_MANA) {
gc->mana.gdma_context = gc;
gc->mana.dev_id = dev;
+ } else if (dev_type == GDMA_DEVICE_MANA_IB) {
+ gc->mana_ib.dev_id = dev;
+ gc->mana_ib.gdma_context = gc;
}
}
@@ -971,6 +974,7 @@ int mana_gd_register_device(struct gdma_dev *gd)
return 0;
}
+EXPORT_SYMBOL_NS(mana_gd_register_device, NET_MANA);
int mana_gd_deregister_device(struct gdma_dev *gd)
{
@@ -1001,6 +1005,7 @@ int mana_gd_deregister_device(struct gdma_dev *gd)
return err;
}
+EXPORT_SYMBOL_NS(mana_gd_deregister_device, NET_MANA);
u32 mana_gd_wq_avail_space(struct gdma_queue *wq)
{
diff --git a/include/net/mana/gdma.h b/include/net/mana/gdma.h
index 88b6ef7ce1a6..000f0d7670f7 100644
--- a/include/net/mana/gdma.h
+++ b/include/net/mana/gdma.h
@@ -66,6 +66,7 @@ enum {
GDMA_DEVICE_NONE = 0,
GDMA_DEVICE_HWC = 1,
GDMA_DEVICE_MANA = 2,
+ GDMA_DEVICE_MANA_IB = 3,
};
struct gdma_resource {
@@ -387,6 +388,9 @@ struct gdma_context {
/* Azure network adapter */
struct gdma_dev mana;
+
+ /* Azure RDMA adapter */
+ struct gdma_dev mana_ib;
};
#define MAX_NUM_GDMA_DEVICES 4
--
2.25.1
^ permalink raw reply related [flat|nested] 7+ messages in thread* [Patch v2 2/3] RDMA/mana_ib: query device capabilities
2023-12-04 23:02 [Patch v2 0/3] Register with RDMA SOC interface and support for CQ longli
2023-12-04 23:02 ` [Patch v2 1/3] RDMA/mana_ib: register RDMA device with GDMA longli
@ 2023-12-04 23:02 ` longli
2023-12-04 23:02 ` [Patch v2 3/3] RDMA/mana_ib: Add CQ interrupt support for RAW QP longli
2 siblings, 0 replies; 7+ messages in thread
From: longli @ 2023-12-04 23:02 UTC (permalink / raw)
To: Jason Gunthorpe, Leon Romanovsky, Ajay Sharma, Dexuan Cui,
K. Y. Srinivasan, Haiyang Zhang, Wei Liu, David S. Miller,
Eric Dumazet, Jakub Kicinski, Paolo Abeni
Cc: linux-rdma, linux-hyperv, netdev, linux-kernel, Long Li
From: Long Li <longli@microsoft.com>
With RDMA device registered, use it to query on hardware capabilities and
cache this information for future query requests to the driver.
Signed-off-by: Long Li <longli@microsoft.com>
---
drivers/infiniband/hw/mana/cq.c | 2 +-
drivers/infiniband/hw/mana/device.c | 7 +++
drivers/infiniband/hw/mana/main.c | 65 ++++++++++++++++++++++------
drivers/infiniband/hw/mana/mana_ib.h | 50 +++++++++++++++++++++
drivers/infiniband/hw/mana/qp.c | 4 +-
include/net/mana/gdma.h | 1 +
6 files changed, 113 insertions(+), 16 deletions(-)
diff --git a/drivers/infiniband/hw/mana/cq.c b/drivers/infiniband/hw/mana/cq.c
index d141cab8a1e6..09a2c263e39b 100644
--- a/drivers/infiniband/hw/mana/cq.c
+++ b/drivers/infiniband/hw/mana/cq.c
@@ -26,7 +26,7 @@ int mana_ib_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
return err;
}
- if (attr->cqe > MAX_SEND_BUFFERS_PER_QUEUE) {
+ if (attr->cqe > mdev->adapter_caps.max_qp_wr) {
ibdev_dbg(ibdev, "CQE %d exceeding limit\n", attr->cqe);
return -EINVAL;
}
diff --git a/drivers/infiniband/hw/mana/device.c b/drivers/infiniband/hw/mana/device.c
index fe025e13a45c..e9494172195b 100644
--- a/drivers/infiniband/hw/mana/device.c
+++ b/drivers/infiniband/hw/mana/device.c
@@ -85,6 +85,13 @@ static int mana_ib_probe(struct auxiliary_device *adev,
}
dev->gdma_dev = &mdev->gdma_context->mana_ib;
+ ret = mana_ib_query_adapter_caps(dev);
+ if (ret) {
+ ibdev_err(&dev->ib_dev, "Failed to query device caps, ret %d",
+ ret);
+ goto free_ib_device;
+ }
+
ret = ib_register_device(&dev->ib_dev, "mana_%d",
mdev->gdma_context->dev);
if (ret)
diff --git a/drivers/infiniband/hw/mana/main.c b/drivers/infiniband/hw/mana/main.c
index 53730306ed9b..8d8f711121d2 100644
--- a/drivers/infiniband/hw/mana/main.c
+++ b/drivers/infiniband/hw/mana/main.c
@@ -486,20 +486,17 @@ int mana_ib_get_port_immutable(struct ib_device *ibdev, u32 port_num,
int mana_ib_query_device(struct ib_device *ibdev, struct ib_device_attr *props,
struct ib_udata *uhw)
{
- props->max_qp = MANA_MAX_NUM_QUEUES;
- props->max_qp_wr = MAX_SEND_BUFFERS_PER_QUEUE;
-
- /*
- * max_cqe could be potentially much bigger.
- * As this version of driver only support RAW QP, set it to the same
- * value as max_qp_wr
- */
- props->max_cqe = MAX_SEND_BUFFERS_PER_QUEUE;
-
+ struct mana_ib_dev *dev = container_of(ibdev,
+ struct mana_ib_dev, ib_dev);
+
+ props->max_qp = dev->adapter_caps.max_qp_count;
+ props->max_qp_wr = dev->adapter_caps.max_qp_wr;
+ props->max_cq = dev->adapter_caps.max_cq_count;
+ props->max_cqe = dev->adapter_caps.max_qp_wr;
+ props->max_mr = dev->adapter_caps.max_mr_count;
props->max_mr_size = MANA_IB_MAX_MR_SIZE;
- props->max_mr = MANA_IB_MAX_MR;
- props->max_send_sge = MAX_TX_WQE_SGL_ENTRIES;
- props->max_recv_sge = MAX_RX_WQE_SGL_ENTRIES;
+ props->max_send_sge = dev->adapter_caps.max_send_sge_count;
+ props->max_recv_sge = dev->adapter_caps.max_recv_sge_count;
return 0;
}
@@ -521,3 +518,45 @@ int mana_ib_query_gid(struct ib_device *ibdev, u32 port, int index,
void mana_ib_disassociate_ucontext(struct ib_ucontext *ibcontext)
{
}
+
+int mana_ib_query_adapter_caps(struct mana_ib_dev *dev)
+{
+ struct mana_ib_adapter_caps *caps = &dev->adapter_caps;
+ struct mana_ib_query_adapter_caps_resp resp = {};
+ struct mana_ib_query_adapter_caps_req req = {};
+ int err;
+
+ mana_gd_init_req_hdr(&req.hdr, MANA_IB_GET_ADAPTER_CAP, sizeof(req),
+ sizeof(resp));
+ req.hdr.resp.msg_version = GDMA_MESSAGE_V3;
+ req.hdr.dev_id = dev->gdma_dev->dev_id;
+
+ err = mana_gd_send_request(dev->gdma_dev->gdma_context, sizeof(req),
+ &req, sizeof(resp), &resp);
+
+ if (err) {
+ ibdev_err(&dev->ib_dev,
+ "Failed to query adapter caps err %d", err);
+ return err;
+ }
+
+ caps->max_sq_id = resp.max_sq_id;
+ caps->max_rq_id = resp.max_rq_id;
+ caps->max_cq_id = resp.max_cq_id;
+ caps->max_qp_count = resp.max_qp_count;
+ caps->max_cq_count = resp.max_cq_count;
+ caps->max_mr_count = resp.max_mr_count;
+ caps->max_pd_count = resp.max_pd_count;
+ caps->max_inbound_read_limit = resp.max_inbound_read_limit;
+ caps->max_outbound_read_limit = resp.max_outbound_read_limit;
+ caps->mw_count = resp.mw_count;
+ caps->max_srq_count = resp.max_srq_count;
+ caps->max_qp_wr = min_t(u32,
+ resp.max_requester_sq_size / GDMA_MAX_SQE_SIZE,
+ resp.max_requester_rq_size / GDMA_MAX_RQE_SIZE);
+ caps->max_inline_data_size = resp.max_inline_data_size;
+ caps->max_send_sge_count = resp.max_send_sge_count;
+ caps->max_recv_sge_count = resp.max_recv_sge_count;
+
+ return 0;
+}
diff --git a/drivers/infiniband/hw/mana/mana_ib.h b/drivers/infiniband/hw/mana/mana_ib.h
index 502cc8672eef..7cb3d8ee4292 100644
--- a/drivers/infiniband/hw/mana/mana_ib.h
+++ b/drivers/infiniband/hw/mana/mana_ib.h
@@ -27,9 +27,28 @@
*/
#define MANA_IB_MAX_MR 0xFFFFFFu
+struct mana_ib_adapter_caps {
+ u32 max_sq_id;
+ u32 max_rq_id;
+ u32 max_cq_id;
+ u32 max_qp_count;
+ u32 max_cq_count;
+ u32 max_mr_count;
+ u32 max_pd_count;
+ u32 max_inbound_read_limit;
+ u32 max_outbound_read_limit;
+ u32 mw_count;
+ u32 max_srq_count;
+ u32 max_qp_wr;
+ u32 max_send_sge_count;
+ u32 max_recv_sge_count;
+ u32 max_inline_data_size;
+};
+
struct mana_ib_dev {
struct ib_device ib_dev;
struct gdma_dev *gdma_dev;
+ struct mana_ib_adapter_caps adapter_caps;
};
struct mana_ib_wq {
@@ -92,6 +111,36 @@ struct mana_ib_rwq_ind_table {
struct ib_rwq_ind_table ib_ind_table;
};
+enum mana_ib_command_code {
+ MANA_IB_GET_ADAPTER_CAP = 0x30001,
+};
+
+struct mana_ib_query_adapter_caps_req {
+ struct gdma_req_hdr hdr;
+}; /*HW Data */
+
+struct mana_ib_query_adapter_caps_resp {
+ struct gdma_resp_hdr hdr;
+ u32 max_sq_id;
+ u32 max_rq_id;
+ u32 max_cq_id;
+ u32 max_qp_count;
+ u32 max_cq_count;
+ u32 max_mr_count;
+ u32 max_pd_count;
+ u32 max_inbound_read_limit;
+ u32 max_outbound_read_limit;
+ u32 mw_count;
+ u32 max_srq_count;
+ u32 max_requester_sq_size;
+ u32 max_responder_sq_size;
+ u32 max_requester_rq_size;
+ u32 max_responder_rq_size;
+ u32 max_send_sge_count;
+ u32 max_recv_sge_count;
+ u32 max_inline_data_size;
+}; /* HW Data */
+
int mana_ib_gd_create_dma_region(struct mana_ib_dev *dev, struct ib_umem *umem,
mana_handle_t *gdma_region);
@@ -159,4 +208,5 @@ int mana_ib_query_gid(struct ib_device *ibdev, u32 port, int index,
void mana_ib_disassociate_ucontext(struct ib_ucontext *ibcontext);
+int mana_ib_query_adapter_caps(struct mana_ib_dev *mdev);
#endif
diff --git a/drivers/infiniband/hw/mana/qp.c b/drivers/infiniband/hw/mana/qp.c
index ae45d28eef5e..4667b18ec1dd 100644
--- a/drivers/infiniband/hw/mana/qp.c
+++ b/drivers/infiniband/hw/mana/qp.c
@@ -130,7 +130,7 @@ static int mana_ib_create_qp_rss(struct ib_qp *ibqp, struct ib_pd *pd,
return ret;
}
- if (attr->cap.max_recv_wr > MAX_SEND_BUFFERS_PER_QUEUE) {
+ if (attr->cap.max_recv_wr > mdev->adapter_caps.max_qp_wr) {
ibdev_dbg(&mdev->ib_dev,
"Requested max_recv_wr %d exceeding limit\n",
attr->cap.max_recv_wr);
@@ -296,7 +296,7 @@ static int mana_ib_create_qp_raw(struct ib_qp *ibqp, struct ib_pd *ibpd,
if (port < 1 || port > mc->num_ports)
return -EINVAL;
- if (attr->cap.max_send_wr > MAX_SEND_BUFFERS_PER_QUEUE) {
+ if (attr->cap.max_send_wr > mdev->adapter_caps.max_qp_wr) {
ibdev_dbg(&mdev->ib_dev,
"Requested max_send_wr %d exceeding limit\n",
attr->cap.max_send_wr);
diff --git a/include/net/mana/gdma.h b/include/net/mana/gdma.h
index 000f0d7670f7..797971e2d5a5 100644
--- a/include/net/mana/gdma.h
+++ b/include/net/mana/gdma.h
@@ -150,6 +150,7 @@ struct gdma_general_req {
#define GDMA_MESSAGE_V1 1
#define GDMA_MESSAGE_V2 2
+#define GDMA_MESSAGE_V3 3
struct gdma_general_resp {
struct gdma_resp_hdr hdr;
--
2.25.1
^ permalink raw reply related [flat|nested] 7+ messages in thread* [Patch v2 3/3] RDMA/mana_ib: Add CQ interrupt support for RAW QP
2023-12-04 23:02 [Patch v2 0/3] Register with RDMA SOC interface and support for CQ longli
2023-12-04 23:02 ` [Patch v2 1/3] RDMA/mana_ib: register RDMA device with GDMA longli
2023-12-04 23:02 ` [Patch v2 2/3] RDMA/mana_ib: query device capabilities longli
@ 2023-12-04 23:02 ` longli
2023-12-09 17:33 ` Simon Horman
2 siblings, 1 reply; 7+ messages in thread
From: longli @ 2023-12-04 23:02 UTC (permalink / raw)
To: Jason Gunthorpe, Leon Romanovsky, Ajay Sharma, Dexuan Cui,
K. Y. Srinivasan, Haiyang Zhang, Wei Liu, David S. Miller,
Eric Dumazet, Jakub Kicinski, Paolo Abeni
Cc: linux-rdma, linux-hyperv, netdev, linux-kernel, Long Li
From: Long Li <longli@microsoft.com>
At probing time, the MANA core code allocates EQs for supporting interrupts
on Ethernet queues. The same interrupt mechanisum is used by RAW QP.
Use the same EQs for delivering interrupts on the CQ for the RAW QP.
Signed-off-by: Long Li <longli@microsoft.com>
---
drivers/infiniband/hw/mana/cq.c | 32 ++++++++++++-
drivers/infiniband/hw/mana/mana_ib.h | 3 ++
drivers/infiniband/hw/mana/qp.c | 72 ++++++++++++++++++++++++++--
3 files changed, 102 insertions(+), 5 deletions(-)
diff --git a/drivers/infiniband/hw/mana/cq.c b/drivers/infiniband/hw/mana/cq.c
index 09a2c263e39b..83ebd070535a 100644
--- a/drivers/infiniband/hw/mana/cq.c
+++ b/drivers/infiniband/hw/mana/cq.c
@@ -12,13 +12,20 @@ int mana_ib_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
struct ib_device *ibdev = ibcq->device;
struct mana_ib_create_cq ucmd = {};
struct mana_ib_dev *mdev;
+ struct gdma_context *gc;
int err;
mdev = container_of(ibdev, struct mana_ib_dev, ib_dev);
+ gc = mdev->gdma_dev->gdma_context;
if (udata->inlen < sizeof(ucmd))
return -EINVAL;
+ if (attr->comp_vector > gc->max_num_queues)
+ return -EINVAL;
+
+ cq->comp_vector = attr->comp_vector;
+
err = ib_copy_from_udata(&ucmd, udata, min(sizeof(ucmd), udata->inlen));
if (err) {
ibdev_dbg(ibdev,
@@ -56,6 +63,7 @@ int mana_ib_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
/*
* The CQ ID is not known at this time. The ID is generated at create_qp
*/
+ cq->id = INVALID_QUEUE_ID;
return 0;
@@ -69,11 +77,33 @@ int mana_ib_destroy_cq(struct ib_cq *ibcq, struct ib_udata *udata)
struct mana_ib_cq *cq = container_of(ibcq, struct mana_ib_cq, ibcq);
struct ib_device *ibdev = ibcq->device;
struct mana_ib_dev *mdev;
+ struct gdma_context *gc;
+ int err;
mdev = container_of(ibdev, struct mana_ib_dev, ib_dev);
+ gc = mdev->gdma_dev->gdma_context;
+
+ err = mana_ib_gd_destroy_dma_region(mdev, cq->gdma_region);
+ if (err) {
+ ibdev_dbg(ibdev,
+ "Failed to destroy dma region, %d\n", err);
+ return err;
+ }
+
+ if (cq->id != INVALID_QUEUE_ID) {
+ kfree(gc->cq_table[cq->id]);
+ gc->cq_table[cq->id] = NULL;
+ }
- mana_ib_gd_destroy_dma_region(mdev, cq->gdma_region);
ib_umem_release(cq->umem);
return 0;
}
+
+void mana_ib_cq_handler(void *ctx, struct gdma_queue *gdma_cq)
+{
+ struct mana_ib_cq *cq = ctx;
+
+ if (cq->ibcq.comp_handler)
+ cq->ibcq.comp_handler(&cq->ibcq, cq->ibcq.cq_context);
+}
diff --git a/drivers/infiniband/hw/mana/mana_ib.h b/drivers/infiniband/hw/mana/mana_ib.h
index 7cb3d8ee4292..53bb4905afd5 100644
--- a/drivers/infiniband/hw/mana/mana_ib.h
+++ b/drivers/infiniband/hw/mana/mana_ib.h
@@ -86,6 +86,7 @@ struct mana_ib_cq {
int cqe;
u64 gdma_region;
u64 id;
+ u32 comp_vector;
};
struct mana_ib_qp {
@@ -209,4 +210,6 @@ int mana_ib_query_gid(struct ib_device *ibdev, u32 port, int index,
void mana_ib_disassociate_ucontext(struct ib_ucontext *ibcontext);
int mana_ib_query_adapter_caps(struct mana_ib_dev *mdev);
+
+void mana_ib_cq_handler(void *ctx, struct gdma_queue *gdma_cq);
#endif
diff --git a/drivers/infiniband/hw/mana/qp.c b/drivers/infiniband/hw/mana/qp.c
index 4667b18ec1dd..186d9829bb93 100644
--- a/drivers/infiniband/hw/mana/qp.c
+++ b/drivers/infiniband/hw/mana/qp.c
@@ -99,25 +99,34 @@ static int mana_ib_create_qp_rss(struct ib_qp *ibqp, struct ib_pd *pd,
struct mana_ib_qp *qp = container_of(ibqp, struct mana_ib_qp, ibqp);
struct mana_ib_dev *mdev =
container_of(pd->device, struct mana_ib_dev, ib_dev);
+ struct ib_ucontext *ib_ucontext = pd->uobject->context;
struct ib_rwq_ind_table *ind_tbl = attr->rwq_ind_tbl;
struct mana_ib_create_qp_rss_resp resp = {};
struct mana_ib_create_qp_rss ucmd = {};
+ struct mana_ib_ucontext *mana_ucontext;
+ struct gdma_queue **gdma_cq_allocated;
mana_handle_t *mana_ind_table;
struct mana_port_context *mpc;
+ struct gdma_queue *gdma_cq;
unsigned int ind_tbl_size;
struct mana_context *mc;
struct net_device *ndev;
+ struct gdma_context *gc;
struct mana_ib_cq *cq;
struct mana_ib_wq *wq;
struct gdma_dev *gd;
+ struct mana_eq *eq;
struct ib_cq *ibcq;
struct ib_wq *ibwq;
int i = 0;
u32 port;
int ret;
- gd = &mdev->gdma_dev->gdma_context->mana;
+ gc = mdev->gdma_dev->gdma_context;
+ gd = &gc->mana;
mc = gd->driver_data;
+ mana_ucontext =
+ container_of(ib_ucontext, struct mana_ib_ucontext, ibucontext);
if (!udata || udata->inlen < sizeof(ucmd))
return -EINVAL;
@@ -179,6 +188,13 @@ static int mana_ib_create_qp_rss(struct ib_qp *ibqp, struct ib_pd *pd,
goto fail;
}
+ gdma_cq_allocated = kcalloc(ind_tbl_size, sizeof(*gdma_cq_allocated),
+ GFP_KERNEL);
+ if (!gdma_cq_allocated) {
+ ret = -ENOMEM;
+ goto fail;
+ }
+
qp->port = port;
for (i = 0; i < ind_tbl_size; i++) {
@@ -197,7 +213,8 @@ static int mana_ib_create_qp_rss(struct ib_qp *ibqp, struct ib_pd *pd,
cq_spec.gdma_region = cq->gdma_region;
cq_spec.queue_size = cq->cqe * COMP_ENTRY_SIZE;
cq_spec.modr_ctx_id = 0;
- cq_spec.attached_eq = GDMA_CQ_NO_EQ;
+ eq = &mc->eqs[cq->comp_vector % gc->max_num_queues];
+ cq_spec.attached_eq = eq->eq->id;
ret = mana_create_wq_obj(mpc, mpc->port_handle, GDMA_RQ,
&wq_spec, &cq_spec, &wq->rx_object);
@@ -219,6 +236,21 @@ static int mana_ib_create_qp_rss(struct ib_qp *ibqp, struct ib_pd *pd,
resp.entries[i].wqid = wq->id;
mana_ind_table[i] = wq->rx_object;
+
+ /* Create CQ table entry */
+ WARN_ON(gc->cq_table[cq->id]);
+ gdma_cq = kzalloc(sizeof(*gdma_cq), GFP_KERNEL);
+ if (!gdma_cq) {
+ ret = -ENOMEM;
+ goto fail;
+ }
+ gdma_cq_allocated[i] = gdma_cq;
+
+ gdma_cq->cq.context = cq;
+ gdma_cq->type = GDMA_CQ;
+ gdma_cq->cq.callback = mana_ib_cq_handler;
+ gdma_cq->id = cq->id;
+ gc->cq_table[cq->id] = gdma_cq;
}
resp.num_entries = i;
@@ -238,6 +270,7 @@ static int mana_ib_create_qp_rss(struct ib_qp *ibqp, struct ib_pd *pd,
goto fail;
}
+ kfree(gdma_cq_allocated);
kfree(mana_ind_table);
return 0;
@@ -247,8 +280,15 @@ static int mana_ib_create_qp_rss(struct ib_qp *ibqp, struct ib_pd *pd,
ibwq = ind_tbl->ind_tbl[i];
wq = container_of(ibwq, struct mana_ib_wq, ibwq);
mana_destroy_wq_obj(mpc, GDMA_RQ, wq->rx_object);
+
+ if (gdma_cq_allocated[i]) {
+ gc->cq_table[gdma_cq_allocated[i]->id] =
+ NULL;
+ kfree(gdma_cq_allocated[i]);
+ }
}
+ kfree(gdma_cq_allocated);
kfree(mana_ind_table);
return ret;
@@ -273,11 +313,14 @@ static int mana_ib_create_qp_raw(struct ib_qp *ibqp, struct ib_pd *ibpd,
struct mana_obj_spec wq_spec = {};
struct mana_obj_spec cq_spec = {};
struct mana_port_context *mpc;
+ struct gdma_queue *gdma_cq;
struct mana_context *mc;
struct net_device *ndev;
struct ib_umem *umem;
- int err;
+ struct mana_eq *eq;
+ int eq_vec;
u32 port;
+ int err;
mc = gd->driver_data;
@@ -354,7 +397,9 @@ static int mana_ib_create_qp_raw(struct ib_qp *ibqp, struct ib_pd *ibpd,
cq_spec.gdma_region = send_cq->gdma_region;
cq_spec.queue_size = send_cq->cqe * COMP_ENTRY_SIZE;
cq_spec.modr_ctx_id = 0;
- cq_spec.attached_eq = GDMA_CQ_NO_EQ;
+ eq_vec = send_cq->comp_vector % gd->gdma_context->max_num_queues;
+ eq = &mc->eqs[eq_vec];
+ cq_spec.attached_eq = eq->eq->id;
err = mana_create_wq_obj(mpc, mpc->port_handle, GDMA_SQ, &wq_spec,
&cq_spec, &qp->tx_object);
@@ -372,6 +417,20 @@ static int mana_ib_create_qp_raw(struct ib_qp *ibqp, struct ib_pd *ibpd,
qp->sq_id = wq_spec.queue_index;
send_cq->id = cq_spec.queue_index;
+ /* Create CQ table entry */
+ WARN_ON(gd->gdma_context->cq_table[send_cq->id]);
+ gdma_cq = kzalloc(sizeof(*gdma_cq), GFP_KERNEL);
+ if (!gdma_cq) {
+ err = -ENOMEM;
+ goto err_destroy_wq_obj;
+ }
+
+ gdma_cq->cq.context = send_cq;
+ gdma_cq->type = GDMA_CQ;
+ gdma_cq->cq.callback = mana_ib_cq_handler;
+ gdma_cq->id = send_cq->id;
+ gd->gdma_context->cq_table[send_cq->id] = gdma_cq;
+
ibdev_dbg(&mdev->ib_dev,
"ret %d qp->tx_object 0x%llx sq id %llu cq id %llu\n", err,
qp->tx_object, qp->sq_id, send_cq->id);
@@ -391,6 +450,11 @@ static int mana_ib_create_qp_raw(struct ib_qp *ibqp, struct ib_pd *ibpd,
return 0;
err_destroy_wq_obj:
+ if (gdma_cq) {
+ kfree(gdma_cq);
+ gd->gdma_context->cq_table[send_cq->id] = NULL;
+ }
+
mana_destroy_wq_obj(mpc, GDMA_SQ, qp->tx_object);
err_destroy_dma_region:
--
2.25.1
^ permalink raw reply related [flat|nested] 7+ messages in thread* Re: [Patch v2 3/3] RDMA/mana_ib: Add CQ interrupt support for RAW QP
2023-12-04 23:02 ` [Patch v2 3/3] RDMA/mana_ib: Add CQ interrupt support for RAW QP longli
@ 2023-12-09 17:33 ` Simon Horman
2023-12-12 2:50 ` Long Li
0 siblings, 1 reply; 7+ messages in thread
From: Simon Horman @ 2023-12-09 17:33 UTC (permalink / raw)
To: longli
Cc: Jason Gunthorpe, Leon Romanovsky, Ajay Sharma, Dexuan Cui,
K. Y. Srinivasan, Haiyang Zhang, Wei Liu, David S. Miller,
Eric Dumazet, Jakub Kicinski, Paolo Abeni, linux-rdma,
linux-hyperv, netdev, linux-kernel, Long Li
On Mon, Dec 04, 2023 at 03:02:59PM -0800, longli@linuxonhyperv.com wrote:
> From: Long Li <longli@microsoft.com>
>
> At probing time, the MANA core code allocates EQs for supporting interrupts
> on Ethernet queues. The same interrupt mechanisum is used by RAW QP.
>
> Use the same EQs for delivering interrupts on the CQ for the RAW QP.
>
> Signed-off-by: Long Li <longli@microsoft.com>
Hi Long Li,
some minor feedback from my side.
...
> diff --git a/drivers/infiniband/hw/mana/qp.c b/drivers/infiniband/hw/mana/qp.c
> index 4667b18ec1dd..186d9829bb93 100644
> --- a/drivers/infiniband/hw/mana/qp.c
> +++ b/drivers/infiniband/hw/mana/qp.c
> @@ -99,25 +99,34 @@ static int mana_ib_create_qp_rss(struct ib_qp *ibqp, struct ib_pd *pd,
> struct mana_ib_qp *qp = container_of(ibqp, struct mana_ib_qp, ibqp);
> struct mana_ib_dev *mdev =
> container_of(pd->device, struct mana_ib_dev, ib_dev);
> + struct ib_ucontext *ib_ucontext = pd->uobject->context;
> struct ib_rwq_ind_table *ind_tbl = attr->rwq_ind_tbl;
> struct mana_ib_create_qp_rss_resp resp = {};
> struct mana_ib_create_qp_rss ucmd = {};
> + struct mana_ib_ucontext *mana_ucontext;
> + struct gdma_queue **gdma_cq_allocated;
> mana_handle_t *mana_ind_table;
> struct mana_port_context *mpc;
> + struct gdma_queue *gdma_cq;
> unsigned int ind_tbl_size;
> struct mana_context *mc;
> struct net_device *ndev;
> + struct gdma_context *gc;
> struct mana_ib_cq *cq;
> struct mana_ib_wq *wq;
> struct gdma_dev *gd;
> + struct mana_eq *eq;
> struct ib_cq *ibcq;
> struct ib_wq *ibwq;
> int i = 0;
> u32 port;
> int ret;
>
> - gd = &mdev->gdma_dev->gdma_context->mana;
> + gc = mdev->gdma_dev->gdma_context;
> + gd = &gc->mana;
> mc = gd->driver_data;
> + mana_ucontext =
> + container_of(ib_ucontext, struct mana_ib_ucontext, ibucontext);
>
> if (!udata || udata->inlen < sizeof(ucmd))
> return -EINVAL;
nit: mana_ucontext appears to be set but unused.
Flagged by W=1 builds.
> @@ -179,6 +188,13 @@ static int mana_ib_create_qp_rss(struct ib_qp *ibqp, struct ib_pd *pd,
> goto fail;
> }
>
> + gdma_cq_allocated = kcalloc(ind_tbl_size, sizeof(*gdma_cq_allocated),
> + GFP_KERNEL);
> + if (!gdma_cq_allocated) {
> + ret = -ENOMEM;
> + goto fail;
> + }
> +
> qp->port = port;
>
> for (i = 0; i < ind_tbl_size; i++) {
...
> @@ -219,6 +236,21 @@ static int mana_ib_create_qp_rss(struct ib_qp *ibqp, struct ib_pd *pd,
> resp.entries[i].wqid = wq->id;
>
> mana_ind_table[i] = wq->rx_object;
> +
> + /* Create CQ table entry */
> + WARN_ON(gc->cq_table[cq->id]);
> + gdma_cq = kzalloc(sizeof(*gdma_cq), GFP_KERNEL);
> + if (!gdma_cq) {
> + ret = -ENOMEM;
> + goto fail;
> + }
> + gdma_cq_allocated[i] = gdma_cq;
> +
> + gdma_cq->cq.context = cq;
> + gdma_cq->type = GDMA_CQ;
> + gdma_cq->cq.callback = mana_ib_cq_handler;
> + gdma_cq->id = cq->id;
> + gc->cq_table[cq->id] = gdma_cq;
> }
> resp.num_entries = i;
>
> @@ -238,6 +270,7 @@ static int mana_ib_create_qp_rss(struct ib_qp *ibqp, struct ib_pd *pd,
> goto fail;
> }
>
> + kfree(gdma_cq_allocated);
> kfree(mana_ind_table);
>
> return 0;
> @@ -247,8 +280,15 @@ static int mana_ib_create_qp_rss(struct ib_qp *ibqp, struct ib_pd *pd,
> ibwq = ind_tbl->ind_tbl[i];
> wq = container_of(ibwq, struct mana_ib_wq, ibwq);
> mana_destroy_wq_obj(mpc, GDMA_RQ, wq->rx_object);
> +
> + if (gdma_cq_allocated[i]) {
nit: It is not clear to me that condition can ever be false.
If we get here then gdma_cq_allocated[i] is a valid pointer.
> + gc->cq_table[gdma_cq_allocated[i]->id] =
> + NULL;
> + kfree(gdma_cq_allocated[i]);
> + }
> }
>
> + kfree(gdma_cq_allocated);
> kfree(mana_ind_table);
>
> return ret;
...
^ permalink raw reply [flat|nested] 7+ messages in thread* RE: [Patch v2 3/3] RDMA/mana_ib: Add CQ interrupt support for RAW QP
2023-12-09 17:33 ` Simon Horman
@ 2023-12-12 2:50 ` Long Li
0 siblings, 0 replies; 7+ messages in thread
From: Long Li @ 2023-12-12 2:50 UTC (permalink / raw)
To: Simon Horman, longli@linuxonhyperv.com
Cc: Jason Gunthorpe, Leon Romanovsky, Ajay Sharma, Dexuan Cui,
KY Srinivasan, Haiyang Zhang, Wei Liu, David S. Miller,
Eric Dumazet, Jakub Kicinski, Paolo Abeni,
linux-rdma@vger.kernel.org, linux-hyperv@vger.kernel.org,
netdev@vger.kernel.org, linux-kernel@vger.kernel.org
> Hi Long Li,
>
> some minor feedback from my side.
>
> ...
>
> > diff --git a/drivers/infiniband/hw/mana/qp.c
> > b/drivers/infiniband/hw/mana/qp.c index 4667b18ec1dd..186d9829bb93
> > 100644
> > --- a/drivers/infiniband/hw/mana/qp.c
> > +++ b/drivers/infiniband/hw/mana/qp.c
> > @@ -99,25 +99,34 @@ static int mana_ib_create_qp_rss(struct ib_qp *ibqp,
> struct ib_pd *pd,
> > struct mana_ib_qp *qp = container_of(ibqp, struct mana_ib_qp, ibqp);
> > struct mana_ib_dev *mdev =
> > container_of(pd->device, struct mana_ib_dev, ib_dev);
> > + struct ib_ucontext *ib_ucontext = pd->uobject->context;
> > struct ib_rwq_ind_table *ind_tbl = attr->rwq_ind_tbl;
> > struct mana_ib_create_qp_rss_resp resp = {};
> > struct mana_ib_create_qp_rss ucmd = {};
> > + struct mana_ib_ucontext *mana_ucontext;
> > + struct gdma_queue **gdma_cq_allocated;
> > mana_handle_t *mana_ind_table;
> > struct mana_port_context *mpc;
> > + struct gdma_queue *gdma_cq;
> > unsigned int ind_tbl_size;
> > struct mana_context *mc;
> > struct net_device *ndev;
> > + struct gdma_context *gc;
> > struct mana_ib_cq *cq;
> > struct mana_ib_wq *wq;
> > struct gdma_dev *gd;
> > + struct mana_eq *eq;
> > struct ib_cq *ibcq;
> > struct ib_wq *ibwq;
> > int i = 0;
> > u32 port;
> > int ret;
> >
> > - gd = &mdev->gdma_dev->gdma_context->mana;
> > + gc = mdev->gdma_dev->gdma_context;
> > + gd = &gc->mana;
> > mc = gd->driver_data;
> > + mana_ucontext =
> > + container_of(ib_ucontext, struct mana_ib_ucontext, ibucontext);
> >
> > if (!udata || udata->inlen < sizeof(ucmd))
> > return -EINVAL;
>
> nit: mana_ucontext appears to be set but unused.
Thank you, will fix this in v3.
>
> Flagged by W=1 builds.
>
> > @@ -179,6 +188,13 @@ static int mana_ib_create_qp_rss(struct ib_qp *ibqp,
> struct ib_pd *pd,
> > goto fail;
> > }
> >
> > + gdma_cq_allocated = kcalloc(ind_tbl_size, sizeof(*gdma_cq_allocated),
> > + GFP_KERNEL);
> > + if (!gdma_cq_allocated) {
> > + ret = -ENOMEM;
> > + goto fail;
> > + }
> > +
> > qp->port = port;
> >
> > for (i = 0; i < ind_tbl_size; i++) {
>
> ...
>
> > @@ -219,6 +236,21 @@ static int mana_ib_create_qp_rss(struct ib_qp *ibqp,
> struct ib_pd *pd,
> > resp.entries[i].wqid = wq->id;
> >
> > mana_ind_table[i] = wq->rx_object;
> > +
> > + /* Create CQ table entry */
> > + WARN_ON(gc->cq_table[cq->id]);
> > + gdma_cq = kzalloc(sizeof(*gdma_cq), GFP_KERNEL);
> > + if (!gdma_cq) {
> > + ret = -ENOMEM;
> > + goto fail;
> > + }
> > + gdma_cq_allocated[i] = gdma_cq;
> > +
> > + gdma_cq->cq.context = cq;
> > + gdma_cq->type = GDMA_CQ;
> > + gdma_cq->cq.callback = mana_ib_cq_handler;
> > + gdma_cq->id = cq->id;
> > + gc->cq_table[cq->id] = gdma_cq;
> > }
> > resp.num_entries = i;
> >
> > @@ -238,6 +270,7 @@ static int mana_ib_create_qp_rss(struct ib_qp *ibqp,
> struct ib_pd *pd,
> > goto fail;
> > }
> >
> > + kfree(gdma_cq_allocated);
> > kfree(mana_ind_table);
> >
> > return 0;
> > @@ -247,8 +280,15 @@ static int mana_ib_create_qp_rss(struct ib_qp *ibqp,
> struct ib_pd *pd,
> > ibwq = ind_tbl->ind_tbl[i];
> > wq = container_of(ibwq, struct mana_ib_wq, ibwq);
> > mana_destroy_wq_obj(mpc, GDMA_RQ, wq->rx_object);
> > +
> > + if (gdma_cq_allocated[i]) {
>
> nit: It is not clear to me that condition can ever be false.
> If we get here then gdma_cq_allocated[i] is a valid pointer.
I'm removing this check.
I spotted another bug here: mana_create_wq_obj() could fail and break the loop early. I will do a proper cleanup and send v3.
Thanks,
Long
>
> > + gc->cq_table[gdma_cq_allocated[i]->id] =
> > + NULL;
> > + kfree(gdma_cq_allocated[i]);
> > + }
> > }
> >
> > + kfree(gdma_cq_allocated);
> > kfree(mana_ind_table);
> >
> > return ret;
>
> ...
^ permalink raw reply [flat|nested] 7+ messages in thread