[patch rdma-next v3 1/2] net: mana: Change the function signature of mana_get_primary_netdev_rcu

[patch rdma-next v3 1/2] net: mana: Change the function signature of mana_get_primary_netdev_rcu

[longli]

From: Long Li <longli@microsoft.com>

Change mana_get_primary_netdev_rcu() to mana_get_primary_netdev(), and
return the ndev with refcount held. The caller is responsible for dropping
the refcount by calling dev_put().

Signed-off-by: Long Li <longli@microsoft.com>
---
 drivers/infiniband/hw/mana/device.c           |  7 +++----
 drivers/net/ethernet/microsoft/mana/mana_en.c | 11 +++++++----
 include/net/mana/mana.h                       |  2 +-
 3 files changed, 11 insertions(+), 9 deletions(-)

diff --git a/drivers/infiniband/hw/mana/device.c b/drivers/infiniband/hw/mana/device.c
index 3416a85f8738..afe3d8d20b3b 100644
--- a/drivers/infiniband/hw/mana/device.c
+++ b/drivers/infiniband/hw/mana/device.c
@@ -84,10 +84,8 @@ static int mana_ib_probe(struct auxiliary_device *adev,
 	dev->ib_dev.num_comp_vectors = mdev->gdma_context->max_num_queues;
 	dev->ib_dev.dev.parent = mdev->gdma_context->dev;
 
-	rcu_read_lock(); /* required to get primary netdev */
-	ndev = mana_get_primary_netdev_rcu(mc, 0);
+	ndev = mana_get_primary_netdev(mc, 0);
 	if (!ndev) {
-		rcu_read_unlock();
 		ret = -ENODEV;
 		ibdev_err(&dev->ib_dev, "Failed to get netdev for IB port 1");
 		goto free_ib_device;
@@ -95,7 +93,8 @@ static int mana_ib_probe(struct auxiliary_device *adev,
 	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);
-	rcu_read_unlock();
+	/* mana_get_primary_netdev() returns ndev with refcount held */
+	dev_put(ndev);
 	if (ret) {
 		ibdev_err(&dev->ib_dev, "Failed to set ib netdev, ret %d", ret);
 		goto free_ib_device;
diff --git a/drivers/net/ethernet/microsoft/mana/mana_en.c b/drivers/net/ethernet/microsoft/mana/mana_en.c
index aa1e47233fe5..cfa664d5a137 100644
--- a/drivers/net/ethernet/microsoft/mana/mana_en.c
+++ b/drivers/net/ethernet/microsoft/mana/mana_en.c
@@ -3131,21 +3131,24 @@ void mana_remove(struct gdma_dev *gd, bool suspending)
 	kfree(ac);
 }
 
-struct net_device *mana_get_primary_netdev_rcu(struct mana_context *ac, u32 port_index)
+struct net_device *mana_get_primary_netdev(struct mana_context *ac, u32 port_index)
 {
 	struct net_device *ndev;
 
-	RCU_LOCKDEP_WARN(!rcu_read_lock_held(),
-			 "Taking primary netdev without holding the RCU read lock");
 	if (port_index >= ac->num_ports)
 		return NULL;
 
+	rcu_read_lock();
+
 	/* When mana is used in netvsc, the upper netdevice should be returned. */
 	if (ac->ports[port_index]->flags & IFF_SLAVE)
 		ndev = netdev_master_upper_dev_get_rcu(ac->ports[port_index]);
 	else
 		ndev = ac->ports[port_index];
 
+	dev_hold(ndev);
+	rcu_read_unlock();
+
 	return ndev;
 }
-EXPORT_SYMBOL_NS(mana_get_primary_netdev_rcu, "NET_MANA");
+EXPORT_SYMBOL_NS(mana_get_primary_netdev, "NET_MANA");
diff --git a/include/net/mana/mana.h b/include/net/mana/mana.h
index 0d00b24eacaf..770359ce7d3b 100644
--- a/include/net/mana/mana.h
+++ b/include/net/mana/mana.h
@@ -827,5 +827,5 @@ int mana_cfg_vport(struct mana_port_context *apc, u32 protection_dom_id,
 		   u32 doorbell_pg_id);
 void mana_uncfg_vport(struct mana_port_context *apc);
 
-struct net_device *mana_get_primary_netdev_rcu(struct mana_context *ac, u32 port_index);
+struct net_device *mana_get_primary_netdev(struct mana_context *ac, u32 port_index);
 #endif /* _MANA_H */
-- 
2.34.1

[patch rdma-next v3 2/2] RDMA/mana_ib: Handle net event for pointing to the current netdev

[longli]

From: Long Li <longli@microsoft.com>

When running under Hyper-V, the master device to the RDMA device is always
bonded to this RDMA device. This is not user-configurable.

The master device can be unbind/bind from the kernel. During those events,
the RDMA device should set to the current netdev to relect the change of
master device from those events.

Signed-off-by: Long Li <longli@microsoft.com>
---
Changes
v2: Add missing error handling when register_netdevice_notifier() fails.
v3: Change mana_get_primary_netdev() to return with netdev refcount held.

 drivers/infiniband/hw/mana/device.c  | 46 ++++++++++++++++++++++++++--
 drivers/infiniband/hw/mana/mana_ib.h |  1 +
 2 files changed, 45 insertions(+), 2 deletions(-)

diff --git a/drivers/infiniband/hw/mana/device.c b/drivers/infiniband/hw/mana/device.c
index afe3d8d20b3b..dc03f2d8c1b6 100644
--- a/drivers/infiniband/hw/mana/device.c
+++ b/drivers/infiniband/hw/mana/device.c
@@ -51,6 +51,37 @@ static const struct ib_device_ops mana_ib_dev_ops = {
 			   ib_ind_table),
 };
 
+static int mana_ib_netdev_event(struct notifier_block *this,
+				unsigned long event, void *ptr)
+{
+	struct mana_ib_dev *dev = container_of(this, struct mana_ib_dev, nb);
+	struct net_device *event_dev = netdev_notifier_info_to_dev(ptr);
+	struct gdma_context *gc = dev->gdma_dev->gdma_context;
+	struct mana_context *mc = gc->mana.driver_data;
+	struct net_device *ndev;
+
+	if (event_dev != mc->ports[0])
+		return NOTIFY_DONE;
+
+	switch (event) {
+	case NETDEV_CHANGEUPPER:
+		ndev = mana_get_primary_netdev(mc, 0);
+		/*
+		 * RDMA core will setup GID based on updated netdev.
+		 * It's not possible to race with the core as rtnl lock is being
+		 * held.
+		 */
+		ib_device_set_netdev(&dev->ib_dev, ndev, 1);
+
+		/* mana_get_primary_netdev() returns ndev with refcount held */
+		dev_put(ndev);
+
+		return NOTIFY_OK;
+	default:
+		return NOTIFY_DONE;
+	}
+}
+
 static int mana_ib_probe(struct auxiliary_device *adev,
 			 const struct auxiliary_device_id *id)
 {
@@ -108,17 +139,25 @@ static int mana_ib_probe(struct auxiliary_device *adev,
 	}
 	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_device;
+		goto deregister_net_notifier;
 	}
 
 	ret = mana_ib_create_eqs(dev);
 	if (ret) {
 		ibdev_err(&dev->ib_dev, "Failed to create EQs, ret %d", ret);
-		goto deregister_device;
+		goto deregister_net_notifier;
 	}
 
 	ret = mana_ib_gd_create_rnic_adapter(dev);
@@ -147,6 +186,8 @@ static int mana_ib_probe(struct auxiliary_device *adev,
 	mana_ib_gd_destroy_rnic_adapter(dev);
 destroy_eqs:
 	mana_ib_destroy_eqs(dev);
+deregister_net_notifier:
+	unregister_netdevice_notifier(&dev->nb);
 deregister_device:
 	mana_gd_deregister_device(dev->gdma_dev);
 free_ib_device:
@@ -162,6 +203,7 @@ static void mana_ib_remove(struct auxiliary_device *adev)
 	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);
 }
diff --git a/drivers/infiniband/hw/mana/mana_ib.h b/drivers/infiniband/hw/mana/mana_ib.h
index b53a5b4de908..d88187072899 100644
--- a/drivers/infiniband/hw/mana/mana_ib.h
+++ b/drivers/infiniband/hw/mana/mana_ib.h
@@ -64,6 +64,7 @@ struct mana_ib_dev {
 	struct gdma_queue **eqs;
 	struct xarray qp_table_wq;
 	struct mana_ib_adapter_caps adapter_caps;
+	struct notifier_block nb;
 };
 
 struct mana_ib_wq {
-- 
2.34.1

Re: [patch rdma-next v3 1/2] net: mana: Change the function signature of mana_get_primary_netdev_rcu

[Jakub Kicinski]

On Tue,  4 Mar 2025 16:00:01 -0800 longli@linuxonhyperv.com wrote:
> +	dev_hold(ndev);

netdev_hold(), please

RE: [EXTERNAL] Re: [patch rdma-next v3 1/2] net: mana: Change the function signature of mana_get_primary_netdev_rcu

[Long Li]

> On Tue,  4 Mar 2025 16:00:01 -0800 longli@linuxonhyperv.com wrote:
> > +	dev_hold(ndev);
> 
> netdev_hold(), please

Will change to netdev_hold().

Is it mandatory to use a tracker, or is it okay to use NULL like the following code:
https://web.git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/net/ethernet/marvell/octeon_ep_vf/octep_vf_main.c?h=v6.14-rc5#n837

Thanks,
Long

Re: [patch rdma-next v3 1/2] net: mana: Change the function signature of mana_get_primary_netdev_rcu

[Jason Gunthorpe]

On Tue, Mar 04, 2025 at 04:00:01PM -0800, longli@linuxonhyperv.com wrote:
>  
> -struct net_device *mana_get_primary_netdev_rcu(struct mana_context *ac, u32 port_index)
> +struct net_device *mana_get_primary_netdev(struct mana_context *ac, u32 port_index)
>  {
>  	struct net_device *ndev;
>  
> -	RCU_LOCKDEP_WARN(!rcu_read_lock_held(),
> -			 "Taking primary netdev without holding the RCU read lock");
>  	if (port_index >= ac->num_ports)
>  		return NULL;
>  
> +	rcu_read_lock();
> +
>  	/* When mana is used in netvsc, the upper netdevice should be returned. */
>  	if (ac->ports[port_index]->flags & IFF_SLAVE)
>  		ndev = netdev_master_upper_dev_get_rcu(ac->ports[port_index]);
>  	else
>  		ndev = ac->ports[port_index];
>  
> +	dev_hold(ndev);
> +	rcu_read_unlock();

That's much better, yes

Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>

Jason