[PATCH rdma-next v4 0/4] RDMA/mana_ib: allow separate mana_ib for each mana client

[PATCH rdma-next v4 0/4] RDMA/mana_ib: allow separate mana_ib for each mana client

[Konstantin Taranov]

From: Konstantin Taranov <kotaranov@microsoft.com>

Microsoft mana adapter has 2 devices in the HW: mana ethernet device and RNIC device.
Both devices can implement RDMA drivers and, so far, they have been sharing
one ib device context. However, they are different devices with different
capabilities in the HW and have different lifetime model.

This series allows us to model the aforementioned two devices as separate ib devices.
The mana_ib will continue supporting two devices but as individual ib devices.
It enables the driver to dynamically destroy and create the auxiliary device over
RNIC, when the HW reboots the RNIC module. Without this separation, the reboot
would cause destruction of the ib device serving DPDK clients from the uninterrupted
ethernet HW module.

v2:
- renamed aux device from mana.dpdk to mana.eth (patch 1 and 2)
- Fixed a possible race between servicing and pci threads (patch 4)

v3:
- Added vendorid and partid in mana_ib_query_device (patch 2)

v4:
- rabased on latest rdma-next

Konstantin Taranov (3):
  net: mana: Probe rdma device in mana driver
  RDMA/mana_ib: Add support of mana_ib for RNIC and ETH nic
  RDMA/mana_ib: unify mana_ib functions to support any gdma device

Shiraz Saleem (1):
  net: mana: Add support for auxiliary device servicing events

 drivers/infiniband/hw/mana/cq.c               |   4 +-
 drivers/infiniband/hw/mana/device.c           | 174 +++++++++---------
 drivers/infiniband/hw/mana/main.c             |  82 +++++++--
 drivers/infiniband/hw/mana/mana_ib.h          |   6 +
 drivers/infiniband/hw/mana/qp.c               |   5 +-
 .../net/ethernet/microsoft/mana/gdma_main.c   |  26 ++-
 .../net/ethernet/microsoft/mana/hw_channel.c  |  20 ++
 drivers/net/ethernet/microsoft/mana/mana_en.c | 108 ++++++++++-
 include/net/mana/gdma.h                       |  19 ++
 include/net/mana/hw_channel.h                 |   9 +
 include/net/mana/mana.h                       |   3 +
 11 files changed, 333 insertions(+), 123 deletions(-)

-- 
2.43.0

[PATCH rdma-next v4 1/4] net: mana: Probe rdma device in mana driver

[Konstantin Taranov]

From: Konstantin Taranov <kotaranov@microsoft.com>

Initialize gdma device for rdma inside mana module.
For each gdma device, initialize an auxiliary ib device.

Signed-off-by: Konstantin Taranov <kotaranov@microsoft.com>
---
 .../net/ethernet/microsoft/mana/gdma_main.c   | 15 ++++++-
 drivers/net/ethernet/microsoft/mana/mana_en.c | 39 +++++++++++++++++--
 include/net/mana/mana.h                       |  3 ++
 3 files changed, 52 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ethernet/microsoft/mana/gdma_main.c b/drivers/net/ethernet/microsoft/mana/gdma_main.c
index 8ee1aa3..59e7814 100644
--- a/drivers/net/ethernet/microsoft/mana/gdma_main.c
+++ b/drivers/net/ethernet/microsoft/mana/gdma_main.c
@@ -1005,7 +1005,6 @@ 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)
 {
@@ -1036,7 +1035,6 @@ 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)
 {
@@ -1579,8 +1577,14 @@ static int mana_gd_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
 	if (err)
 		goto cleanup_gd;
 
+	err = mana_rdma_probe(&gc->mana_ib);
+	if (err)
+		goto cleanup_mana;
+
 	return 0;
 
+cleanup_mana:
+	mana_remove(&gc->mana, false);
 cleanup_gd:
 	mana_gd_cleanup(pdev);
 unmap_bar:
@@ -1608,6 +1612,7 @@ static void mana_gd_remove(struct pci_dev *pdev)
 {
 	struct gdma_context *gc = pci_get_drvdata(pdev);
 
+	mana_rdma_remove(&gc->mana_ib);
 	mana_remove(&gc->mana, false);
 
 	mana_gd_cleanup(pdev);
@@ -1631,6 +1636,7 @@ static int mana_gd_suspend(struct pci_dev *pdev, pm_message_t state)
 {
 	struct gdma_context *gc = pci_get_drvdata(pdev);
 
+	mana_rdma_remove(&gc->mana_ib);
 	mana_remove(&gc->mana, true);
 
 	mana_gd_cleanup(pdev);
@@ -1655,6 +1661,10 @@ static int mana_gd_resume(struct pci_dev *pdev)
 	if (err)
 		return err;
 
+	err = mana_rdma_probe(&gc->mana_ib);
+	if (err)
+		return err;
+
 	return 0;
 }
 
@@ -1665,6 +1675,7 @@ static void mana_gd_shutdown(struct pci_dev *pdev)
 
 	dev_info(&pdev->dev, "Shutdown was called\n");
 
+	mana_rdma_remove(&gc->mana_ib);
 	mana_remove(&gc->mana, true);
 
 	mana_gd_cleanup(pdev);
diff --git a/drivers/net/ethernet/microsoft/mana/mana_en.c b/drivers/net/ethernet/microsoft/mana/mana_en.c
index 5be0585..2013d0e 100644
--- a/drivers/net/ethernet/microsoft/mana/mana_en.c
+++ b/drivers/net/ethernet/microsoft/mana/mana_en.c
@@ -2944,7 +2944,7 @@ static void remove_adev(struct gdma_dev *gd)
 	gd->adev = NULL;
 }
 
-static int add_adev(struct gdma_dev *gd)
+static int add_adev(struct gdma_dev *gd, const char *name)
 {
 	struct auxiliary_device *adev;
 	struct mana_adev *madev;
@@ -2960,7 +2960,7 @@ static int add_adev(struct gdma_dev *gd)
 		goto idx_fail;
 	adev->id = ret;
 
-	adev->name = "rdma";
+	adev->name = name;
 	adev->dev.parent = gd->gdma_context->dev;
 	adev->dev.release = adev_release;
 	madev->mdev = gd;
@@ -3076,7 +3076,7 @@ int mana_probe(struct gdma_dev *gd, bool resuming)
 		}
 	}
 
-	err = add_adev(gd);
+	err = add_adev(gd, "eth");
 out:
 	if (err) {
 		mana_remove(gd, false);
@@ -3150,6 +3150,39 @@ out:
 	dev_dbg(dev, "%s succeeded\n", __func__);
 }
 
+int mana_rdma_probe(struct gdma_dev *gd)
+{
+	int err = 0;
+
+	if (gd->dev_id.type != GDMA_DEVICE_MANA_IB) {
+		/* RDMA device is not detected on pci */
+		return err;
+	}
+
+	err = mana_gd_register_device(gd);
+	if (err)
+		return err;
+
+	err = add_adev(gd, "rdma");
+	if (err)
+		mana_gd_deregister_device(gd);
+
+	return err;
+}
+
+void mana_rdma_remove(struct gdma_dev *gd)
+{
+	if (gd->dev_id.type != GDMA_DEVICE_MANA_IB) {
+		/* RDMA device is not detected on pci */
+		return;
+	}
+
+	if (gd->adev)
+		remove_adev(gd);
+
+	mana_gd_deregister_device(gd);
+}
+
 struct net_device *mana_get_primary_netdev(struct mana_context *ac,
 					   u32 port_index,
 					   netdevice_tracker *tracker)
diff --git a/include/net/mana/mana.h b/include/net/mana/mana.h
index 0f78065..5857efc 100644
--- a/include/net/mana/mana.h
+++ b/include/net/mana/mana.h
@@ -488,6 +488,9 @@ int mana_detach(struct net_device *ndev, bool from_close);
 int mana_probe(struct gdma_dev *gd, bool resuming);
 void mana_remove(struct gdma_dev *gd, bool suspending);
 
+int mana_rdma_probe(struct gdma_dev *gd);
+void mana_rdma_remove(struct gdma_dev *gd);
+
 void mana_xdp_tx(struct sk_buff *skb, struct net_device *ndev);
 int mana_xdp_xmit(struct net_device *ndev, int n, struct xdp_frame **frames,
 		  u32 flags);
-- 
2.43.0

RE: [PATCH rdma-next v4 1/4] net: mana: Probe rdma device in mana driver

[Long Li]

> -----Original Message-----
> From: Konstantin Taranov <kotaranov@linux.microsoft.com>
> Sent: Wednesday, May 7, 2025 8:59 AM
> To: Konstantin Taranov <kotaranov@microsoft.com>; pabeni@redhat.com;
> Haiyang Zhang <haiyangz@microsoft.com>; KY Srinivasan <kys@microsoft.com>;
> edumazet@google.com; kuba@kernel.org; davem@davemloft.net; Dexuan Cui
> <decui@microsoft.com>; wei.liu@kernel.org; Long Li <longli@microsoft.com>;
> jgg@ziepe.ca; leon@kernel.org
> Cc: linux-rdma@vger.kernel.org; linux-kernel@vger.kernel.org;
> netdev@vger.kernel.org
> Subject: [PATCH rdma-next v4 1/4] net: mana: Probe rdma device in mana driver
> 
> From: Konstantin Taranov <kotaranov@microsoft.com>
> 
> Initialize gdma device for rdma inside mana module.
> For each gdma device, initialize an auxiliary ib device.
> 
> Signed-off-by: Konstantin Taranov <kotaranov@microsoft.com>

Reviewed-by: Long Li <longli@microsoft.com>

> ---
>  .../net/ethernet/microsoft/mana/gdma_main.c   | 15 ++++++-
>  drivers/net/ethernet/microsoft/mana/mana_en.c | 39 +++++++++++++++++--
>  include/net/mana/mana.h                       |  3 ++
>  3 files changed, 52 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/net/ethernet/microsoft/mana/gdma_main.c
> b/drivers/net/ethernet/microsoft/mana/gdma_main.c
> index 8ee1aa3..59e7814 100644
> --- a/drivers/net/ethernet/microsoft/mana/gdma_main.c
> +++ b/drivers/net/ethernet/microsoft/mana/gdma_main.c
> @@ -1005,7 +1005,6 @@ 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)  { @@ -1036,7 +1035,6
> @@ 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)  { @@ -1579,8
> +1577,14 @@ static int mana_gd_probe(struct pci_dev *pdev, const struct
> pci_device_id *ent)
>  	if (err)
>  		goto cleanup_gd;
> 
> +	err = mana_rdma_probe(&gc->mana_ib);
> +	if (err)
> +		goto cleanup_mana;
> +
>  	return 0;
> 
> +cleanup_mana:
> +	mana_remove(&gc->mana, false);
>  cleanup_gd:
>  	mana_gd_cleanup(pdev);
>  unmap_bar:
> @@ -1608,6 +1612,7 @@ static void mana_gd_remove(struct pci_dev *pdev)  {
>  	struct gdma_context *gc = pci_get_drvdata(pdev);
> 
> +	mana_rdma_remove(&gc->mana_ib);
>  	mana_remove(&gc->mana, false);
> 
>  	mana_gd_cleanup(pdev);
> @@ -1631,6 +1636,7 @@ static int mana_gd_suspend(struct pci_dev *pdev,
> pm_message_t state)  {
>  	struct gdma_context *gc = pci_get_drvdata(pdev);
> 
> +	mana_rdma_remove(&gc->mana_ib);
>  	mana_remove(&gc->mana, true);
> 
>  	mana_gd_cleanup(pdev);
> @@ -1655,6 +1661,10 @@ static int mana_gd_resume(struct pci_dev *pdev)
>  	if (err)
>  		return err;
> 
> +	err = mana_rdma_probe(&gc->mana_ib);
> +	if (err)
> +		return err;
> +
>  	return 0;
>  }
> 
> @@ -1665,6 +1675,7 @@ static void mana_gd_shutdown(struct pci_dev *pdev)
> 
>  	dev_info(&pdev->dev, "Shutdown was called\n");
> 
> +	mana_rdma_remove(&gc->mana_ib);
>  	mana_remove(&gc->mana, true);
> 
>  	mana_gd_cleanup(pdev);
> diff --git a/drivers/net/ethernet/microsoft/mana/mana_en.c
> b/drivers/net/ethernet/microsoft/mana/mana_en.c
> index 5be0585..2013d0e 100644
> --- a/drivers/net/ethernet/microsoft/mana/mana_en.c
> +++ b/drivers/net/ethernet/microsoft/mana/mana_en.c
> @@ -2944,7 +2944,7 @@ static void remove_adev(struct gdma_dev *gd)
>  	gd->adev = NULL;
>  }
> 
> -static int add_adev(struct gdma_dev *gd)
> +static int add_adev(struct gdma_dev *gd, const char *name)
>  {
>  	struct auxiliary_device *adev;
>  	struct mana_adev *madev;
> @@ -2960,7 +2960,7 @@ static int add_adev(struct gdma_dev *gd)
>  		goto idx_fail;
>  	adev->id = ret;
> 
> -	adev->name = "rdma";
> +	adev->name = name;
>  	adev->dev.parent = gd->gdma_context->dev;
>  	adev->dev.release = adev_release;
>  	madev->mdev = gd;
> @@ -3076,7 +3076,7 @@ int mana_probe(struct gdma_dev *gd, bool resuming)
>  		}
>  	}
> 
> -	err = add_adev(gd);
> +	err = add_adev(gd, "eth");
>  out:
>  	if (err) {
>  		mana_remove(gd, false);
> @@ -3150,6 +3150,39 @@ out:
>  	dev_dbg(dev, "%s succeeded\n", __func__);  }
> 
> +int mana_rdma_probe(struct gdma_dev *gd) {
> +	int err = 0;
> +
> +	if (gd->dev_id.type != GDMA_DEVICE_MANA_IB) {
> +		/* RDMA device is not detected on pci */
> +		return err;
> +	}
> +
> +	err = mana_gd_register_device(gd);
> +	if (err)
> +		return err;
> +
> +	err = add_adev(gd, "rdma");
> +	if (err)
> +		mana_gd_deregister_device(gd);
> +
> +	return err;
> +}
> +
> +void mana_rdma_remove(struct gdma_dev *gd) {
> +	if (gd->dev_id.type != GDMA_DEVICE_MANA_IB) {
> +		/* RDMA device is not detected on pci */
> +		return;
> +	}
> +
> +	if (gd->adev)
> +		remove_adev(gd);
> +
> +	mana_gd_deregister_device(gd);
> +}
> +
>  struct net_device *mana_get_primary_netdev(struct mana_context *ac,
>  					   u32 port_index,
>  					   netdevice_tracker *tracker)
> diff --git a/include/net/mana/mana.h b/include/net/mana/mana.h index
> 0f78065..5857efc 100644
> --- a/include/net/mana/mana.h
> +++ b/include/net/mana/mana.h
> @@ -488,6 +488,9 @@ int mana_detach(struct net_device *ndev, bool
> from_close);  int mana_probe(struct gdma_dev *gd, bool resuming);  void
> mana_remove(struct gdma_dev *gd, bool suspending);
> 
> +int mana_rdma_probe(struct gdma_dev *gd); void mana_rdma_remove(struct
> +gdma_dev *gd);
> +
>  void mana_xdp_tx(struct sk_buff *skb, struct net_device *ndev);  int
> mana_xdp_xmit(struct net_device *ndev, int n, struct xdp_frame **frames,
>  		  u32 flags);
> --
> 2.43.0

[PATCH rdma-next v4 2/4] RDMA/mana_ib: Add support of mana_ib for RNIC and ETH nic

[Konstantin Taranov]

From: Konstantin Taranov <kotaranov@microsoft.com>

Allow mana_ib to be created over ethernet gdma device and
over rnic gdma device. The HW has two devices with different
capabilities and different use-cases. Initialize required
resources depending on the used gdma device.

Signed-off-by: Konstantin Taranov <kotaranov@microsoft.com>
---
 drivers/infiniband/hw/mana/device.c  | 174 +++++++++++++--------------
 drivers/infiniband/hw/mana/main.c    |  55 ++++++++-
 drivers/infiniband/hw/mana/mana_ib.h |   6 +
 3 files changed, 138 insertions(+), 97 deletions(-)

diff --git a/drivers/infiniband/hw/mana/device.c b/drivers/infiniband/hw/mana/device.c
index b310893..165c0a1 100644
--- a/drivers/infiniband/hw/mana/device.c
+++ b/drivers/infiniband/hw/mana/device.c
@@ -101,103 +101,95 @@ static int mana_ib_probe(struct auxiliary_device *adev,
 			 const struct auxiliary_device_id *id)
 {
 	struct mana_adev *madev = container_of(adev, struct mana_adev, adev);
+	struct gdma_context *gc = madev->mdev->gdma_context;
+	struct mana_context *mc = gc->mana.driver_data;
 	struct gdma_dev *mdev = madev->mdev;
 	struct net_device *ndev;
-	struct mana_context *mc;
 	struct mana_ib_dev *dev;
 	u8 mac_addr[ETH_ALEN];
 	int ret;
 
-	mc = mdev->driver_data;
-
 	dev = ib_alloc_device(mana_ib_dev, ib_dev);
 	if (!dev)
 		return -ENOMEM;
 
 	ib_set_device_ops(&dev->ib_dev, &mana_ib_dev_ops);
-
-	dev->ib_dev.phys_port_cnt = mc->num_ports;
-
-	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->ib_dev.node_type = RDMA_NODE_IB_CA;
-
-	/*
-	 * num_comp_vectors needs to set to the max MSIX index
-	 * when interrupts and event queues are implemented
-	 */
-	dev->ib_dev.num_comp_vectors = mdev->gdma_context->max_num_queues;
-	dev->ib_dev.dev.parent = mdev->gdma_context->dev;
-
-	ndev = mana_get_primary_netdev(mc, 0, &dev->dev_tracker);
-	if (!ndev) {
-		ret = -ENODEV;
-		ibdev_err(&dev->ib_dev, "Failed to get netdev for IB port 1");
-		goto free_ib_device;
-	}
-	ether_addr_copy(mac_addr, ndev->dev_addr);
-	addrconf_addr_eui48((u8 *)&dev->ib_dev.node_guid, ndev->dev_addr);
-	ret = ib_device_set_netdev(&dev->ib_dev, ndev, 1);
-	/* mana_get_primary_netdev() returns ndev with refcount held */
-	netdev_put(ndev, &dev->dev_tracker);
-	if (ret) {
-		ibdev_err(&dev->ib_dev, "Failed to set ib netdev, ret %d", ret);
-		goto free_ib_device;
-	}
-
-	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;
-
-	dev->nb.notifier_call = mana_ib_netdev_event;
-	ret = register_netdevice_notifier(&dev->nb);
-	if (ret) {
-		ibdev_err(&dev->ib_dev, "Failed to register net notifier, %d",
-			  ret);
-		goto deregister_device;
-	}
-
-	ret = mana_ib_gd_query_adapter_caps(dev);
-	if (ret) {
-		ibdev_err(&dev->ib_dev, "Failed to query device caps, ret %d",
-			  ret);
-		goto deregister_net_notifier;
-	}
-
-	ib_set_device_ops(&dev->ib_dev, &mana_ib_stats_ops);
-
-	ret = mana_ib_create_eqs(dev);
-	if (ret) {
-		ibdev_err(&dev->ib_dev, "Failed to create EQs, ret %d", ret);
-		goto deregister_net_notifier;
-	}
-
-	ret = mana_ib_gd_create_rnic_adapter(dev);
-	if (ret)
-		goto destroy_eqs;
-
+	dev->ib_dev.num_comp_vectors = gc->max_num_queues;
+	dev->ib_dev.dev.parent = gc->dev;
+	dev->gdma_dev = mdev;
 	xa_init_flags(&dev->qp_table_wq, XA_FLAGS_LOCK_IRQ);
-	ret = mana_ib_gd_config_mac(dev, ADDR_OP_ADD, mac_addr);
-	if (ret) {
-		ibdev_err(&dev->ib_dev, "Failed to add Mac address, ret %d",
-			  ret);
-		goto destroy_rnic;
+
+	if (mana_ib_is_rnic(dev)) {
+		dev->ib_dev.phys_port_cnt = 1;
+		ndev = mana_get_primary_netdev(mc, 0, &dev->dev_tracker);
+		if (!ndev) {
+			ret = -ENODEV;
+			ibdev_err(&dev->ib_dev, "Failed to get netdev for IB port 1");
+			goto free_ib_device;
+		}
+		ether_addr_copy(mac_addr, ndev->dev_addr);
+		addrconf_addr_eui48((u8 *)&dev->ib_dev.node_guid, ndev->dev_addr);
+		ret = ib_device_set_netdev(&dev->ib_dev, ndev, 1);
+		/* mana_get_primary_netdev() returns ndev with refcount held */
+		netdev_put(ndev, &dev->dev_tracker);
+		if (ret) {
+			ibdev_err(&dev->ib_dev, "Failed to set ib netdev, ret %d", ret);
+			goto free_ib_device;
+		}
+
+		dev->nb.notifier_call = mana_ib_netdev_event;
+		ret = register_netdevice_notifier(&dev->nb);
+		if (ret) {
+			ibdev_err(&dev->ib_dev, "Failed to register net notifier, %d",
+				  ret);
+			goto free_ib_device;
+		}
+
+		ret = mana_ib_gd_query_adapter_caps(dev);
+		if (ret) {
+			ibdev_err(&dev->ib_dev, "Failed to query device caps, ret %d", ret);
+			goto deregister_net_notifier;
+		}
+
+		ib_set_device_ops(&dev->ib_dev, &mana_ib_stats_ops);
+
+		ret = mana_ib_create_eqs(dev);
+		if (ret) {
+			ibdev_err(&dev->ib_dev, "Failed to create EQs, ret %d", ret);
+			goto deregister_net_notifier;
+		}
+
+		ret = mana_ib_gd_create_rnic_adapter(dev);
+		if (ret)
+			goto destroy_eqs;
+
+		ret = mana_ib_gd_config_mac(dev, ADDR_OP_ADD, mac_addr);
+		if (ret) {
+			ibdev_err(&dev->ib_dev, "Failed to add Mac address, ret %d", ret);
+			goto destroy_rnic;
+		}
+	} else {
+		dev->ib_dev.phys_port_cnt = mc->num_ports;
+		ret = mana_eth_query_adapter_caps(dev);
+		if (ret) {
+			ibdev_err(&dev->ib_dev, "Failed to query ETH device caps, ret %d", ret);
+			goto free_ib_device;
+		}
 	}
 
-	dev->av_pool = dma_pool_create("mana_ib_av", mdev->gdma_context->dev,
-				       MANA_AV_BUFFER_SIZE, MANA_AV_BUFFER_SIZE, 0);
+	dev->av_pool = dma_pool_create("mana_ib_av", gc->dev, MANA_AV_BUFFER_SIZE,
+				       MANA_AV_BUFFER_SIZE, 0);
 	if (!dev->av_pool) {
 		ret = -ENOMEM;
 		goto destroy_rnic;
 	}
 
-	ret = ib_register_device(&dev->ib_dev, "mana_%d",
-				 mdev->gdma_context->dev);
+	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);
+
+	ret = ib_register_device(&dev->ib_dev, mana_ib_is_rnic(dev) ? "mana_%d" : "manae_%d",
+				 gc->dev);
 	if (ret)
 		goto deallocate_pool;
 
@@ -208,15 +200,16 @@ static int mana_ib_probe(struct auxiliary_device *adev,
 deallocate_pool:
 	dma_pool_destroy(dev->av_pool);
 destroy_rnic:
-	xa_destroy(&dev->qp_table_wq);
-	mana_ib_gd_destroy_rnic_adapter(dev);
+	if (mana_ib_is_rnic(dev))
+		mana_ib_gd_destroy_rnic_adapter(dev);
 destroy_eqs:
-	mana_ib_destroy_eqs(dev);
+	if (mana_ib_is_rnic(dev))
+		mana_ib_destroy_eqs(dev);
 deregister_net_notifier:
-	unregister_netdevice_notifier(&dev->nb);
-deregister_device:
-	mana_gd_deregister_device(dev->gdma_dev);
+	if (mana_ib_is_rnic(dev))
+		unregister_netdevice_notifier(&dev->nb);
 free_ib_device:
+	xa_destroy(&dev->qp_table_wq);
 	ib_dealloc_device(&dev->ib_dev);
 	return ret;
 }
@@ -227,25 +220,24 @@ static void mana_ib_remove(struct auxiliary_device *adev)
 
 	ib_unregister_device(&dev->ib_dev);
 	dma_pool_destroy(dev->av_pool);
+	if (mana_ib_is_rnic(dev)) {
+		mana_ib_gd_destroy_rnic_adapter(dev);
+		mana_ib_destroy_eqs(dev);
+		unregister_netdevice_notifier(&dev->nb);
+	}
 	xa_destroy(&dev->qp_table_wq);
-	mana_ib_gd_destroy_rnic_adapter(dev);
-	mana_ib_destroy_eqs(dev);
-	unregister_netdevice_notifier(&dev->nb);
-	mana_gd_deregister_device(dev->gdma_dev);
 	ib_dealloc_device(&dev->ib_dev);
 }
 
 static const struct auxiliary_device_id mana_id_table[] = {
-	{
-		.name = "mana.rdma",
-	},
+	{ .name = "mana.rdma", },
+	{ .name = "mana.eth", },
 	{},
 };
 
 MODULE_DEVICE_TABLE(auxiliary, mana_id_table);
 
 static struct auxiliary_driver mana_driver = {
-	.name = "rdma",
 	.probe = mana_ib_probe,
 	.remove = mana_ib_remove,
 	.id_table = mana_id_table,
diff --git a/drivers/infiniband/hw/mana/main.c b/drivers/infiniband/hw/mana/main.c
index bb0f685..3837e30 100644
--- a/drivers/infiniband/hw/mana/main.c
+++ b/drivers/infiniband/hw/mana/main.c
@@ -4,6 +4,7 @@
  */
 
 #include "mana_ib.h"
+#include "linux/pci.h"
 
 void mana_ib_uncfg_vport(struct mana_ib_dev *dev, struct mana_ib_pd *pd,
 			 u32 port)
@@ -551,6 +552,7 @@ int mana_ib_mmap(struct ib_ucontext *ibcontext, struct vm_area_struct *vma)
 int mana_ib_get_port_immutable(struct ib_device *ibdev, u32 port_num,
 			       struct ib_port_immutable *immutable)
 {
+	struct mana_ib_dev *dev = container_of(ibdev, struct mana_ib_dev, ib_dev);
 	struct ib_port_attr attr;
 	int err;
 
@@ -560,10 +562,12 @@ int mana_ib_get_port_immutable(struct ib_device *ibdev, u32 port_num,
 
 	immutable->pkey_tbl_len = attr.pkey_tbl_len;
 	immutable->gid_tbl_len = attr.gid_tbl_len;
-	immutable->core_cap_flags = RDMA_CORE_PORT_RAW_PACKET;
-	if (port_num == 1) {
-		immutable->core_cap_flags |= RDMA_CORE_PORT_IBA_ROCE_UDP_ENCAP;
+
+	if (mana_ib_is_rnic(dev)) {
+		immutable->core_cap_flags = RDMA_CORE_PORT_IBA_ROCE_UDP_ENCAP;
 		immutable->max_mad_size = IB_MGMT_MAD_SIZE;
+	} else {
+		immutable->core_cap_flags = RDMA_CORE_PORT_RAW_PACKET;
 	}
 
 	return 0;
@@ -572,10 +576,12 @@ 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)
 {
-	struct mana_ib_dev *dev = container_of(ibdev,
-			struct mana_ib_dev, ib_dev);
+	struct mana_ib_dev *dev = container_of(ibdev, struct mana_ib_dev, ib_dev);
+	struct pci_dev *pdev = to_pci_dev(mdev_to_gc(dev)->dev);
 
 	memset(props, 0, sizeof(*props));
+	props->vendor_id = pdev->vendor;
+	props->vendor_part_id = dev->gdma_dev->dev_id.type;
 	props->max_mr_size = MANA_IB_MAX_MR_SIZE;
 	props->page_size_cap = dev->adapter_caps.page_size_cap;
 	props->max_qp = dev->adapter_caps.max_qp_count;
@@ -596,6 +602,8 @@ int mana_ib_query_device(struct ib_device *ibdev, struct ib_device_attr *props,
 	props->max_ah = INT_MAX;
 	props->max_pkeys = 1;
 	props->local_ca_ack_delay = MANA_CA_ACK_DELAY;
+	if (!mana_ib_is_rnic(dev))
+		props->raw_packet_caps = IB_RAW_PACKET_CAP_IP_CSUM;
 
 	return 0;
 }
@@ -603,6 +611,7 @@ int mana_ib_query_device(struct ib_device *ibdev, struct ib_device_attr *props,
 int mana_ib_query_port(struct ib_device *ibdev, u32 port,
 		       struct ib_port_attr *props)
 {
+	struct mana_ib_dev *dev = container_of(ibdev, struct mana_ib_dev, ib_dev);
 	struct net_device *ndev = mana_ib_get_netdev(ibdev, port);
 
 	if (!ndev)
@@ -623,7 +632,7 @@ int mana_ib_query_port(struct ib_device *ibdev, u32 port,
 	props->active_width = IB_WIDTH_4X;
 	props->active_speed = IB_SPEED_EDR;
 	props->pkey_tbl_len = 1;
-	if (port == 1) {
+	if (mana_ib_is_rnic(dev)) {
 		props->gid_tbl_len = 16;
 		props->port_cap_flags = IB_PORT_CM_SUP;
 		props->ip_gids = true;
@@ -703,6 +712,37 @@ int mana_ib_gd_query_adapter_caps(struct mana_ib_dev *dev)
 	return 0;
 }
 
+int mana_eth_query_adapter_caps(struct mana_ib_dev *dev)
+{
+	struct mana_ib_adapter_caps *caps = &dev->adapter_caps;
+	struct gdma_query_max_resources_resp resp = {};
+	struct gdma_general_req req = {};
+	int err;
+
+	mana_gd_init_req_hdr(&req.hdr, GDMA_QUERY_MAX_RESOURCES,
+			     sizeof(req), sizeof(resp));
+
+	err = mana_gd_send_request(mdev_to_gc(dev), 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_qp_count = min_t(u32, resp.max_sq, resp.max_rq);
+	caps->max_cq_count = resp.max_cq;
+	caps->max_mr_count = resp.max_mst;
+	caps->max_pd_count = 0x6000;
+	caps->max_qp_wr = min_t(u32,
+				0x100000 / GDMA_MAX_SQE_SIZE,
+				0x100000 / GDMA_MAX_RQE_SIZE);
+	caps->max_send_sge_count = 30;
+	caps->max_recv_sge_count = 15;
+	caps->page_size_cap = PAGE_SZ_BM;
+
+	return 0;
+}
+
 static void
 mana_ib_event_handler(void *ctx, struct gdma_queue *q, struct gdma_event *event)
 {
@@ -921,6 +961,9 @@ int mana_ib_gd_create_cq(struct mana_ib_dev *mdev, struct mana_ib_cq *cq, u32 do
 	struct mana_rnic_create_cq_req req = {};
 	int err;
 
+	if (!mdev->eqs)
+		return -EINVAL;
+
 	mana_gd_init_req_hdr(&req.hdr, MANA_IB_CREATE_CQ, sizeof(req), sizeof(resp));
 	req.hdr.dev_id = gc->mana_ib.dev_id;
 	req.adapter = mdev->adapter_handle;
diff --git a/drivers/infiniband/hw/mana/mana_ib.h b/drivers/infiniband/hw/mana/mana_ib.h
index f0dbd90..42bebd6 100644
--- a/drivers/infiniband/hw/mana/mana_ib.h
+++ b/drivers/infiniband/hw/mana/mana_ib.h
@@ -544,6 +544,11 @@ static inline void mana_put_qp_ref(struct mana_ib_qp *qp)
 		complete(&qp->free);
 }
 
+static inline bool mana_ib_is_rnic(struct mana_ib_dev *mdev)
+{
+	return mdev->gdma_dev->dev_id.type == GDMA_DEVICE_MANA_IB;
+}
+
 static inline struct net_device *mana_ib_get_netdev(struct ib_device *ibdev, u32 port)
 {
 	struct mana_ib_dev *mdev = container_of(ibdev, struct mana_ib_dev, ib_dev);
@@ -643,6 +648,7 @@ 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_gd_query_adapter_caps(struct mana_ib_dev *mdev);
+int mana_eth_query_adapter_caps(struct mana_ib_dev *mdev);
 
 int mana_ib_create_eqs(struct mana_ib_dev *mdev);
 
-- 
2.43.0

RE: [PATCH rdma-next v4 2/4] RDMA/mana_ib: Add support of mana_ib for RNIC and ETH nic

[Long Li]

> -----Original Message-----
> From: Konstantin Taranov <kotaranov@linux.microsoft.com>
> Sent: Wednesday, May 7, 2025 8:59 AM
> To: Konstantin Taranov <kotaranov@microsoft.com>; pabeni@redhat.com;
> Haiyang Zhang <haiyangz@microsoft.com>; KY Srinivasan <kys@microsoft.com>;
> edumazet@google.com; kuba@kernel.org; davem@davemloft.net; Dexuan Cui
> <decui@microsoft.com>; wei.liu@kernel.org; Long Li <longli@microsoft.com>;
> jgg@ziepe.ca; leon@kernel.org
> Cc: linux-rdma@vger.kernel.org; linux-kernel@vger.kernel.org;
> netdev@vger.kernel.org
> Subject: [PATCH rdma-next v4 2/4] RDMA/mana_ib: Add support of mana_ib for
> RNIC and ETH nic
> 
> From: Konstantin Taranov <kotaranov@microsoft.com>
> 
> Allow mana_ib to be created over ethernet gdma device and over rnic gdma
> device. The HW has two devices with different capabilities and different use-
> cases. Initialize required resources depending on the used gdma device.
> 
> Signed-off-by: Konstantin Taranov <kotaranov@microsoft.com>

Reviewed-by: Long Li <longli@microsoft.com>

> ---
>  drivers/infiniband/hw/mana/device.c  | 174 +++++++++++++--------------
>  drivers/infiniband/hw/mana/main.c    |  55 ++++++++-
>  drivers/infiniband/hw/mana/mana_ib.h |   6 +
>  3 files changed, 138 insertions(+), 97 deletions(-)
> 
> diff --git a/drivers/infiniband/hw/mana/device.c
> b/drivers/infiniband/hw/mana/device.c
> index b310893..165c0a1 100644
> --- a/drivers/infiniband/hw/mana/device.c
> +++ b/drivers/infiniband/hw/mana/device.c
> @@ -101,103 +101,95 @@ static int mana_ib_probe(struct auxiliary_device
> *adev,
>  			 const struct auxiliary_device_id *id)  {
>  	struct mana_adev *madev = container_of(adev, struct mana_adev,
> adev);
> +	struct gdma_context *gc = madev->mdev->gdma_context;
> +	struct mana_context *mc = gc->mana.driver_data;
>  	struct gdma_dev *mdev = madev->mdev;
>  	struct net_device *ndev;
> -	struct mana_context *mc;
>  	struct mana_ib_dev *dev;
>  	u8 mac_addr[ETH_ALEN];
>  	int ret;
> 
> -	mc = mdev->driver_data;
> -
>  	dev = ib_alloc_device(mana_ib_dev, ib_dev);
>  	if (!dev)
>  		return -ENOMEM;
> 
>  	ib_set_device_ops(&dev->ib_dev, &mana_ib_dev_ops);
> -
> -	dev->ib_dev.phys_port_cnt = mc->num_ports;
> -
> -	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->ib_dev.node_type = RDMA_NODE_IB_CA;
> -
> -	/*
> -	 * num_comp_vectors needs to set to the max MSIX index
> -	 * when interrupts and event queues are implemented
> -	 */
> -	dev->ib_dev.num_comp_vectors = mdev->gdma_context-
> >max_num_queues;
> -	dev->ib_dev.dev.parent = mdev->gdma_context->dev;
> -
> -	ndev = mana_get_primary_netdev(mc, 0, &dev->dev_tracker);
> -	if (!ndev) {
> -		ret = -ENODEV;
> -		ibdev_err(&dev->ib_dev, "Failed to get netdev for IB port 1");
> -		goto free_ib_device;
> -	}
> -	ether_addr_copy(mac_addr, ndev->dev_addr);
> -	addrconf_addr_eui48((u8 *)&dev->ib_dev.node_guid, ndev->dev_addr);
> -	ret = ib_device_set_netdev(&dev->ib_dev, ndev, 1);
> -	/* mana_get_primary_netdev() returns ndev with refcount held */
> -	netdev_put(ndev, &dev->dev_tracker);
> -	if (ret) {
> -		ibdev_err(&dev->ib_dev, "Failed to set ib netdev, ret %d", ret);
> -		goto free_ib_device;
> -	}
> -
> -	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;
> -
> -	dev->nb.notifier_call = mana_ib_netdev_event;
> -	ret = register_netdevice_notifier(&dev->nb);
> -	if (ret) {
> -		ibdev_err(&dev->ib_dev, "Failed to register net notifier, %d",
> -			  ret);
> -		goto deregister_device;
> -	}
> -
> -	ret = mana_ib_gd_query_adapter_caps(dev);
> -	if (ret) {
> -		ibdev_err(&dev->ib_dev, "Failed to query device caps, ret %d",
> -			  ret);
> -		goto deregister_net_notifier;
> -	}
> -
> -	ib_set_device_ops(&dev->ib_dev, &mana_ib_stats_ops);
> -
> -	ret = mana_ib_create_eqs(dev);
> -	if (ret) {
> -		ibdev_err(&dev->ib_dev, "Failed to create EQs, ret %d", ret);
> -		goto deregister_net_notifier;
> -	}
> -
> -	ret = mana_ib_gd_create_rnic_adapter(dev);
> -	if (ret)
> -		goto destroy_eqs;
> -
> +	dev->ib_dev.num_comp_vectors = gc->max_num_queues;
> +	dev->ib_dev.dev.parent = gc->dev;
> +	dev->gdma_dev = mdev;
>  	xa_init_flags(&dev->qp_table_wq, XA_FLAGS_LOCK_IRQ);
> -	ret = mana_ib_gd_config_mac(dev, ADDR_OP_ADD, mac_addr);
> -	if (ret) {
> -		ibdev_err(&dev->ib_dev, "Failed to add Mac address, ret %d",
> -			  ret);
> -		goto destroy_rnic;
> +
> +	if (mana_ib_is_rnic(dev)) {
> +		dev->ib_dev.phys_port_cnt = 1;
> +		ndev = mana_get_primary_netdev(mc, 0, &dev->dev_tracker);
> +		if (!ndev) {
> +			ret = -ENODEV;
> +			ibdev_err(&dev->ib_dev, "Failed to get netdev for IB
> port 1");
> +			goto free_ib_device;
> +		}
> +		ether_addr_copy(mac_addr, ndev->dev_addr);
> +		addrconf_addr_eui48((u8 *)&dev->ib_dev.node_guid, ndev-
> >dev_addr);
> +		ret = ib_device_set_netdev(&dev->ib_dev, ndev, 1);
> +		/* mana_get_primary_netdev() returns ndev with refcount held
> */
> +		netdev_put(ndev, &dev->dev_tracker);
> +		if (ret) {
> +			ibdev_err(&dev->ib_dev, "Failed to set ib netdev,
> ret %d", ret);
> +			goto free_ib_device;
> +		}
> +
> +		dev->nb.notifier_call = mana_ib_netdev_event;
> +		ret = register_netdevice_notifier(&dev->nb);
> +		if (ret) {
> +			ibdev_err(&dev->ib_dev, "Failed to register net
> notifier, %d",
> +				  ret);
> +			goto free_ib_device;
> +		}
> +
> +		ret = mana_ib_gd_query_adapter_caps(dev);
> +		if (ret) {
> +			ibdev_err(&dev->ib_dev, "Failed to query device caps,
> ret %d", ret);
> +			goto deregister_net_notifier;
> +		}
> +
> +		ib_set_device_ops(&dev->ib_dev, &mana_ib_stats_ops);
> +
> +		ret = mana_ib_create_eqs(dev);
> +		if (ret) {
> +			ibdev_err(&dev->ib_dev, "Failed to create EQs, ret %d",
> ret);
> +			goto deregister_net_notifier;
> +		}
> +
> +		ret = mana_ib_gd_create_rnic_adapter(dev);
> +		if (ret)
> +			goto destroy_eqs;
> +
> +		ret = mana_ib_gd_config_mac(dev, ADDR_OP_ADD, mac_addr);
> +		if (ret) {
> +			ibdev_err(&dev->ib_dev, "Failed to add Mac address,
> ret %d", ret);
> +			goto destroy_rnic;
> +		}
> +	} else {
> +		dev->ib_dev.phys_port_cnt = mc->num_ports;
> +		ret = mana_eth_query_adapter_caps(dev);
> +		if (ret) {
> +			ibdev_err(&dev->ib_dev, "Failed to query ETH device
> caps, ret %d", ret);
> +			goto free_ib_device;
> +		}
>  	}
> 
> -	dev->av_pool = dma_pool_create("mana_ib_av", mdev->gdma_context-
> >dev,
> -				       MANA_AV_BUFFER_SIZE,
> MANA_AV_BUFFER_SIZE, 0);
> +	dev->av_pool = dma_pool_create("mana_ib_av", gc->dev,
> MANA_AV_BUFFER_SIZE,
> +				       MANA_AV_BUFFER_SIZE, 0);
>  	if (!dev->av_pool) {
>  		ret = -ENOMEM;
>  		goto destroy_rnic;
>  	}
> 
> -	ret = ib_register_device(&dev->ib_dev, "mana_%d",
> -				 mdev->gdma_context->dev);
> +	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);
> +
> +	ret = ib_register_device(&dev->ib_dev, mana_ib_is_rnic(dev) ?
> "mana_%d" : "manae_%d",
> +				 gc->dev);
>  	if (ret)
>  		goto deallocate_pool;
> 
> @@ -208,15 +200,16 @@ static int mana_ib_probe(struct auxiliary_device
> *adev,
>  deallocate_pool:
>  	dma_pool_destroy(dev->av_pool);
>  destroy_rnic:
> -	xa_destroy(&dev->qp_table_wq);
> -	mana_ib_gd_destroy_rnic_adapter(dev);
> +	if (mana_ib_is_rnic(dev))
> +		mana_ib_gd_destroy_rnic_adapter(dev);
>  destroy_eqs:
> -	mana_ib_destroy_eqs(dev);
> +	if (mana_ib_is_rnic(dev))
> +		mana_ib_destroy_eqs(dev);
>  deregister_net_notifier:
> -	unregister_netdevice_notifier(&dev->nb);
> -deregister_device:
> -	mana_gd_deregister_device(dev->gdma_dev);
> +	if (mana_ib_is_rnic(dev))
> +		unregister_netdevice_notifier(&dev->nb);
>  free_ib_device:
> +	xa_destroy(&dev->qp_table_wq);
>  	ib_dealloc_device(&dev->ib_dev);
>  	return ret;
>  }
> @@ -227,25 +220,24 @@ static void mana_ib_remove(struct auxiliary_device
> *adev)
> 
>  	ib_unregister_device(&dev->ib_dev);
>  	dma_pool_destroy(dev->av_pool);
> +	if (mana_ib_is_rnic(dev)) {
> +		mana_ib_gd_destroy_rnic_adapter(dev);
> +		mana_ib_destroy_eqs(dev);
> +		unregister_netdevice_notifier(&dev->nb);
> +	}
>  	xa_destroy(&dev->qp_table_wq);
> -	mana_ib_gd_destroy_rnic_adapter(dev);
> -	mana_ib_destroy_eqs(dev);
> -	unregister_netdevice_notifier(&dev->nb);
> -	mana_gd_deregister_device(dev->gdma_dev);
>  	ib_dealloc_device(&dev->ib_dev);
>  }
> 
>  static const struct auxiliary_device_id mana_id_table[] = {
> -	{
> -		.name = "mana.rdma",
> -	},
> +	{ .name = "mana.rdma", },
> +	{ .name = "mana.eth", },
>  	{},
>  };
> 
>  MODULE_DEVICE_TABLE(auxiliary, mana_id_table);
> 
>  static struct auxiliary_driver mana_driver = {
> -	.name = "rdma",
>  	.probe = mana_ib_probe,
>  	.remove = mana_ib_remove,
>  	.id_table = mana_id_table,
> diff --git a/drivers/infiniband/hw/mana/main.c
> b/drivers/infiniband/hw/mana/main.c
> index bb0f685..3837e30 100644
> --- a/drivers/infiniband/hw/mana/main.c
> +++ b/drivers/infiniband/hw/mana/main.c
> @@ -4,6 +4,7 @@
>   */
> 
>  #include "mana_ib.h"
> +#include "linux/pci.h"
> 
>  void mana_ib_uncfg_vport(struct mana_ib_dev *dev, struct mana_ib_pd *pd,
>  			 u32 port)
> @@ -551,6 +552,7 @@ int mana_ib_mmap(struct ib_ucontext *ibcontext, struct
> vm_area_struct *vma)  int mana_ib_get_port_immutable(struct ib_device
> *ibdev, u32 port_num,
>  			       struct ib_port_immutable *immutable)  {
> +	struct mana_ib_dev *dev = container_of(ibdev, struct mana_ib_dev,
> +ib_dev);
>  	struct ib_port_attr attr;
>  	int err;
> 
> @@ -560,10 +562,12 @@ int mana_ib_get_port_immutable(struct ib_device
> *ibdev, u32 port_num,
> 
>  	immutable->pkey_tbl_len = attr.pkey_tbl_len;
>  	immutable->gid_tbl_len = attr.gid_tbl_len;
> -	immutable->core_cap_flags = RDMA_CORE_PORT_RAW_PACKET;
> -	if (port_num == 1) {
> -		immutable->core_cap_flags |=
> RDMA_CORE_PORT_IBA_ROCE_UDP_ENCAP;
> +
> +	if (mana_ib_is_rnic(dev)) {
> +		immutable->core_cap_flags =
> RDMA_CORE_PORT_IBA_ROCE_UDP_ENCAP;
>  		immutable->max_mad_size = IB_MGMT_MAD_SIZE;
> +	} else {
> +		immutable->core_cap_flags =
> RDMA_CORE_PORT_RAW_PACKET;
>  	}
> 
>  	return 0;
> @@ -572,10 +576,12 @@ 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)
>  {
> -	struct mana_ib_dev *dev = container_of(ibdev,
> -			struct mana_ib_dev, ib_dev);
> +	struct mana_ib_dev *dev = container_of(ibdev, struct mana_ib_dev,
> ib_dev);
> +	struct pci_dev *pdev = to_pci_dev(mdev_to_gc(dev)->dev);
> 
>  	memset(props, 0, sizeof(*props));
> +	props->vendor_id = pdev->vendor;
> +	props->vendor_part_id = dev->gdma_dev->dev_id.type;
>  	props->max_mr_size = MANA_IB_MAX_MR_SIZE;
>  	props->page_size_cap = dev->adapter_caps.page_size_cap;
>  	props->max_qp = dev->adapter_caps.max_qp_count; @@ -596,6 +602,8
> @@ int mana_ib_query_device(struct ib_device *ibdev, struct ib_device_attr
> *props,
>  	props->max_ah = INT_MAX;
>  	props->max_pkeys = 1;
>  	props->local_ca_ack_delay = MANA_CA_ACK_DELAY;
> +	if (!mana_ib_is_rnic(dev))
> +		props->raw_packet_caps = IB_RAW_PACKET_CAP_IP_CSUM;
> 
>  	return 0;
>  }
> @@ -603,6 +611,7 @@ int mana_ib_query_device(struct ib_device *ibdev, struct
> ib_device_attr *props,  int mana_ib_query_port(struct ib_device *ibdev, u32 port,
>  		       struct ib_port_attr *props)
>  {
> +	struct mana_ib_dev *dev = container_of(ibdev, struct mana_ib_dev,
> +ib_dev);
>  	struct net_device *ndev = mana_ib_get_netdev(ibdev, port);
> 
>  	if (!ndev)
> @@ -623,7 +632,7 @@ int mana_ib_query_port(struct ib_device *ibdev, u32
> port,
>  	props->active_width = IB_WIDTH_4X;
>  	props->active_speed = IB_SPEED_EDR;
>  	props->pkey_tbl_len = 1;
> -	if (port == 1) {
> +	if (mana_ib_is_rnic(dev)) {
>  		props->gid_tbl_len = 16;
>  		props->port_cap_flags = IB_PORT_CM_SUP;
>  		props->ip_gids = true;
> @@ -703,6 +712,37 @@ int mana_ib_gd_query_adapter_caps(struct
> mana_ib_dev *dev)
>  	return 0;
>  }
> 
> +int mana_eth_query_adapter_caps(struct mana_ib_dev *dev) {
> +	struct mana_ib_adapter_caps *caps = &dev->adapter_caps;
> +	struct gdma_query_max_resources_resp resp = {};
> +	struct gdma_general_req req = {};
> +	int err;
> +
> +	mana_gd_init_req_hdr(&req.hdr, GDMA_QUERY_MAX_RESOURCES,
> +			     sizeof(req), sizeof(resp));
> +
> +	err = mana_gd_send_request(mdev_to_gc(dev), 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_qp_count = min_t(u32, resp.max_sq, resp.max_rq);
> +	caps->max_cq_count = resp.max_cq;
> +	caps->max_mr_count = resp.max_mst;
> +	caps->max_pd_count = 0x6000;
> +	caps->max_qp_wr = min_t(u32,
> +				0x100000 / GDMA_MAX_SQE_SIZE,
> +				0x100000 / GDMA_MAX_RQE_SIZE);
> +	caps->max_send_sge_count = 30;
> +	caps->max_recv_sge_count = 15;
> +	caps->page_size_cap = PAGE_SZ_BM;
> +
> +	return 0;
> +}
> +
>  static void
>  mana_ib_event_handler(void *ctx, struct gdma_queue *q, struct gdma_event
> *event)  { @@ -921,6 +961,9 @@ int mana_ib_gd_create_cq(struct
> mana_ib_dev *mdev, struct mana_ib_cq *cq, u32 do
>  	struct mana_rnic_create_cq_req req = {};
>  	int err;
> 
> +	if (!mdev->eqs)
> +		return -EINVAL;
> +
>  	mana_gd_init_req_hdr(&req.hdr, MANA_IB_CREATE_CQ, sizeof(req),
> sizeof(resp));
>  	req.hdr.dev_id = gc->mana_ib.dev_id;
>  	req.adapter = mdev->adapter_handle;
> diff --git a/drivers/infiniband/hw/mana/mana_ib.h
> b/drivers/infiniband/hw/mana/mana_ib.h
> index f0dbd90..42bebd6 100644
> --- a/drivers/infiniband/hw/mana/mana_ib.h
> +++ b/drivers/infiniband/hw/mana/mana_ib.h
> @@ -544,6 +544,11 @@ static inline void mana_put_qp_ref(struct mana_ib_qp
> *qp)
>  		complete(&qp->free);
>  }
> 
> +static inline bool mana_ib_is_rnic(struct mana_ib_dev *mdev) {
> +	return mdev->gdma_dev->dev_id.type == GDMA_DEVICE_MANA_IB; }
> +
>  static inline struct net_device *mana_ib_get_netdev(struct ib_device *ibdev,
> u32 port)  {
>  	struct mana_ib_dev *mdev = container_of(ibdev, struct mana_ib_dev,
> ib_dev); @@ -643,6 +648,7 @@ 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_gd_query_adapter_caps(struct mana_ib_dev *mdev);
> +int mana_eth_query_adapter_caps(struct mana_ib_dev *mdev);
> 
>  int mana_ib_create_eqs(struct mana_ib_dev *mdev);
> 
> --
> 2.43.0

[PATCH rdma-next v4 3/4] RDMA/mana_ib: unify mana_ib functions to support any gdma device

[Konstantin Taranov]

From: Konstantin Taranov <kotaranov@microsoft.com>

Use the installed gdma_device instead of hard-coded device
in requests to the HW.

Signed-off-by: Konstantin Taranov <kotaranov@microsoft.com>
---
 drivers/infiniband/hw/mana/cq.c   |  4 +---
 drivers/infiniband/hw/mana/main.c | 27 +++++++++++++--------------
 drivers/infiniband/hw/mana/qp.c   |  5 ++---
 3 files changed, 16 insertions(+), 20 deletions(-)

diff --git a/drivers/infiniband/hw/mana/cq.c b/drivers/infiniband/hw/mana/cq.c
index 0fc4e26..28e154b 100644
--- a/drivers/infiniband/hw/mana/cq.c
+++ b/drivers/infiniband/hw/mana/cq.c
@@ -15,14 +15,12 @@ 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;
 	bool is_rnic_cq;
 	u32 doorbell;
 	u32 buf_size;
 	int err;
 
 	mdev = container_of(ibdev, struct mana_ib_dev, ib_dev);
-	gc = mdev_to_gc(mdev);
 
 	cq->comp_vector = attr->comp_vector % ibdev->num_comp_vectors;
 	cq->cq_handle = INVALID_MANA_HANDLE;
@@ -65,7 +63,7 @@ int mana_ib_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
 			ibdev_dbg(ibdev, "Failed to create kernel queue for create cq, %d\n", err);
 			return err;
 		}
-		doorbell = gc->mana_ib.doorbell;
+		doorbell = mdev->gdma_dev->doorbell;
 	}
 
 	if (is_rnic_cq) {
diff --git a/drivers/infiniband/hw/mana/main.c b/drivers/infiniband/hw/mana/main.c
index 3837e30..41a24a1 100644
--- a/drivers/infiniband/hw/mana/main.c
+++ b/drivers/infiniband/hw/mana/main.c
@@ -244,7 +244,6 @@ void mana_ib_dealloc_ucontext(struct ib_ucontext *ibcontext)
 int mana_ib_create_kernel_queue(struct mana_ib_dev *mdev, u32 size, enum gdma_queue_type type,
 				struct mana_ib_queue *queue)
 {
-	struct gdma_context *gc = mdev_to_gc(mdev);
 	struct gdma_queue_spec spec = {};
 	int err;
 
@@ -253,7 +252,7 @@ int mana_ib_create_kernel_queue(struct mana_ib_dev *mdev, u32 size, enum gdma_qu
 	spec.type = type;
 	spec.monitor_avl_buf = false;
 	spec.queue_size = size;
-	err = mana_gd_create_mana_wq_cq(&gc->mana_ib, &spec, &queue->kmem);
+	err = mana_gd_create_mana_wq_cq(mdev->gdma_dev, &spec, &queue->kmem);
 	if (err)
 		return err;
 	/* take ownership into mana_ib from mana */
@@ -784,7 +783,7 @@ int mana_ib_create_eqs(struct mana_ib_dev *mdev)
 	spec.eq.log2_throttle_limit = LOG2_EQ_THROTTLE;
 	spec.eq.msix_index = 0;
 
-	err = mana_gd_create_mana_eq(&gc->mana_ib, &spec, &mdev->fatal_err_eq);
+	err = mana_gd_create_mana_eq(mdev->gdma_dev, &spec, &mdev->fatal_err_eq);
 	if (err)
 		return err;
 
@@ -835,7 +834,7 @@ int mana_ib_gd_create_rnic_adapter(struct mana_ib_dev *mdev)
 
 	mana_gd_init_req_hdr(&req.hdr, MANA_IB_CREATE_ADAPTER, sizeof(req), sizeof(resp));
 	req.hdr.req.msg_version = GDMA_MESSAGE_V2;
-	req.hdr.dev_id = gc->mana_ib.dev_id;
+	req.hdr.dev_id = mdev->gdma_dev->dev_id;
 	req.notify_eq_id = mdev->fatal_err_eq->id;
 
 	if (mdev->adapter_caps.feature_flags & MANA_IB_FEATURE_CLIENT_ERROR_CQE_SUPPORT)
@@ -860,7 +859,7 @@ int mana_ib_gd_destroy_rnic_adapter(struct mana_ib_dev *mdev)
 
 	gc = mdev_to_gc(mdev);
 	mana_gd_init_req_hdr(&req.hdr, MANA_IB_DESTROY_ADAPTER, sizeof(req), sizeof(resp));
-	req.hdr.dev_id = gc->mana_ib.dev_id;
+	req.hdr.dev_id = mdev->gdma_dev->dev_id;
 	req.adapter = mdev->adapter_handle;
 
 	err = mana_gd_send_request(gc, sizeof(req), &req, sizeof(resp), &resp);
@@ -887,7 +886,7 @@ int mana_ib_gd_add_gid(const struct ib_gid_attr *attr, void **context)
 	}
 
 	mana_gd_init_req_hdr(&req.hdr, MANA_IB_CONFIG_IP_ADDR, sizeof(req), sizeof(resp));
-	req.hdr.dev_id = gc->mana_ib.dev_id;
+	req.hdr.dev_id = mdev->gdma_dev->dev_id;
 	req.adapter = mdev->adapter_handle;
 	req.op = ADDR_OP_ADD;
 	req.sgid_type = (ntype == RDMA_NETWORK_IPV6) ? SGID_TYPE_IPV6 : SGID_TYPE_IPV4;
@@ -917,7 +916,7 @@ int mana_ib_gd_del_gid(const struct ib_gid_attr *attr, void **context)
 	}
 
 	mana_gd_init_req_hdr(&req.hdr, MANA_IB_CONFIG_IP_ADDR, sizeof(req), sizeof(resp));
-	req.hdr.dev_id = gc->mana_ib.dev_id;
+	req.hdr.dev_id = mdev->gdma_dev->dev_id;
 	req.adapter = mdev->adapter_handle;
 	req.op = ADDR_OP_REMOVE;
 	req.sgid_type = (ntype == RDMA_NETWORK_IPV6) ? SGID_TYPE_IPV6 : SGID_TYPE_IPV4;
@@ -940,7 +939,7 @@ int mana_ib_gd_config_mac(struct mana_ib_dev *mdev, enum mana_ib_addr_op op, u8
 	int err;
 
 	mana_gd_init_req_hdr(&req.hdr, MANA_IB_CONFIG_MAC_ADDR, sizeof(req), sizeof(resp));
-	req.hdr.dev_id = gc->mana_ib.dev_id;
+	req.hdr.dev_id = mdev->gdma_dev->dev_id;
 	req.adapter = mdev->adapter_handle;
 	req.op = op;
 	copy_in_reverse(req.mac_addr, mac, ETH_ALEN);
@@ -965,7 +964,7 @@ int mana_ib_gd_create_cq(struct mana_ib_dev *mdev, struct mana_ib_cq *cq, u32 do
 		return -EINVAL;
 
 	mana_gd_init_req_hdr(&req.hdr, MANA_IB_CREATE_CQ, sizeof(req), sizeof(resp));
-	req.hdr.dev_id = gc->mana_ib.dev_id;
+	req.hdr.dev_id = mdev->gdma_dev->dev_id;
 	req.adapter = mdev->adapter_handle;
 	req.gdma_region = cq->queue.gdma_region;
 	req.eq_id = mdev->eqs[cq->comp_vector]->id;
@@ -997,7 +996,7 @@ int mana_ib_gd_destroy_cq(struct mana_ib_dev *mdev, struct mana_ib_cq *cq)
 		return 0;
 
 	mana_gd_init_req_hdr(&req.hdr, MANA_IB_DESTROY_CQ, sizeof(req), sizeof(resp));
-	req.hdr.dev_id = gc->mana_ib.dev_id;
+	req.hdr.dev_id = mdev->gdma_dev->dev_id;
 	req.adapter = mdev->adapter_handle;
 	req.cq_handle = cq->cq_handle;
 
@@ -1023,7 +1022,7 @@ int mana_ib_gd_create_rc_qp(struct mana_ib_dev *mdev, struct mana_ib_qp *qp,
 	int err, i;
 
 	mana_gd_init_req_hdr(&req.hdr, MANA_IB_CREATE_RC_QP, sizeof(req), sizeof(resp));
-	req.hdr.dev_id = gc->mana_ib.dev_id;
+	req.hdr.dev_id = mdev->gdma_dev->dev_id;
 	req.adapter = mdev->adapter_handle;
 	req.pd_handle = pd->pd_handle;
 	req.send_cq_handle = send_cq->cq_handle;
@@ -1059,7 +1058,7 @@ int mana_ib_gd_destroy_rc_qp(struct mana_ib_dev *mdev, struct mana_ib_qp *qp)
 	int err;
 
 	mana_gd_init_req_hdr(&req.hdr, MANA_IB_DESTROY_RC_QP, sizeof(req), sizeof(resp));
-	req.hdr.dev_id = gc->mana_ib.dev_id;
+	req.hdr.dev_id = mdev->gdma_dev->dev_id;
 	req.adapter = mdev->adapter_handle;
 	req.rc_qp_handle = qp->qp_handle;
 	err = mana_gd_send_request(gc, sizeof(req), &req, sizeof(resp), &resp);
@@ -1082,7 +1081,7 @@ int mana_ib_gd_create_ud_qp(struct mana_ib_dev *mdev, struct mana_ib_qp *qp,
 	int err, i;
 
 	mana_gd_init_req_hdr(&req.hdr, MANA_IB_CREATE_UD_QP, sizeof(req), sizeof(resp));
-	req.hdr.dev_id = gc->mana_ib.dev_id;
+	req.hdr.dev_id = mdev->gdma_dev->dev_id;
 	req.adapter = mdev->adapter_handle;
 	req.pd_handle = pd->pd_handle;
 	req.send_cq_handle = send_cq->cq_handle;
@@ -1117,7 +1116,7 @@ int mana_ib_gd_destroy_ud_qp(struct mana_ib_dev *mdev, struct mana_ib_qp *qp)
 	int err;
 
 	mana_gd_init_req_hdr(&req.hdr, MANA_IB_DESTROY_UD_QP, sizeof(req), sizeof(resp));
-	req.hdr.dev_id = gc->mana_ib.dev_id;
+	req.hdr.dev_id = mdev->gdma_dev->dev_id;
 	req.adapter = mdev->adapter_handle;
 	req.qp_handle = qp->qp_handle;
 	err = mana_gd_send_request(gc, sizeof(req), &req, sizeof(resp), &resp);
diff --git a/drivers/infiniband/hw/mana/qp.c b/drivers/infiniband/hw/mana/qp.c
index c928af5..14fd7d6 100644
--- a/drivers/infiniband/hw/mana/qp.c
+++ b/drivers/infiniband/hw/mana/qp.c
@@ -635,7 +635,6 @@ static int mana_ib_create_ud_qp(struct ib_qp *ibqp, struct ib_pd *ibpd,
 {
 	struct mana_ib_dev *mdev = container_of(ibpd->device, struct mana_ib_dev, ib_dev);
 	struct mana_ib_qp *qp = container_of(ibqp, struct mana_ib_qp, ibqp);
-	struct gdma_context *gc = mdev_to_gc(mdev);
 	u32 doorbell, queue_size;
 	int i, err;
 
@@ -654,7 +653,7 @@ static int mana_ib_create_ud_qp(struct ib_qp *ibqp, struct ib_pd *ibpd,
 			goto destroy_queues;
 		}
 	}
-	doorbell = gc->mana_ib.doorbell;
+	doorbell = mdev->gdma_dev->doorbell;
 
 	err = create_shadow_queue(&qp->shadow_rq, attr->cap.max_recv_wr,
 				  sizeof(struct ud_rq_shadow_wqe));
@@ -736,7 +735,7 @@ static int mana_ib_gd_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 	int err;
 
 	mana_gd_init_req_hdr(&req.hdr, MANA_IB_SET_QP_STATE, sizeof(req), sizeof(resp));
-	req.hdr.dev_id = gc->mana_ib.dev_id;
+	req.hdr.dev_id = mdev->gdma_dev->dev_id;
 	req.adapter = mdev->adapter_handle;
 	req.qp_handle = qp->qp_handle;
 	req.qp_state = attr->qp_state;
-- 
2.43.0

RE: [PATCH rdma-next v4 3/4] RDMA/mana_ib: unify mana_ib functions to support any gdma device

[Long Li]

> -----Original Message-----
> From: Konstantin Taranov <kotaranov@linux.microsoft.com>
> Sent: Wednesday, May 7, 2025 8:59 AM
> To: Konstantin Taranov <kotaranov@microsoft.com>; pabeni@redhat.com;
> Haiyang Zhang <haiyangz@microsoft.com>; KY Srinivasan <kys@microsoft.com>;
> edumazet@google.com; kuba@kernel.org; davem@davemloft.net; Dexuan Cui
> <decui@microsoft.com>; wei.liu@kernel.org; Long Li <longli@microsoft.com>;
> jgg@ziepe.ca; leon@kernel.org
> Cc: linux-rdma@vger.kernel.org; linux-kernel@vger.kernel.org;
> netdev@vger.kernel.org
> Subject: [PATCH rdma-next v4 3/4] RDMA/mana_ib: unify mana_ib functions to
> support any gdma device
> 
> From: Konstantin Taranov <kotaranov@microsoft.com>
> 
> Use the installed gdma_device instead of hard-coded device in requests to the
> HW.
> 
> Signed-off-by: Konstantin Taranov <kotaranov@microsoft.com>

Reviewed-by: Long Li <longli@microsoft.com>

> ---
>  drivers/infiniband/hw/mana/cq.c   |  4 +---
>  drivers/infiniband/hw/mana/main.c | 27 +++++++++++++--------------
>  drivers/infiniband/hw/mana/qp.c   |  5 ++---
>  3 files changed, 16 insertions(+), 20 deletions(-)
> 
> diff --git a/drivers/infiniband/hw/mana/cq.c b/drivers/infiniband/hw/mana/cq.c
> index 0fc4e26..28e154b 100644
> --- a/drivers/infiniband/hw/mana/cq.c
> +++ b/drivers/infiniband/hw/mana/cq.c
> @@ -15,14 +15,12 @@ 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;
>  	bool is_rnic_cq;
>  	u32 doorbell;
>  	u32 buf_size;
>  	int err;
> 
>  	mdev = container_of(ibdev, struct mana_ib_dev, ib_dev);
> -	gc = mdev_to_gc(mdev);
> 
>  	cq->comp_vector = attr->comp_vector % ibdev->num_comp_vectors;
>  	cq->cq_handle = INVALID_MANA_HANDLE;
> @@ -65,7 +63,7 @@ int mana_ib_create_cq(struct ib_cq *ibcq, const struct
> ib_cq_init_attr *attr,
>  			ibdev_dbg(ibdev, "Failed to create kernel queue for
> create cq, %d\n", err);
>  			return err;
>  		}
> -		doorbell = gc->mana_ib.doorbell;
> +		doorbell = mdev->gdma_dev->doorbell;
>  	}
> 
>  	if (is_rnic_cq) {
> diff --git a/drivers/infiniband/hw/mana/main.c
> b/drivers/infiniband/hw/mana/main.c
> index 3837e30..41a24a1 100644
> --- a/drivers/infiniband/hw/mana/main.c
> +++ b/drivers/infiniband/hw/mana/main.c
> @@ -244,7 +244,6 @@ void mana_ib_dealloc_ucontext(struct ib_ucontext
> *ibcontext)  int mana_ib_create_kernel_queue(struct mana_ib_dev *mdev, u32
> size, enum gdma_queue_type type,
>  				struct mana_ib_queue *queue)
>  {
> -	struct gdma_context *gc = mdev_to_gc(mdev);
>  	struct gdma_queue_spec spec = {};
>  	int err;
> 
> @@ -253,7 +252,7 @@ int mana_ib_create_kernel_queue(struct mana_ib_dev
> *mdev, u32 size, enum gdma_qu
>  	spec.type = type;
>  	spec.monitor_avl_buf = false;
>  	spec.queue_size = size;
> -	err = mana_gd_create_mana_wq_cq(&gc->mana_ib, &spec, &queue-
> >kmem);
> +	err = mana_gd_create_mana_wq_cq(mdev->gdma_dev, &spec, &queue-
> >kmem);
>  	if (err)
>  		return err;
>  	/* take ownership into mana_ib from mana */ @@ -784,7 +783,7 @@
> int mana_ib_create_eqs(struct mana_ib_dev *mdev)
>  	spec.eq.log2_throttle_limit = LOG2_EQ_THROTTLE;
>  	spec.eq.msix_index = 0;
> 
> -	err = mana_gd_create_mana_eq(&gc->mana_ib, &spec, &mdev-
> >fatal_err_eq);
> +	err = mana_gd_create_mana_eq(mdev->gdma_dev, &spec,
> +&mdev->fatal_err_eq);
>  	if (err)
>  		return err;
> 
> @@ -835,7 +834,7 @@ int mana_ib_gd_create_rnic_adapter(struct
> mana_ib_dev *mdev)
> 
>  	mana_gd_init_req_hdr(&req.hdr, MANA_IB_CREATE_ADAPTER,
> sizeof(req), sizeof(resp));
>  	req.hdr.req.msg_version = GDMA_MESSAGE_V2;
> -	req.hdr.dev_id = gc->mana_ib.dev_id;
> +	req.hdr.dev_id = mdev->gdma_dev->dev_id;
>  	req.notify_eq_id = mdev->fatal_err_eq->id;
> 
>  	if (mdev->adapter_caps.feature_flags &
> MANA_IB_FEATURE_CLIENT_ERROR_CQE_SUPPORT)
> @@ -860,7 +859,7 @@ int mana_ib_gd_destroy_rnic_adapter(struct
> mana_ib_dev *mdev)
> 
>  	gc = mdev_to_gc(mdev);
>  	mana_gd_init_req_hdr(&req.hdr, MANA_IB_DESTROY_ADAPTER,
> sizeof(req), sizeof(resp));
> -	req.hdr.dev_id = gc->mana_ib.dev_id;
> +	req.hdr.dev_id = mdev->gdma_dev->dev_id;
>  	req.adapter = mdev->adapter_handle;
> 
>  	err = mana_gd_send_request(gc, sizeof(req), &req, sizeof(resp), &resp);
> @@ -887,7 +886,7 @@ int mana_ib_gd_add_gid(const struct ib_gid_attr *attr,
> void **context)
>  	}
> 
>  	mana_gd_init_req_hdr(&req.hdr, MANA_IB_CONFIG_IP_ADDR,
> sizeof(req), sizeof(resp));
> -	req.hdr.dev_id = gc->mana_ib.dev_id;
> +	req.hdr.dev_id = mdev->gdma_dev->dev_id;
>  	req.adapter = mdev->adapter_handle;
>  	req.op = ADDR_OP_ADD;
>  	req.sgid_type = (ntype == RDMA_NETWORK_IPV6) ? SGID_TYPE_IPV6 :
> SGID_TYPE_IPV4; @@ -917,7 +916,7 @@ int mana_ib_gd_del_gid(const struct
> ib_gid_attr *attr, void **context)
>  	}
> 
>  	mana_gd_init_req_hdr(&req.hdr, MANA_IB_CONFIG_IP_ADDR,
> sizeof(req), sizeof(resp));
> -	req.hdr.dev_id = gc->mana_ib.dev_id;
> +	req.hdr.dev_id = mdev->gdma_dev->dev_id;
>  	req.adapter = mdev->adapter_handle;
>  	req.op = ADDR_OP_REMOVE;
>  	req.sgid_type = (ntype == RDMA_NETWORK_IPV6) ? SGID_TYPE_IPV6 :
> SGID_TYPE_IPV4; @@ -940,7 +939,7 @@ int mana_ib_gd_config_mac(struct
> mana_ib_dev *mdev, enum mana_ib_addr_op op, u8
>  	int err;
> 
>  	mana_gd_init_req_hdr(&req.hdr, MANA_IB_CONFIG_MAC_ADDR,
> sizeof(req), sizeof(resp));
> -	req.hdr.dev_id = gc->mana_ib.dev_id;
> +	req.hdr.dev_id = mdev->gdma_dev->dev_id;
>  	req.adapter = mdev->adapter_handle;
>  	req.op = op;
>  	copy_in_reverse(req.mac_addr, mac, ETH_ALEN); @@ -965,7 +964,7
> @@ int mana_ib_gd_create_cq(struct mana_ib_dev *mdev, struct mana_ib_cq
> *cq, u32 do
>  		return -EINVAL;
> 
>  	mana_gd_init_req_hdr(&req.hdr, MANA_IB_CREATE_CQ, sizeof(req),
> sizeof(resp));
> -	req.hdr.dev_id = gc->mana_ib.dev_id;
> +	req.hdr.dev_id = mdev->gdma_dev->dev_id;
>  	req.adapter = mdev->adapter_handle;
>  	req.gdma_region = cq->queue.gdma_region;
>  	req.eq_id = mdev->eqs[cq->comp_vector]->id; @@ -997,7 +996,7 @@
> int mana_ib_gd_destroy_cq(struct mana_ib_dev *mdev, struct mana_ib_cq *cq)
>  		return 0;
> 
>  	mana_gd_init_req_hdr(&req.hdr, MANA_IB_DESTROY_CQ, sizeof(req),
> sizeof(resp));
> -	req.hdr.dev_id = gc->mana_ib.dev_id;
> +	req.hdr.dev_id = mdev->gdma_dev->dev_id;
>  	req.adapter = mdev->adapter_handle;
>  	req.cq_handle = cq->cq_handle;
> 
> @@ -1023,7 +1022,7 @@ int mana_ib_gd_create_rc_qp(struct mana_ib_dev
> *mdev, struct mana_ib_qp *qp,
>  	int err, i;
> 
>  	mana_gd_init_req_hdr(&req.hdr, MANA_IB_CREATE_RC_QP, sizeof(req),
> sizeof(resp));
> -	req.hdr.dev_id = gc->mana_ib.dev_id;
> +	req.hdr.dev_id = mdev->gdma_dev->dev_id;
>  	req.adapter = mdev->adapter_handle;
>  	req.pd_handle = pd->pd_handle;
>  	req.send_cq_handle = send_cq->cq_handle; @@ -1059,7 +1058,7 @@
> int mana_ib_gd_destroy_rc_qp(struct mana_ib_dev *mdev, struct mana_ib_qp
> *qp)
>  	int err;
> 
>  	mana_gd_init_req_hdr(&req.hdr, MANA_IB_DESTROY_RC_QP,
> sizeof(req), sizeof(resp));
> -	req.hdr.dev_id = gc->mana_ib.dev_id;
> +	req.hdr.dev_id = mdev->gdma_dev->dev_id;
>  	req.adapter = mdev->adapter_handle;
>  	req.rc_qp_handle = qp->qp_handle;
>  	err = mana_gd_send_request(gc, sizeof(req), &req, sizeof(resp), &resp);
> @@ -1082,7 +1081,7 @@ int mana_ib_gd_create_ud_qp(struct mana_ib_dev
> *mdev, struct mana_ib_qp *qp,
>  	int err, i;
> 
>  	mana_gd_init_req_hdr(&req.hdr, MANA_IB_CREATE_UD_QP, sizeof(req),
> sizeof(resp));
> -	req.hdr.dev_id = gc->mana_ib.dev_id;
> +	req.hdr.dev_id = mdev->gdma_dev->dev_id;
>  	req.adapter = mdev->adapter_handle;
>  	req.pd_handle = pd->pd_handle;
>  	req.send_cq_handle = send_cq->cq_handle; @@ -1117,7 +1116,7 @@
> int mana_ib_gd_destroy_ud_qp(struct mana_ib_dev *mdev, struct mana_ib_qp
> *qp)
>  	int err;
> 
>  	mana_gd_init_req_hdr(&req.hdr, MANA_IB_DESTROY_UD_QP,
> sizeof(req), sizeof(resp));
> -	req.hdr.dev_id = gc->mana_ib.dev_id;
> +	req.hdr.dev_id = mdev->gdma_dev->dev_id;
>  	req.adapter = mdev->adapter_handle;
>  	req.qp_handle = qp->qp_handle;
>  	err = mana_gd_send_request(gc, sizeof(req), &req, sizeof(resp), &resp);
> diff --git a/drivers/infiniband/hw/mana/qp.c b/drivers/infiniband/hw/mana/qp.c
> index c928af5..14fd7d6 100644
> --- a/drivers/infiniband/hw/mana/qp.c
> +++ b/drivers/infiniband/hw/mana/qp.c
> @@ -635,7 +635,6 @@ static int mana_ib_create_ud_qp(struct ib_qp *ibqp,
> struct ib_pd *ibpd,  {
>  	struct mana_ib_dev *mdev = container_of(ibpd->device, struct
> mana_ib_dev, ib_dev);
>  	struct mana_ib_qp *qp = container_of(ibqp, struct mana_ib_qp, ibqp);
> -	struct gdma_context *gc = mdev_to_gc(mdev);
>  	u32 doorbell, queue_size;
>  	int i, err;
> 
> @@ -654,7 +653,7 @@ static int mana_ib_create_ud_qp(struct ib_qp *ibqp,
> struct ib_pd *ibpd,
>  			goto destroy_queues;
>  		}
>  	}
> -	doorbell = gc->mana_ib.doorbell;
> +	doorbell = mdev->gdma_dev->doorbell;
> 
>  	err = create_shadow_queue(&qp->shadow_rq, attr->cap.max_recv_wr,
>  				  sizeof(struct ud_rq_shadow_wqe)); @@ -736,7
> +735,7 @@ static int mana_ib_gd_modify_qp(struct ib_qp *ibqp, struct
> ib_qp_attr *attr,
>  	int err;
> 
>  	mana_gd_init_req_hdr(&req.hdr, MANA_IB_SET_QP_STATE, sizeof(req),
> sizeof(resp));
> -	req.hdr.dev_id = gc->mana_ib.dev_id;
> +	req.hdr.dev_id = mdev->gdma_dev->dev_id;
>  	req.adapter = mdev->adapter_handle;
>  	req.qp_handle = qp->qp_handle;
>  	req.qp_state = attr->qp_state;
> --
> 2.43.0

[PATCH rdma-next v4 4/4] net: mana: Add support for auxiliary device servicing events

[Konstantin Taranov]

From: Shiraz Saleem <shirazsaleem@microsoft.com>

Handle soc servcing events which require the rdma auxiliary device resources to
be cleaned up during a suspend, and re-initialized during a resume.

Signed-off-by: Shiraz Saleem <shirazsaleem@microsoft.com>
Signed-off-by: Konstantin Taranov <kotaranov@microsoft.com>
---
 .../net/ethernet/microsoft/mana/gdma_main.c   | 11 ++-
 .../net/ethernet/microsoft/mana/hw_channel.c  | 20 ++++++
 drivers/net/ethernet/microsoft/mana/mana_en.c | 69 +++++++++++++++++++
 include/net/mana/gdma.h                       | 19 +++++
 include/net/mana/hw_channel.h                 |  9 +++
 5 files changed, 127 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/microsoft/mana/gdma_main.c b/drivers/net/ethernet/microsoft/mana/gdma_main.c
index 59e7814..3504507 100644
--- a/drivers/net/ethernet/microsoft/mana/gdma_main.c
+++ b/drivers/net/ethernet/microsoft/mana/gdma_main.c
@@ -391,6 +391,7 @@ static void mana_gd_process_eqe(struct gdma_queue *eq)
 	case GDMA_EQE_HWC_INIT_EQ_ID_DB:
 	case GDMA_EQE_HWC_INIT_DATA:
 	case GDMA_EQE_HWC_INIT_DONE:
+	case GDMA_EQE_HWC_SOC_SERVICE:
 	case GDMA_EQE_RNIC_QP_FATAL:
 		if (!eq->eq.callback)
 			break;
@@ -1468,10 +1469,14 @@ static int mana_gd_setup(struct pci_dev *pdev)
 	mana_gd_init_registers(pdev);
 	mana_smc_init(&gc->shm_channel, gc->dev, gc->shm_base);
 
+	gc->service_wq = alloc_ordered_workqueue("gdma_service_wq", 0);
+	if (!gc->service_wq)
+		return -ENOMEM;
+
 	err = mana_gd_setup_irqs(pdev);
 	if (err) {
 		dev_err(gc->dev, "Failed to setup IRQs: %d\n", err);
-		return err;
+		goto free_workqueue;
 	}
 
 	err = mana_hwc_create_channel(gc);
@@ -1497,6 +1502,8 @@ destroy_hwc:
 	mana_hwc_destroy_channel(gc);
 remove_irq:
 	mana_gd_remove_irqs(pdev);
+free_workqueue:
+	destroy_workqueue(gc->service_wq);
 	dev_err(&pdev->dev, "%s failed (error %d)\n", __func__, err);
 	return err;
 }
@@ -1508,6 +1515,8 @@ static void mana_gd_cleanup(struct pci_dev *pdev)
 	mana_hwc_destroy_channel(gc);
 
 	mana_gd_remove_irqs(pdev);
+
+	destroy_workqueue(gc->service_wq);
 	dev_dbg(&pdev->dev, "mana gdma cleanup successful\n");
 }
 
diff --git a/drivers/net/ethernet/microsoft/mana/hw_channel.c b/drivers/net/ethernet/microsoft/mana/hw_channel.c
index 1ba4960..60f6bc1 100644
--- a/drivers/net/ethernet/microsoft/mana/hw_channel.c
+++ b/drivers/net/ethernet/microsoft/mana/hw_channel.c
@@ -112,11 +112,13 @@ out:
 static void mana_hwc_init_event_handler(void *ctx, struct gdma_queue *q_self,
 					struct gdma_event *event)
 {
+	union hwc_init_soc_service_type service_data;
 	struct hw_channel_context *hwc = ctx;
 	struct gdma_dev *gd = hwc->gdma_dev;
 	union hwc_init_type_data type_data;
 	union hwc_init_eq_id_db eq_db;
 	u32 type, val;
+	int ret;
 
 	switch (event->type) {
 	case GDMA_EQE_HWC_INIT_EQ_ID_DB:
@@ -199,7 +201,25 @@ static void mana_hwc_init_event_handler(void *ctx, struct gdma_queue *q_self,
 		}
 
 		break;
+	case GDMA_EQE_HWC_SOC_SERVICE:
+		service_data.as_uint32 = event->details[0];
+		type = service_data.type;
+		val = service_data.value;
 
+		switch (type) {
+		case GDMA_SERVICE_TYPE_RDMA_SUSPEND:
+		case GDMA_SERVICE_TYPE_RDMA_RESUME:
+			ret = mana_rdma_service_event(gd->gdma_context, type);
+			if (ret)
+				dev_err(hwc->dev, "Failed to schedule adev service event: %d\n",
+					ret);
+			break;
+		default:
+			dev_warn(hwc->dev, "Received unknown SOC service type %u\n", type);
+			break;
+		}
+
+		break;
 	default:
 		dev_warn(hwc->dev, "Received unknown gdma event %u\n", event->type);
 		/* Ignore unknown events, which should never happen. */
diff --git a/drivers/net/ethernet/microsoft/mana/mana_en.c b/drivers/net/ethernet/microsoft/mana/mana_en.c
index 2013d0e..39e01e2 100644
--- a/drivers/net/ethernet/microsoft/mana/mana_en.c
+++ b/drivers/net/ethernet/microsoft/mana/mana_en.c
@@ -2992,6 +2992,70 @@ idx_fail:
 	return ret;
 }
 
+static void mana_handle_rdma_servicing(struct work_struct *work)
+{
+	struct mana_service_work *serv_work =
+		container_of(work, struct mana_service_work, work);
+	struct gdma_dev *gd = serv_work->gdma_dev;
+	struct device *dev = gd->gdma_context->dev;
+	int ret;
+
+	if (READ_ONCE(gd->rdma_teardown))
+		goto out;
+
+	switch (serv_work->event) {
+	case GDMA_SERVICE_TYPE_RDMA_SUSPEND:
+		if (!gd->adev || gd->is_suspended)
+			break;
+
+		remove_adev(gd);
+		gd->is_suspended = true;
+		break;
+
+	case GDMA_SERVICE_TYPE_RDMA_RESUME:
+		if (!gd->is_suspended)
+			break;
+
+		ret = add_adev(gd, "rdma");
+		if (ret)
+			dev_err(dev, "Failed to add adev on resume: %d\n", ret);
+		else
+			gd->is_suspended = false;
+		break;
+
+	default:
+		dev_warn(dev, "unknown adev service event %u\n",
+			 serv_work->event);
+		break;
+	}
+
+out:
+	kfree(serv_work);
+}
+
+int mana_rdma_service_event(struct gdma_context *gc, enum gdma_service_type event)
+{
+	struct gdma_dev *gd = &gc->mana_ib;
+	struct mana_service_work *serv_work;
+
+	if (gd->dev_id.type != GDMA_DEVICE_MANA_IB) {
+		/* RDMA device is not detected on pci */
+		return 0;
+	}
+
+	serv_work = kzalloc(sizeof(*serv_work), GFP_ATOMIC);
+	if (!serv_work)
+		return -ENOMEM;
+
+	serv_work->event = event;
+	serv_work->gdma_dev = gd;
+
+	INIT_WORK(&serv_work->work, mana_handle_rdma_servicing);
+	queue_work(gc->service_wq, &serv_work->work);
+
+	return 0;
+}
+
 int mana_probe(struct gdma_dev *gd, bool resuming)
 {
 	struct gdma_context *gc = gd->gdma_context;
@@ -3172,11 +3236,16 @@ int mana_rdma_probe(struct gdma_dev *gd)
 
 void mana_rdma_remove(struct gdma_dev *gd)
 {
+	struct gdma_context *gc = gd->gdma_context;
+
 	if (gd->dev_id.type != GDMA_DEVICE_MANA_IB) {
 		/* RDMA device is not detected on pci */
 		return;
 	}
 
+	WRITE_ONCE(gd->rdma_teardown, true);
+	flush_workqueue(gc->service_wq);
+
 	if (gd->adev)
 		remove_adev(gd);
 
diff --git a/include/net/mana/gdma.h b/include/net/mana/gdma.h
index ffa9820..3ce56a8 100644
--- a/include/net/mana/gdma.h
+++ b/include/net/mana/gdma.h
@@ -60,6 +60,7 @@ enum gdma_eqe_type {
 	GDMA_EQE_HWC_INIT_DONE		= 131,
 	GDMA_EQE_HWC_SOC_RECONFIG	= 132,
 	GDMA_EQE_HWC_SOC_RECONFIG_DATA	= 133,
+	GDMA_EQE_HWC_SOC_SERVICE	= 134,
 	GDMA_EQE_RNIC_QP_FATAL		= 176,
 };
 
@@ -70,6 +71,18 @@ enum {
 	GDMA_DEVICE_MANA_IB	= 3,
 };
 
+enum gdma_service_type {
+	GDMA_SERVICE_TYPE_NONE		= 0,
+	GDMA_SERVICE_TYPE_RDMA_SUSPEND	= 1,
+	GDMA_SERVICE_TYPE_RDMA_RESUME	= 2,
+};
+
+struct mana_service_work {
+	struct work_struct work;
+	struct gdma_dev *gdma_dev;
+	enum gdma_service_type event;
+};
+
 struct gdma_resource {
 	/* Protect the bitmap */
 	spinlock_t lock;
@@ -224,6 +237,8 @@ struct gdma_dev {
 	void *driver_data;
 
 	struct auxiliary_device *adev;
+	bool is_suspended;
+	bool rdma_teardown;
 };
 
 /* MANA_PAGE_SIZE is the DMA unit */
@@ -409,6 +424,8 @@ struct gdma_context {
 	struct gdma_dev		mana_ib;
 
 	u64 pf_cap_flags1;
+
+	struct workqueue_struct *service_wq;
 };
 
 static inline bool mana_gd_is_mana(struct gdma_dev *gd)
@@ -891,4 +908,6 @@ int mana_gd_destroy_dma_region(struct gdma_context *gc, u64 dma_region_handle);
 void mana_register_debugfs(void);
 void mana_unregister_debugfs(void);
 
+int mana_rdma_service_event(struct gdma_context *gc, enum gdma_service_type event);
+
 #endif /* _GDMA_H */
diff --git a/include/net/mana/hw_channel.h b/include/net/mana/hw_channel.h
index 158b125..83cf933 100644
--- a/include/net/mana/hw_channel.h
+++ b/include/net/mana/hw_channel.h
@@ -49,6 +49,15 @@ union hwc_init_type_data {
 	};
 }; /* HW DATA */
 
+union hwc_init_soc_service_type {
+	u32 as_uint32;
+
+	struct {
+		u32 value	: 28;
+		u32 type	:  4;
+	};
+}; /* HW DATA */
+
 struct hwc_rx_oob {
 	u32 type	: 6;
 	u32 eom		: 1;
-- 
2.43.0

Re: [PATCH rdma-next v4 4/4] net: mana: Add support for auxiliary device servicing events

[ALOK TIWARI]

On 07-05-2025 21:29, Konstantin Taranov wrote:
> From: Shiraz Saleem <shirazsaleem@microsoft.com>
> 
> Handle soc servcing events which require the rdma auxiliary device resources to

typo servcing ->servicing

> be cleaned up during a suspend, and re-initialized during a resume.
> 
> Signed-off-by: Shiraz Saleem <shirazsaleem@microsoft.com>
> Signed-off-by: Konstantin Taranov <kotaranov@microsoft.com>
> ---
>   .../net/ethernet/microsoft/mana/gdma_main.c   | 11 ++-
>   .../net/ethernet/microsoft/mana/hw_channel.c  | 20 ++++++
>   drivers/net/ethernet/microsoft/mana/mana_en.c | 69 +++++++++++++++++++
>   include/net/mana/gdma.h                       | 19 +++++
>   include/net/mana/hw_channel.h                 |  9 +++
>   5 files changed, 127 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/net/ethernet/microsoft/mana/gdma_main.c b/drivers/net/ethernet/microsoft/mana/gdma_main.c
> index 59e7814..3504507 100644
> --- a/drivers/net/ethernet/microsoft/mana/gdma_main.c
> +++ b/drivers/net/ethernet/microsoft/mana/gdma_main.c
> @@ -391,6 +391,7 @@ static void mana_gd_process_eqe(struct gdma_queue *eq)
>   	case GDMA_EQE_HWC_INIT_EQ_ID_DB:
>   	case GDMA_EQE_HWC_INIT_DATA:
>   	case GDMA_EQE_HWC_INIT_DONE:
> +	case GDMA_EQE_HWC_SOC_SERVICE:
>   	case GDMA_EQE_RNIC_QP_FATAL:
>   		if (!eq->eq.callback)
>   			break;
> @@ -1468,10 +1469,14 @@ static int mana_gd_setup(struct pci_dev *pdev)
>   	mana_gd_init_registers(pdev);
>   	mana_smc_init(&gc->shm_channel, gc->dev, gc->shm_base);
>   
> +	gc->service_wq = alloc_ordered_workqueue("gdma_service_wq", 0);
> +	if (!gc->service_wq)
> +		return -ENOMEM;
> +
>   	err = mana_gd_setup_irqs(pdev);
>   	if (err) {
>   		dev_err(gc->dev, "Failed to setup IRQs: %d\n", err);
> -		return err;
> +		goto free_workqueue;
>   	}
>   
>   	err = mana_hwc_create_channel(gc);
> @@ -1497,6 +1502,8 @@ destroy_hwc:
>   	mana_hwc_destroy_channel(gc);
>   remove_irq:
>   	mana_gd_remove_irqs(pdev);
> +free_workqueue:
> +	destroy_workqueue(gc->service_wq);
>   	dev_err(&pdev->dev, "%s failed (error %d)\n", __func__, err);
>   	return err;
>   }
> @@ -1508,6 +1515,8 @@ static void mana_gd_cleanup(struct pci_dev *pdev)
>   	mana_hwc_destroy_channel(gc);
>   
>   	mana_gd_remove_irqs(pdev);
> +
> +	destroy_workqueue(gc->service_wq);
>   	dev_dbg(&pdev->dev, "mana gdma cleanup successful\n");
>   }
>   
> diff --git a/drivers/net/ethernet/microsoft/mana/hw_channel.c b/drivers/net/ethernet/microsoft/mana/hw_channel.c
> index 1ba4960..60f6bc1 100644
> --- a/drivers/net/ethernet/microsoft/mana/hw_channel.c
> +++ b/drivers/net/ethernet/microsoft/mana/hw_channel.c
> @@ -112,11 +112,13 @@ out:
>   static void mana_hwc_init_event_handler(void *ctx, struct gdma_queue *q_self,
>   					struct gdma_event *event)
>   {
> +	union hwc_init_soc_service_type service_data;
>   	struct hw_channel_context *hwc = ctx;
>   	struct gdma_dev *gd = hwc->gdma_dev;
>   	union hwc_init_type_data type_data;
>   	union hwc_init_eq_id_db eq_db;
>   	u32 type, val;
> +	int ret;
>   
>   	switch (event->type) {
>   	case GDMA_EQE_HWC_INIT_EQ_ID_DB:
> @@ -199,7 +201,25 @@ static void mana_hwc_init_event_handler(void *ctx, struct gdma_queue *q_self,
>   		}
>   
>   		break;
> +	case GDMA_EQE_HWC_SOC_SERVICE:
> +		service_data.as_uint32 = event->details[0];
> +		type = service_data.type;
> +		val = service_data.value;

what is use of val?

>   
> +		switch (type) {
> +		case GDMA_SERVICE_TYPE_RDMA_SUSPEND:
> +		case GDMA_SERVICE_TYPE_RDMA_RESUME:
> +			ret = mana_rdma_service_event(gd->gdma_context, type);
> +			if (ret)
> +				dev_err(hwc->dev, "Failed to schedule adev service event: %d\n",
> +					ret);
> +			break;
> +		default:
> +			dev_warn(hwc->dev, "Received unknown SOC service type %u\n", type);
> +			break;
> +		}
> +
> +		break;
>   	default:
>   		dev_warn(hwc->dev, "Received unknown gdma event %u\n", event->type);
>   		/* Ignore unknown events, which should never happen. */
> diff --git a/drivers/net/ethernet/microsoft/mana/mana_en.c b/drivers/net/ethernet/microsoft/mana/mana_en.c
> index 2013d0e..39e01e2 100644
> --- a/drivers/net/ethernet/microsoft/mana/mana_en.c
> +++ b/drivers/net/ethernet/microsoft/mana/mana_en.c
> @@ -2992,6 +2992,70 @@ idx_fail:
>   	return ret;
>   }
>   
> +static void mana_handle_rdma_servicing(struct work_struct *work)

this name does not sound clearer and more aligned with typical naming 
conventions.
Since it is an RDMA service event handler, it could be named to 
mana_rdma_service_handle. What are your thoughts on this?"

> +{
> +	struct mana_service_work *serv_work =
> +		container_of(work, struct mana_service_work, work);
> +	struct gdma_dev *gd = serv_work->gdma_dev;
> +	struct device *dev = gd->gdma_context->dev;
> +	int ret;
> +
> +	if (READ_ONCE(gd->rdma_teardown))
> +		goto out;
> +
> +	switch (serv_work->event) {
> +	case GDMA_SERVICE_TYPE_RDMA_SUSPEND:
> +		if (!gd->adev || gd->is_suspended)
> +			break;
> +
> +		remove_adev(gd);
> +		gd->is_suspended = true;
> +		break;
> +
> +	case GDMA_SERVICE_TYPE_RDMA_RESUME:
> +		if (!gd->is_suspended)
> +			break;
> +
> +		ret = add_adev(gd, "rdma");
> +		if (ret)
> +			dev_err(dev, "Failed to add adev on resume: %d\n", ret);
> +		else
> +			gd->is_suspended = false;
> +		break;
> +
> +	default:
> +		dev_warn(dev, "unknown adev service event %u\n",
> +			 serv_work->event);
> +		break;
> +	}
> +
> +out:
> +	kfree(serv_work);
> +}
> +
> +int mana_rdma_service_event(struct gdma_context *gc, enum gdma_service_type event)
> +{
> +	struct gdma_dev *gd = &gc->mana_ib;
> +	struct mana_service_work *serv_work;
> +
> +	if (gd->dev_id.type != GDMA_DEVICE_MANA_IB) {
> +		/* RDMA device is not detected on pci */
> +		return 0;
> +	}
> +
> +	serv_work = kzalloc(sizeof(*serv_work), GFP_ATOMIC);
> +	if (!serv_work)
> +		return -ENOMEM;
> +
> +	serv_work->event = event;
> +	serv_work->gdma_dev = gd;
> +
> +	INIT_WORK(&serv_work->work, mana_handle_rdma_servicing);
> +	queue_work(gc->service_wq, &serv_work->work);
> +
> +	return 0;
> +}
> +
>   int mana_probe(struct gdma_dev *gd, bool resuming)
>   {
>   	struct gdma_context *gc = gd->gdma_context;
> @@ -3172,11 +3236,16 @@ int mana_rdma_probe(struct gdma_dev *gd)
>   
>   void mana_rdma_remove(struct gdma_dev *gd)
>   {
> +	struct gdma_context *gc = gd->gdma_context;
> +
>   	if (gd->dev_id.type != GDMA_DEVICE_MANA_IB) {
>   		/* RDMA device is not detected on pci */
>   		return;
>   	}
>   
> +	WRITE_ONCE(gd->rdma_teardown, true);
> +	flush_workqueue(gc->service_wq);
> +
>   	if (gd->adev)
>   		remove_adev(gd);
>   
> diff --git a/include/net/mana/gdma.h b/include/net/mana/gdma.h
> index ffa9820..3ce56a8 100644
> --- a/include/net/mana/gdma.h
> +++ b/include/net/mana/gdma.h
> @@ -60,6 +60,7 @@ enum gdma_eqe_type {
>   	GDMA_EQE_HWC_INIT_DONE		= 131,
>   	GDMA_EQE_HWC_SOC_RECONFIG	= 132,
>   	GDMA_EQE_HWC_SOC_RECONFIG_DATA	= 133,
> +	GDMA_EQE_HWC_SOC_SERVICE	= 134,
>   	GDMA_EQE_RNIC_QP_FATAL		= 176,
>   };
>   
> @@ -70,6 +71,18 @@ enum {
>   	GDMA_DEVICE_MANA_IB	= 3,
>   };
>   
> +enum gdma_service_type {
> +	GDMA_SERVICE_TYPE_NONE		= 0,
> +	GDMA_SERVICE_TYPE_RDMA_SUSPEND	= 1,
> +	GDMA_SERVICE_TYPE_RDMA_RESUME	= 2,
> +};
> +
> +struct mana_service_work {
> +	struct work_struct work;
> +	struct gdma_dev *gdma_dev;
> +	enum gdma_service_type event;
> +};
> +
>   struct gdma_resource {
>   	/* Protect the bitmap */
>   	spinlock_t lock;
> @@ -224,6 +237,8 @@ struct gdma_dev {
>   	void *driver_data;
>   
>   	struct auxiliary_device *adev;
> +	bool is_suspended;
> +	bool rdma_teardown;
>   };
>   
>   /* MANA_PAGE_SIZE is the DMA unit */
> @@ -409,6 +424,8 @@ struct gdma_context {
>   	struct gdma_dev		mana_ib;
>   
>   	u64 pf_cap_flags1;
> +
> +	struct workqueue_struct *service_wq;
>   };

Thanks,
Alok

Re: [PATCH rdma-next v4 4/4] net: mana: Add support for auxiliary device servicing events

[Leon Romanovsky]

On Mon, May 12, 2025 at 10:59:43AM +0530, ALOK TIWARI wrote:
> 
> 
> On 07-05-2025 21:29, Konstantin Taranov wrote:
> > From: Shiraz Saleem <shirazsaleem@microsoft.com>
> > 
> > Handle soc servcing events which require the rdma auxiliary device resources to
> 
> typo servcing ->servicing

<...>

> > +		val = service_data.value;
> 
> what is use of val?
> 

<...>

> > +static void mana_handle_rdma_servicing(struct work_struct *work)
> 
> this name does not sound clearer and more aligned with typical naming
> conventions.
> Since it is an RDMA service event handler, it could be named to
> mana_rdma_service_handle. What are your thoughts on this?"

I fixed all these comments locally when applied the series.
Thank you for the review.

Re: [PATCH rdma-next v4 0/4] RDMA/mana_ib: allow separate mana_ib for each mana client

[Leon Romanovsky]

On Wed, 07 May 2025 08:59:01 -0700, Konstantin Taranov wrote:
> From: Konstantin Taranov <kotaranov@microsoft.com>
> 
> Microsoft mana adapter has 2 devices in the HW: mana ethernet device and RNIC device.
> Both devices can implement RDMA drivers and, so far, they have been sharing
> one ib device context. However, they are different devices with different
> capabilities in the HW and have different lifetime model.
> 
> [...]

Applied, thanks!

[1/4] net: mana: Probe rdma device in mana driver
      https://git.kernel.org/rdma/rdma/c/ced82fce77e933
[2/4] RDMA/mana_ib: Add support of mana_ib for RNIC and ETH nic
      https://git.kernel.org/rdma/rdma/c/c390828d4d7b45
[3/4] RDMA/mana_ib: unify mana_ib functions to support any gdma device
      https://git.kernel.org/rdma/rdma/c/d4293f96ce0b6d
[4/4] net: mana: Add support for auxiliary device servicing events
      https://git.kernel.org/rdma/rdma/c/51bb42e5db2b2f

Best regards,
-- 
Leon Romanovsky <leon@kernel.org>