[Patch v1 0/4]

[Patch v1 0/4]

[longli]

From: Long Li <longli@microsoft.com>

This patchset add support for registering a RDMA device handle with SoC
for support of upcoming RC queue pairs and CQ interrupts.

This patchset is partially based on Ajay Sharma's work:
https://lore.kernel.org/netdev/1697494322-26814-1-git-send-email-sharmaajay@linuxonhyperv.com

Long Li (4):
  RDMA/mana_ib: register RDMA device with GDMA
  RDMA/mana_ib: create and process EQ events
  RDMA/mana_ib: create RDMA adapter handle
  RDMA/mana_ib: query device capabilities

 drivers/infiniband/hw/mana/cq.c               |   2 +-
 drivers/infiniband/hw/mana/device.c           |  52 +++++-
 drivers/infiniband/hw/mana/main.c             | 166 ++++++++++++++++--
 drivers/infiniband/hw/mana/mana_ib.h          |  85 +++++++++
 drivers/infiniband/hw/mana/qp.c               |  36 +++-
 .../net/ethernet/microsoft/mana/gdma_main.c   | 152 +++++++++-------
 drivers/net/ethernet/microsoft/mana/mana_en.c |   3 +
 include/net/mana/gdma.h                       |  19 +-
 8 files changed, 420 insertions(+), 95 deletions(-)

-- 
2.34.1

[Patch v1 1/4] RDMA/mana_ib: register RDMA device with GDMA

[longli]

From: Long Li <longli@microsoft.com>

MANA hardware supports RC queue pairs over RoCE. Software needs to register
with the RDMA management interface on the SoC to access this feature.

Signed-off-by: Long Li <longli@microsoft.com>
---
 drivers/infiniband/hw/mana/device.c           | 25 ++++++++++++++++---
 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, 41 insertions(+), 12 deletions(-)

diff --git a/drivers/infiniband/hw/mana/device.c b/drivers/infiniband/hw/mana/device.c
index d4541b8707e4..ee29ddf36cf3 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 +77,33 @@ 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) {
+		ibdev_err(&dev->ib_dev, "Failed to register device, ret %d",
+			  ret);
+		goto free_ib_device;
+	}
+	dev->gdma_dev = &mdev->gdma_context->mana_ib;
+
 	if (ret) {
 		ib_dealloc_device(&dev->ib_dev);
 		return ret;
 	}
 
+	ret = ib_register_device(&dev->ib_dev, "mana_%d",
+				 mdev->gdma_context->dev);
+	if (ret)
+		goto destroy_adapter;
+
 	dev_set_drvdata(&adev->dev, dev);
 
 	return 0;
+
+destroy_adapter:
+	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 +111,9 @@ 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..02e50ed632ee 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(mana_gd_register_device);
 
 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(mana_gd_deregister_device);
 
 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
-- 
2.34.1

[Patch v1 2/4] RDMA/mana_ib: create and process EQ events

[longli]

From: Long Li <longli@microsoft.com>

Before the software can create an RDMA adapter handle with SoC, it needs to
create EQs for processing SoC events from RDMA device. Because MSI-X
vectors are shared between MANA Ethernet device and RDMA device, this
patch adds support to share EQs on MSI-X vectors and creates management
EQ for RDMA device.

Signed-off-by: Long Li <longli@microsoft.com>
---
 drivers/infiniband/hw/mana/device.c           |  13 ++
 drivers/infiniband/hw/mana/main.c             |  52 +++++++
 drivers/infiniband/hw/mana/mana_ib.h          |   4 +
 drivers/infiniband/hw/mana/qp.c               |  15 ++
 .../net/ethernet/microsoft/mana/gdma_main.c   | 147 ++++++++++--------
 drivers/net/ethernet/microsoft/mana/mana_en.c |   3 +
 include/net/mana/gdma.h                       |  14 +-
 7 files changed, 180 insertions(+), 68 deletions(-)

diff --git a/drivers/infiniband/hw/mana/device.c b/drivers/infiniband/hw/mana/device.c
index ee29ddf36cf3..3da4763e1a0c 100644
--- a/drivers/infiniband/hw/mana/device.c
+++ b/drivers/infiniband/hw/mana/device.c
@@ -85,6 +85,14 @@ static int mana_ib_probe(struct auxiliary_device *adev,
 	}
 	dev->gdma_dev = &mdev->gdma_context->mana_ib;
 
+	xa_init(&dev->rq_to_qp_lookup_table);
+
+	ret = mana_ib_create_error_eq(dev);
+	if (ret) {
+		ibdev_err(&dev->ib_dev, "Failed to allocate err eq");
+		goto deregister_device;
+	}
+
 	if (ret) {
 		ib_dealloc_device(&dev->ib_dev);
 		return ret;
@@ -100,6 +108,9 @@ static int mana_ib_probe(struct auxiliary_device *adev,
 	return 0;
 
 destroy_adapter:
+	mana_gd_destroy_queue(dev->gdma_dev->gdma_context, dev->fatal_err_eq);
+	xa_destroy(&dev->rq_to_qp_lookup_table);
+deregister_device:
 	mana_gd_deregister_device(dev->gdma_dev);
 free_ib_device:
 	ib_dealloc_device(&dev->ib_dev);
@@ -112,6 +123,8 @@ static void mana_ib_remove(struct auxiliary_device *adev)
 
 	ib_unregister_device(&dev->ib_dev);
 
+	mana_gd_destroy_queue(dev->gdma_dev->gdma_context, dev->fatal_err_eq);
+	xa_destroy(&dev->rq_to_qp_lookup_table);
 	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 53730306ed9b..032f926bf1ab 100644
--- a/drivers/infiniband/hw/mana/main.c
+++ b/drivers/infiniband/hw/mana/main.c
@@ -521,3 +521,55 @@ int mana_ib_query_gid(struct ib_device *ibdev, u32 port, int index,
 void mana_ib_disassociate_ucontext(struct ib_ucontext *ibcontext)
 {
 }
+
+static void mana_ib_critical_event_handler(void *ctx, struct gdma_queue *queue,
+				      struct gdma_event *event)
+{
+	struct mana_ib_dev *dev = (struct mana_ib_dev *)ctx;
+	struct ib_event mib_event;
+	struct mana_ib_qp *qp;
+	u64 rq_id;
+
+	switch (event->type) {
+	case GDMA_EQE_SOC_EVENT_NOTIFICATION:
+		rq_id = event->details[0] & 0xFFFFFF;
+		qp = xa_load(&dev->rq_to_qp_lookup_table, rq_id);
+		mib_event.event = IB_EVENT_QP_FATAL;
+		mib_event.device = &dev->ib_dev;
+		if (qp && qp->ibqp.event_handler)
+			qp->ibqp.event_handler(&mib_event, qp->ibqp.qp_context);
+		else
+			ibdev_dbg(&dev->ib_dev, "found no qp or event handler");
+		ibdev_dbg(&dev->ib_dev, "Received critical notification");
+		break;
+	default:
+		ibdev_dbg(&dev->ib_dev, "Received unsolicited evt %d",
+			  event->type);
+	}
+}
+
+int mana_ib_create_error_eq(struct mana_ib_dev *dev)
+{
+	struct gdma_queue_spec spec = {};
+	int err;
+
+	spec.type = GDMA_EQ;
+	spec.monitor_avl_buf = false;
+	spec.queue_size = EQ_SIZE;
+	spec.eq.callback = mana_ib_critical_event_handler;
+	spec.eq.context = dev;
+	spec.eq.log2_throttle_limit = LOG2_EQ_THROTTLE;
+	spec.eq.msix_allocated = true;
+	spec.eq.msix_index = 0;
+	spec.doorbell = dev->gdma_dev->doorbell;
+	spec.pdid = dev->gdma_dev->pdid;
+
+	err = mana_gd_create_mana_eq(dev->gdma_dev, &spec,
+				     &dev->fatal_err_eq);
+	if (err)
+		return err;
+
+	dev->fatal_err_eq->eq.disable_needed = true;
+
+	return 0;
+}
diff --git a/drivers/infiniband/hw/mana/mana_ib.h b/drivers/infiniband/hw/mana/mana_ib.h
index 502cc8672eef..a5577c119def 100644
--- a/drivers/infiniband/hw/mana/mana_ib.h
+++ b/drivers/infiniband/hw/mana/mana_ib.h
@@ -30,6 +30,8 @@
 struct mana_ib_dev {
 	struct ib_device ib_dev;
 	struct gdma_dev *gdma_dev;
+	struct gdma_queue *fatal_err_eq;
+	struct xarray rq_to_qp_lookup_table;
 };
 
 struct mana_ib_wq {
@@ -159,4 +161,6 @@ 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_create_error_eq(struct mana_ib_dev *mdev);
+
 #endif
diff --git a/drivers/infiniband/hw/mana/qp.c b/drivers/infiniband/hw/mana/qp.c
index ae45d28eef5e..7ff9c8364551 100644
--- a/drivers/infiniband/hw/mana/qp.c
+++ b/drivers/infiniband/hw/mana/qp.c
@@ -211,6 +211,11 @@ static int mana_ib_create_qp_rss(struct ib_qp *ibqp, struct ib_pd *pd,
 		wq->id = wq_spec.queue_index;
 		cq->id = cq_spec.queue_index;
 
+		ret = xa_err(xa_store(&mdev->rq_to_qp_lookup_table,
+				      wq->id, qp, GFP_KERNEL));
+		if (ret)
+			goto fail;
+
 		ibdev_dbg(&mdev->ib_dev,
 			  "ret %d rx_object 0x%llx wq id %llu cq id %llu\n",
 			  ret, wq->rx_object, wq->id, cq->id);
@@ -246,6 +251,7 @@ static int mana_ib_create_qp_rss(struct ib_qp *ibqp, struct ib_pd *pd,
 	while (i-- > 0) {
 		ibwq = ind_tbl->ind_tbl[i];
 		wq = container_of(ibwq, struct mana_ib_wq, ibwq);
+		xa_erase(&mdev->rq_to_qp_lookup_table, wq->id);
 		mana_destroy_wq_obj(mpc, GDMA_RQ, wq->rx_object);
 	}
 
@@ -372,6 +378,11 @@ 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;
 
+	err = xa_err(xa_store(&mdev->rq_to_qp_lookup_table,
+			      qp->sq_id, qp, GFP_KERNEL));
+	if (err)
+		goto err_destroy_wq_obj;
+
 	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);
@@ -388,9 +399,11 @@ static int mana_ib_create_qp_raw(struct ib_qp *ibqp, struct ib_pd *ibpd,
 		goto err_destroy_wq_obj;
 	}
 
+
 	return 0;
 
 err_destroy_wq_obj:
+	xa_erase(&mdev->rq_to_qp_lookup_table, qp->sq_id);
 	mana_destroy_wq_obj(mpc, GDMA_SQ, qp->tx_object);
 
 err_destroy_dma_region:
@@ -455,6 +468,7 @@ static int mana_ib_destroy_qp_rss(struct mana_ib_qp *qp,
 		wq = container_of(ibwq, struct mana_ib_wq, ibwq);
 		ibdev_dbg(&mdev->ib_dev, "destroying wq->rx_object %llu\n",
 			  wq->rx_object);
+		xa_erase(&mdev->rq_to_qp_lookup_table, wq->id);
 		mana_destroy_wq_obj(mpc, GDMA_RQ, wq->rx_object);
 	}
 
@@ -477,6 +491,7 @@ static int mana_ib_destroy_qp_raw(struct mana_ib_qp *qp, struct ib_udata *udata)
 	mpc = netdev_priv(ndev);
 	pd = container_of(ibpd, struct mana_ib_pd, ibpd);
 
+	xa_erase(&mdev->rq_to_qp_lookup_table, qp->sq_id);
 	mana_destroy_wq_obj(mpc, GDMA_SQ, qp->tx_object);
 
 	if (qp->sq_umem) {
diff --git a/drivers/net/ethernet/microsoft/mana/gdma_main.c b/drivers/net/ethernet/microsoft/mana/gdma_main.c
index 02e50ed632ee..f368056d0b0b 100644
--- a/drivers/net/ethernet/microsoft/mana/gdma_main.c
+++ b/drivers/net/ethernet/microsoft/mana/gdma_main.c
@@ -204,7 +204,8 @@ void mana_gd_free_memory(struct gdma_mem_info *gmi)
 }
 
 static int mana_gd_create_hw_eq(struct gdma_context *gc,
-				struct gdma_queue *queue)
+				struct gdma_queue *queue,
+				u32 doorbell, u32 pdid)
 {
 	struct gdma_create_queue_resp resp = {};
 	struct gdma_create_queue_req req = {};
@@ -218,8 +219,8 @@ static int mana_gd_create_hw_eq(struct gdma_context *gc,
 
 	req.hdr.dev_id = queue->gdma_dev->dev_id;
 	req.type = queue->type;
-	req.pdid = queue->gdma_dev->pdid;
-	req.doolbell_id = queue->gdma_dev->doorbell;
+	req.pdid = pdid;
+	req.doolbell_id = doorbell;
 	req.gdma_region = queue->mem_info.dma_region_handle;
 	req.queue_size = queue->queue_size;
 	req.log2_throttle_limit = queue->eq.log2_throttle_limit;
@@ -393,53 +394,51 @@ static void mana_gd_process_eqe(struct gdma_queue *eq)
 	}
 }
 
-static void mana_gd_process_eq_events(void *arg)
+static void mana_gd_process_eq_events(struct list_head *eq_list)
 {
 	u32 owner_bits, new_bits, old_bits;
 	union gdma_eqe_info eqe_info;
 	struct gdma_eqe *eq_eqe_ptr;
-	struct gdma_queue *eq = arg;
 	struct gdma_context *gc;
+	struct gdma_queue *eq;
 	struct gdma_eqe *eqe;
 	u32 head, num_eqe;
 	int i;
 
-	gc = eq->gdma_dev->gdma_context;
-
-	num_eqe = eq->queue_size / GDMA_EQE_SIZE;
-	eq_eqe_ptr = eq->queue_mem_ptr;
-
-	/* Process up to 5 EQEs at a time, and update the HW head. */
-	for (i = 0; i < 5; i++) {
-		eqe = &eq_eqe_ptr[eq->head % num_eqe];
-		eqe_info.as_uint32 = eqe->eqe_info;
-		owner_bits = eqe_info.owner_bits;
-
-		old_bits = (eq->head / num_eqe - 1) & GDMA_EQE_OWNER_MASK;
-		/* No more entries */
-		if (owner_bits == old_bits)
-			break;
-
-		new_bits = (eq->head / num_eqe) & GDMA_EQE_OWNER_MASK;
-		if (owner_bits != new_bits) {
-			dev_err(gc->dev, "EQ %d: overflow detected\n", eq->id);
-			break;
+	list_for_each_entry_rcu(eq, eq_list, entry) {
+		gc = eq->gdma_dev->gdma_context;
+
+		num_eqe = eq->queue_size / GDMA_EQE_SIZE;
+		eq_eqe_ptr = eq->queue_mem_ptr;
+		/* Process up to 5 EQEs at a time, and update the HW head. */
+		for (i = 0; i < 5; i++) {
+			eqe = &eq_eqe_ptr[eq->head % num_eqe];
+			eqe_info.as_uint32 = eqe->eqe_info;
+			owner_bits = eqe_info.owner_bits;
+
+			old_bits = (eq->head / num_eqe - 1) & GDMA_EQE_OWNER_MASK;
+			/* No more entries */
+			if (owner_bits == old_bits)
+				break;
+
+			new_bits = (eq->head / num_eqe) & GDMA_EQE_OWNER_MASK;
+			if (owner_bits != new_bits) {
+				dev_err(gc->dev, "EQ %d: overflow detected\n",
+					eq->id);
+				break;
+			}
+			/* Per GDMA spec, rmb is necessary after checking owner_bits, before
+			 * reading eqe.
+			 */
+			rmb();
+			mana_gd_process_eqe(eq);
+			eq->head++;
 		}
 
-		/* Per GDMA spec, rmb is necessary after checking owner_bits, before
-		 * reading eqe.
-		 */
-		rmb();
-
-		mana_gd_process_eqe(eq);
-
-		eq->head++;
+		head = eq->head % (num_eqe << GDMA_EQE_OWNER_BITS);
+		mana_gd_ring_doorbell(gc, eq->gdma_dev->doorbell, eq->type,
+				      eq->id, head, SET_ARM_BIT);
 	}
-
-	head = eq->head % (num_eqe << GDMA_EQE_OWNER_BITS);
-
-	mana_gd_ring_doorbell(gc, eq->gdma_dev->doorbell, eq->type, eq->id,
-			      head, SET_ARM_BIT);
 }
 
 static int mana_gd_register_irq(struct gdma_queue *queue,
@@ -457,45 +456,48 @@ static int mana_gd_register_irq(struct gdma_queue *queue,
 	gc = gd->gdma_context;
 	r = &gc->msix_resource;
 	dev = gc->dev;
+	msi_index = spec->eq.msix_index;
 
 	spin_lock_irqsave(&r->lock, flags);
 
-	msi_index = find_first_zero_bit(r->map, r->size);
-	if (msi_index >= r->size || msi_index >= gc->num_msix_usable) {
-		err = -ENOSPC;
-	} else {
-		bitmap_set(r->map, msi_index, 1);
-		queue->eq.msix_index = msi_index;
-	}
-
-	spin_unlock_irqrestore(&r->lock, flags);
+	if (!spec->eq.msix_allocated) {
+		msi_index = find_first_zero_bit(r->map, r->size);
 
-	if (err) {
-		dev_err(dev, "Register IRQ err:%d, msi:%u rsize:%u, nMSI:%u",
-			err, msi_index, r->size, gc->num_msix_usable);
+		if (msi_index >= r->size ||
+		    msi_index >= gc->num_msix_usable)
+			err = -ENOSPC;
+		else
+			bitmap_set(r->map, msi_index, 1);
 
-		return err;
+		if (err) {
+			dev_err(dev, "Register IRQ err:%d, msi:%u rsize:%u, nMSI:%u",
+				err, msi_index, r->size, gc->num_msix_usable);
+				goto out;
+		}
 	}
 
+	queue->eq.msix_index = msi_index;
 	gic = &gc->irq_contexts[msi_index];
 
-	WARN_ON(gic->handler || gic->arg);
-
-	gic->arg = queue;
+	list_add_rcu(&queue->entry, &gic->eq_list);
 
 	gic->handler = mana_gd_process_eq_events;
 
-	return 0;
+out:
+	spin_unlock_irqrestore(&r->lock, flags);
+	return err;
 }
 
-static void mana_gd_deregiser_irq(struct gdma_queue *queue)
+static void mana_gd_deregister_irq(struct gdma_queue *queue)
 {
 	struct gdma_dev *gd = queue->gdma_dev;
 	struct gdma_irq_context *gic;
 	struct gdma_context *gc;
 	struct gdma_resource *r;
 	unsigned int msix_index;
+	struct gdma_queue *eq;
 	unsigned long flags;
+	struct list_head *p;
 
 	gc = gd->gdma_context;
 	r = &gc->msix_resource;
@@ -505,14 +507,24 @@ static void mana_gd_deregiser_irq(struct gdma_queue *queue)
 	if (WARN_ON(msix_index >= gc->num_msix_usable))
 		return;
 
+	spin_lock_irqsave(&r->lock, flags);
+
 	gic = &gc->irq_contexts[msix_index];
-	gic->handler = NULL;
-	gic->arg = NULL;
+	list_for_each_rcu(p, &gic->eq_list) {
+		eq = list_entry(p, struct gdma_queue, entry);
+		if (queue == eq) {
+			list_del(&eq->entry);
+			synchronize_rcu();
+			break;
+		}
+	}
 
-	spin_lock_irqsave(&r->lock, flags);
-	bitmap_clear(r->map, msix_index, 1);
-	spin_unlock_irqrestore(&r->lock, flags);
+	if (list_empty(&gic->eq_list)) {
+		gic->handler = NULL;
+		bitmap_clear(r->map, msix_index, 1);
+	}
 
+	spin_unlock_irqrestore(&r->lock, flags);
 	queue->eq.msix_index = INVALID_PCI_MSIX_INDEX;
 }
 
@@ -575,7 +587,7 @@ static void mana_gd_destroy_eq(struct gdma_context *gc, bool flush_evenets,
 			dev_warn(gc->dev, "Failed to flush EQ: %d\n", err);
 	}
 
-	mana_gd_deregiser_irq(queue);
+	mana_gd_deregister_irq(queue);
 
 	if (queue->eq.disable_needed)
 		mana_gd_disable_queue(queue);
@@ -590,7 +602,7 @@ static int mana_gd_create_eq(struct gdma_dev *gd,
 	u32 log2_num_entries;
 	int err;
 
-	queue->eq.msix_index = INVALID_PCI_MSIX_INDEX;
+	queue->eq.msix_index = spec->eq.msix_index;
 
 	log2_num_entries = ilog2(queue->queue_size / GDMA_EQE_SIZE);
 
@@ -612,7 +624,8 @@ static int mana_gd_create_eq(struct gdma_dev *gd,
 	queue->eq.log2_throttle_limit = spec->eq.log2_throttle_limit ?: 1;
 
 	if (create_hwq) {
-		err = mana_gd_create_hw_eq(gc, queue);
+		err = mana_gd_create_hw_eq(gc, queue,
+					   spec->doorbell, spec->pdid);
 		if (err)
 			goto out;
 
@@ -822,6 +835,7 @@ int mana_gd_create_mana_eq(struct gdma_dev *gd,
 	kfree(queue);
 	return err;
 }
+EXPORT_SYMBOL(mana_gd_create_mana_eq);
 
 int mana_gd_create_mana_wq_cq(struct gdma_dev *gd,
 			      const struct gdma_queue_spec *spec,
@@ -898,6 +912,7 @@ void mana_gd_destroy_queue(struct gdma_context *gc, struct gdma_queue *queue)
 	mana_gd_free_memory(gmi);
 	kfree(queue);
 }
+EXPORT_SYMBOL(mana_gd_destroy_queue);
 
 int mana_gd_verify_vf_version(struct pci_dev *pdev)
 {
@@ -1224,7 +1239,7 @@ static irqreturn_t mana_gd_intr(int irq, void *arg)
 	struct gdma_irq_context *gic = arg;
 
 	if (gic->handler)
-		gic->handler(gic->arg);
+		gic->handler(&gic->eq_list);
 
 	return IRQ_HANDLED;
 }
@@ -1277,7 +1292,7 @@ static int mana_gd_setup_irqs(struct pci_dev *pdev)
 	for (i = 0; i < nvec; i++) {
 		gic = &gc->irq_contexts[i];
 		gic->handler = NULL;
-		gic->arg = NULL;
+		INIT_LIST_HEAD(&gic->eq_list);
 
 		if (!i)
 			snprintf(gic->name, MANA_IRQ_NAME_SZ, "mana_hwc@pci:%s",
diff --git a/drivers/net/ethernet/microsoft/mana/mana_en.c b/drivers/net/ethernet/microsoft/mana/mana_en.c
index fc3d2903a80f..abf63f405940 100644
--- a/drivers/net/ethernet/microsoft/mana/mana_en.c
+++ b/drivers/net/ethernet/microsoft/mana/mana_en.c
@@ -1242,6 +1242,9 @@ static int mana_create_eq(struct mana_context *ac)
 	spec.eq.callback = NULL;
 	spec.eq.context = ac->eqs;
 	spec.eq.log2_throttle_limit = LOG2_EQ_THROTTLE;
+	spec.eq.msix_allocated = false;
+	spec.doorbell = gd->doorbell;
+	spec.pdid = gd->pdid;
 
 	for (i = 0; i < gc->max_num_queues; i++) {
 		err = mana_gd_create_mana_eq(gd, &spec, &ac->eqs[i].eq);
diff --git a/include/net/mana/gdma.h b/include/net/mana/gdma.h
index 000f0d7670f7..e32c75639557 100644
--- a/include/net/mana/gdma.h
+++ b/include/net/mana/gdma.h
@@ -60,6 +60,11 @@ enum gdma_eqe_type {
 	GDMA_EQE_HWC_INIT_DONE		= 131,
 	GDMA_EQE_HWC_SOC_RECONFIG	= 132,
 	GDMA_EQE_HWC_SOC_RECONFIG_DATA	= 133,
+
+	/* RDMA SOC Events */
+	GDMA_EQE_SOC_EVENT_NOTIFICATION = 176,
+	GDMA_EQE_SOC_EVENT_TEST = 177,
+
 };
 
 enum {
@@ -294,6 +299,7 @@ struct gdma_queue {
 
 	u32 head;
 	u32 tail;
+	struct list_head entry;
 
 	/* Extra fields specific to EQ/CQ. */
 	union {
@@ -321,6 +327,8 @@ struct gdma_queue_spec {
 	enum gdma_queue_type type;
 	bool monitor_avl_buf;
 	unsigned int queue_size;
+	u32 doorbell;
+	u32 pdid;
 
 	/* Extra fields specific to EQ/CQ. */
 	union {
@@ -329,6 +337,8 @@ struct gdma_queue_spec {
 			void *context;
 
 			unsigned long log2_throttle_limit;
+			bool msix_allocated;
+			unsigned int msix_index;
 		} eq;
 
 		struct {
@@ -344,8 +354,8 @@ struct gdma_queue_spec {
 #define MANA_IRQ_NAME_SZ 32
 
 struct gdma_irq_context {
-	void (*handler)(void *arg);
-	void *arg;
+	void (*handler)(struct list_head *arg);
+	struct list_head eq_list;
 	char name[MANA_IRQ_NAME_SZ];
 };
 
-- 
2.34.1

[Patch v1 3/4] RDMA/mana_ib: create RDMA adapter handle

[longli]

From: Long Li <longli@microsoft.com>

Create the RDMA device handle with the SoC using the management EQ created
earlier.

Signed-off-by: Long Li <longli@microsoft.com>
---
 drivers/infiniband/hw/mana/device.c  | 10 ++++--
 drivers/infiniband/hw/mana/main.c    | 51 ++++++++++++++++++++++++++++
 drivers/infiniband/hw/mana/mana_ib.h | 30 ++++++++++++++++
 3 files changed, 88 insertions(+), 3 deletions(-)

diff --git a/drivers/infiniband/hw/mana/device.c b/drivers/infiniband/hw/mana/device.c
index 3da4763e1a0c..5e5aa75230c2 100644
--- a/drivers/infiniband/hw/mana/device.c
+++ b/drivers/infiniband/hw/mana/device.c
@@ -93,9 +93,10 @@ static int mana_ib_probe(struct auxiliary_device *adev,
 		goto deregister_device;
 	}
 
+	ret = mana_ib_create_adapter(dev);
 	if (ret) {
-		ib_dealloc_device(&dev->ib_dev);
-		return ret;
+		ibdev_err(&dev->ib_dev, "Failed to create adapter");
+		goto free_error_eq;
 	}
 
 	ret = ib_register_device(&dev->ib_dev, "mana_%d",
@@ -108,8 +109,10 @@ static int mana_ib_probe(struct auxiliary_device *adev,
 	return 0;
 
 destroy_adapter:
-	mana_gd_destroy_queue(dev->gdma_dev->gdma_context, dev->fatal_err_eq);
+	mana_ib_destroy_adapter(dev);
+free_error_eq:
 	xa_destroy(&dev->rq_to_qp_lookup_table);
+	mana_gd_destroy_queue(dev->gdma_dev->gdma_context, dev->fatal_err_eq);
 deregister_device:
 	mana_gd_deregister_device(dev->gdma_dev);
 free_ib_device:
@@ -123,6 +126,7 @@ static void mana_ib_remove(struct auxiliary_device *adev)
 
 	ib_unregister_device(&dev->ib_dev);
 
+	mana_ib_destroy_adapter(dev);
 	mana_gd_destroy_queue(dev->gdma_dev->gdma_context, dev->fatal_err_eq);
 	xa_destroy(&dev->rq_to_qp_lookup_table);
 	mana_gd_deregister_device(dev->gdma_dev);
diff --git a/drivers/infiniband/hw/mana/main.c b/drivers/infiniband/hw/mana/main.c
index 032f926bf1ab..4f4343d14041 100644
--- a/drivers/infiniband/hw/mana/main.c
+++ b/drivers/infiniband/hw/mana/main.c
@@ -522,6 +522,57 @@ void mana_ib_disassociate_ucontext(struct ib_ucontext *ibcontext)
 {
 }
 
+int mana_ib_destroy_adapter(struct mana_ib_dev *dev)
+{
+	struct mana_ib_destroy_adapter_resp resp = {};
+	struct mana_ib_destroy_adapter_req req = {};
+	struct gdma_context *gc;
+	int err;
+
+	gc = dev->gdma_dev->gdma_context;
+
+	mana_gd_init_req_hdr(&req.hdr, MANA_IB_DESTROY_ADAPTER, sizeof(req),
+			     sizeof(resp));
+	req.adapter = dev->adapter_handle;
+	req.hdr.dev_id = gc->mana_ib.dev_id;
+
+	err = mana_gd_send_request(gc, sizeof(req), &req, sizeof(resp), &resp);
+
+	if (err) {
+		ibdev_err(&dev->ib_dev, "Failed to destroy adapter err %d", err);
+		return err;
+	}
+
+	return 0;
+}
+
+int mana_ib_create_adapter(struct mana_ib_dev *dev)
+{
+	struct mana_ib_create_adapter_resp resp = {};
+	struct mana_ib_create_adapter_req req = {};
+	struct gdma_context *gc;
+	int err;
+
+	gc = dev->gdma_dev->gdma_context;
+
+	mana_gd_init_req_hdr(&req.hdr, MANA_IB_CREATE_ADAPTER, sizeof(req),
+			     sizeof(resp));
+	req.notify_eq_id = dev->fatal_err_eq->id;
+	req.hdr.dev_id = gc->mana_ib.dev_id;
+
+	err = mana_gd_send_request(gc, sizeof(req), &req, sizeof(resp), &resp);
+
+	if (err) {
+		ibdev_err(&dev->ib_dev, "Failed to create adapter err %d",
+			  err);
+		return err;
+	}
+
+	dev->adapter_handle = resp.adapter;
+
+	return 0;
+}
+
 static void mana_ib_critical_event_handler(void *ctx, struct gdma_queue *queue,
 				      struct gdma_event *event)
 {
diff --git a/drivers/infiniband/hw/mana/mana_ib.h b/drivers/infiniband/hw/mana/mana_ib.h
index a5577c119def..4286caf0d67c 100644
--- a/drivers/infiniband/hw/mana/mana_ib.h
+++ b/drivers/infiniband/hw/mana/mana_ib.h
@@ -31,6 +31,7 @@ struct mana_ib_dev {
 	struct ib_device ib_dev;
 	struct gdma_dev *gdma_dev;
 	struct gdma_queue *fatal_err_eq;
+	mana_handle_t adapter_handle;
 	struct xarray rq_to_qp_lookup_table;
 };
 
@@ -94,6 +95,31 @@ struct mana_ib_rwq_ind_table {
 	struct ib_rwq_ind_table ib_ind_table;
 };
 
+enum mana_ib_command_code {
+	MANA_IB_CREATE_ADAPTER  = 0x30002,
+	MANA_IB_DESTROY_ADAPTER = 0x30003,
+};
+
+struct mana_ib_create_adapter_req {
+	struct gdma_req_hdr hdr;
+	u32 notify_eq_id;
+	u32 reserved;
+}; /*HW Data */
+
+struct mana_ib_create_adapter_resp {
+	struct gdma_resp_hdr hdr;
+	mana_handle_t adapter;
+}; /* HW Data */
+
+struct mana_ib_destroy_adapter_req {
+	struct gdma_req_hdr hdr;
+	mana_handle_t adapter;
+}; /*HW Data */
+
+struct mana_ib_destroy_adapter_resp {
+	struct gdma_resp_hdr hdr;
+}; /* HW Data */
+
 int mana_ib_gd_create_dma_region(struct mana_ib_dev *dev, struct ib_umem *umem,
 				 mana_handle_t *gdma_region);
 
@@ -163,4 +189,8 @@ void mana_ib_disassociate_ucontext(struct ib_ucontext *ibcontext);
 
 int mana_ib_create_error_eq(struct mana_ib_dev *mdev);
 
+int mana_ib_create_adapter(struct mana_ib_dev *mdev);
+
+int mana_ib_destroy_adapter(struct mana_ib_dev *mdev);
+
 #endif
-- 
2.34.1

[Patch v1 4/4] RDMA/mana_ib: query device capabilities

[longli]

From: Long Li <longli@microsoft.com>

With RDMA device handle created, 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  |  6 +++
 drivers/infiniband/hw/mana/main.c    | 59 +++++++++++++++++++++++-----
 drivers/infiniband/hw/mana/mana_ib.h | 51 ++++++++++++++++++++++++
 drivers/infiniband/hw/mana/qp.c      |  6 ++-
 include/net/mana/gdma.h              |  1 +
 6 files changed, 112 insertions(+), 13 deletions(-)

diff --git a/drivers/infiniband/hw/mana/cq.c b/drivers/infiniband/hw/mana/cq.c
index d141cab8a1e6..71064f17c235 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_requester_sq_size) {
 		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 5e5aa75230c2..06f53df072bb 100644
--- a/drivers/infiniband/hw/mana/device.c
+++ b/drivers/infiniband/hw/mana/device.c
@@ -99,6 +99,12 @@ static int mana_ib_probe(struct auxiliary_device *adev,
 		goto free_error_eq;
 	}
 
+	ret = mana_ib_query_adapter_caps(dev);
+	if (ret) {
+		ibdev_dbg(&dev->ib_dev, "Failed to get device caps");
+		goto destroy_adapter;
+	}
+
 	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 4f4343d14041..bf27fa395a67 100644
--- a/drivers/infiniband/hw/mana/main.c
+++ b/drivers/infiniband/hw/mana/main.c
@@ -486,18 +486,14 @@ 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_requester_sq_size;
+	props->max_cqe = dev->adapter_caps.max_requester_sq_size;
+	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;
 
@@ -624,3 +620,46 @@ int mana_ib_create_error_eq(struct mana_ib_dev *dev)
 
 	return 0;
 }
+
+int mana_ib_query_adapter_caps(struct mana_ib_dev *dev)
+{
+	struct mana_ib_query_adapter_caps_resp resp = {};
+	struct mana_ib_query_adapter_caps_req req = {};
+	struct mana_ib_adapter_caps *caps = &dev->adapter_caps;
+	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_requester_sq_size = resp.max_requester_sq_size;
+	caps->max_responder_sq_size = resp.max_responder_sq_size;
+	caps->max_requester_rq_size = resp.max_requester_rq_size;
+	caps->max_responder_rq_size = resp.max_responder_rq_size;
+	caps->max_inline_data_size = resp.max_inline_data_size;
+	caps->max_send_wqe_size = MAX_TX_WQE_SGL_ENTRIES;
+	caps->max_recv_wqe_size = MAX_RX_WQE_SGL_ENTRIES;
+
+	return 0;
+}
diff --git a/drivers/infiniband/hw/mana/mana_ib.h b/drivers/infiniband/hw/mana/mana_ib.h
index 4286caf0d67c..d7a56b075fbc 100644
--- a/drivers/infiniband/hw/mana/mana_ib.h
+++ b/drivers/infiniband/hw/mana/mana_ib.h
@@ -27,11 +27,33 @@
  */
 #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_requester_sq_size;
+	u32 max_responder_sq_size;
+	u32 max_requester_rq_size;
+	u32 max_responder_rq_size;
+	u32 max_send_wqe_size;
+	u32 max_recv_wqe_size;
+	u32 max_inline_data_size;
+};
+
 struct mana_ib_dev {
 	struct ib_device ib_dev;
 	struct gdma_dev *gdma_dev;
 	struct gdma_queue *fatal_err_eq;
 	mana_handle_t adapter_handle;
+	struct mana_ib_adapter_caps adapter_caps;
 	struct xarray rq_to_qp_lookup_table;
 };
 
@@ -96,6 +118,7 @@ struct mana_ib_rwq_ind_table {
 };
 
 enum mana_ib_command_code {
+	MANA_IB_GET_ADAPTER_CAP = 0x30001,
 	MANA_IB_CREATE_ADAPTER  = 0x30002,
 	MANA_IB_DESTROY_ADAPTER = 0x30003,
 };
@@ -120,6 +143,32 @@ struct mana_ib_destroy_adapter_resp {
 	struct gdma_resp_hdr hdr;
 }; /* HW Data */
 
+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_wqe_size;
+	u32 max_recv_wqe_size;
+	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);
 
@@ -193,4 +242,6 @@ int mana_ib_create_adapter(struct mana_ib_dev *mdev);
 
 int mana_ib_destroy_adapter(struct mana_ib_dev *mdev);
 
+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 7ff9c8364551..7211e93d999c 100644
--- a/drivers/infiniband/hw/mana/qp.c
+++ b/drivers/infiniband/hw/mana/qp.c
@@ -130,7 +130,8 @@ 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_requester_sq_size) {
 		ibdev_dbg(&mdev->ib_dev,
 			  "Requested max_recv_wr %d exceeding limit\n",
 			  attr->cap.max_recv_wr);
@@ -302,7 +303,8 @@ 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_requester_sq_size) {
 		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 e32c75639557..c1850ec7faae 100644
--- a/include/net/mana/gdma.h
+++ b/include/net/mana/gdma.h
@@ -155,6 +155,7 @@ 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;
-- 
2.34.1

Re: [Patch v1 2/4] RDMA/mana_ib: create and process EQ events

[Simon Horman]

On Wed, Nov 22, 2023 at 07:10:08PM -0800, longli@linuxonhyperv.com wrote:
> From: Long Li <longli@microsoft.com>
> 
> Before the software can create an RDMA adapter handle with SoC, it needs to
> create EQs for processing SoC events from RDMA device. Because MSI-X
> vectors are shared between MANA Ethernet device and RDMA device, this
> patch adds support to share EQs on MSI-X vectors and creates management
> EQ for RDMA device.
> 
> Signed-off-by: Long Li <longli@microsoft.com>

...

> diff --git a/drivers/net/ethernet/microsoft/mana/gdma_main.c b/drivers/net/ethernet/microsoft/mana/gdma_main.c

...

> -static void mana_gd_deregiser_irq(struct gdma_queue *queue)
> +static void mana_gd_deregister_irq(struct gdma_queue *queue)
>  {
>  	struct gdma_dev *gd = queue->gdma_dev;
>  	struct gdma_irq_context *gic;
>  	struct gdma_context *gc;
>  	struct gdma_resource *r;
>  	unsigned int msix_index;
> +	struct gdma_queue *eq;
>  	unsigned long flags;
> +	struct list_head *p;
>  
>  	gc = gd->gdma_context;
>  	r = &gc->msix_resource;
> @@ -505,14 +507,24 @@ static void mana_gd_deregiser_irq(struct gdma_queue *queue)
>  	if (WARN_ON(msix_index >= gc->num_msix_usable))
>  		return;
>  
> +	spin_lock_irqsave(&r->lock, flags);
> +
>  	gic = &gc->irq_contexts[msix_index];
> -	gic->handler = NULL;
> -	gic->arg = NULL;
> +	list_for_each_rcu(p, &gic->eq_list) {
> +		eq = list_entry(p, struct gdma_queue, entry);

Hi Long Li,

Sparse complains a bit about this construction:

 .../gdma_main.c:513:9: error: incompatible types in comparison expression (different address spaces):
 .../gdma_main.c:513:9:    struct list_head [noderef] __rcu *
 .../gdma_main.c:513:9:    struct list_head *
 .../gdma_main.c:513:9: error: incompatible types in comparison expression (different address spaces):
 .../gdma_main.c:513:9:    struct list_head [noderef] __rcu *
 .../gdma_main.c:513:9:    struct list_head *

Perhaps using list_for_each_entry_rcu() is appropriate here.


> +		if (queue == eq) {
> +			list_del(&eq->entry);
> +			synchronize_rcu();
> +			break;
> +		}
> +	}
>  
> -	spin_lock_irqsave(&r->lock, flags);
> -	bitmap_clear(r->map, msix_index, 1);
> -	spin_unlock_irqrestore(&r->lock, flags);
> +	if (list_empty(&gic->eq_list)) {
> +		gic->handler = NULL;
> +		bitmap_clear(r->map, msix_index, 1);
> +	}
>  
> +	spin_unlock_irqrestore(&r->lock, flags);
>  	queue->eq.msix_index = INVALID_PCI_MSIX_INDEX;
>  }
>  

...

RE: [Patch v1 2/4] RDMA/mana_ib: create and process EQ events

[Long Li]

> > @@ -505,14 +507,24 @@ static void mana_gd_deregiser_irq(struct
> gdma_queue *queue)
> >       if (WARN_ON(msix_index >= gc->num_msix_usable))
> >               return;
> >
> > +     spin_lock_irqsave(&r->lock, flags);
> > +
> >       gic = &gc->irq_contexts[msix_index];
> > -     gic->handler = NULL;
> > -     gic->arg = NULL;
> > +     list_for_each_rcu(p, &gic->eq_list) {
> > +             eq = list_entry(p, struct gdma_queue, entry);
> 
> Hi Long Li,
> 
> Sparse complains a bit about this construction:
> 
>  .../gdma_main.c:513:9: error: incompatible types in comparison expression
> (different address spaces):
>  .../gdma_main.c:513:9:    struct list_head [noderef] __rcu *
>  .../gdma_main.c:513:9:    struct list_head *
>  .../gdma_main.c:513:9: error: incompatible types in comparison expression
> (different address spaces):
>  .../gdma_main.c:513:9:    struct list_head [noderef] __rcu *
>  .../gdma_main.c:513:9:    struct list_head *
> 
> Perhaps using list_for_each_entry_rcu() is appropriate here.

Thank you, I will fix this.

Long