diff mbox series

[for-next,1/2] {topost} RDMA/hns: Deliver net device event to ofed

Message ID 1570184954-21384-2-git-send-email-liweihang@hisilicon.com (mailing list archive)
State Changes Requested
Delegated to: Jason Gunthorpe
Headers show
Series RDMA/hns: Add support of net device event reporting to ULP | expand

Commit Message

Weihang Li Oct. 4, 2019, 10:29 a.m. UTC 
From: Lang Cheng <chenglang@huawei.com>

Driver can notify ULP with IB event when net link down/up.

Signed-off-by: Lang Cheng <chenglang@huawei.com>
Signed-off-by: Weihang Li <liweihang@hisilicon.com>
---
 drivers/infiniband/hw/hns/hns_roce_device.h | 12 +++++++++
 drivers/infiniband/hw/hns/hns_roce_main.c   | 40 +++++++++++++++++++++--------
 2 files changed, 42 insertions(+), 10 deletions(-)

Comments

Jason Gunthorpe Oct. 4, 2019, 6:22 p.m. UTC | #1 
On Fri, Oct 04, 2019 at 06:29:13PM +0800, Weihang Li wrote:
> From: Lang Cheng <chenglang@huawei.com>
> 
> Driver can notify ULP with IB event when net link down/up.
> 
> Signed-off-by: Lang Cheng <chenglang@huawei.com>
> Signed-off-by: Weihang Li <liweihang@hisilicon.com>
>  drivers/infiniband/hw/hns/hns_roce_device.h | 12 +++++++++
>  drivers/infiniband/hw/hns/hns_roce_main.c   | 40 +++++++++++++++++++++--------
>  2 files changed, 42 insertions(+), 10 deletions(-)
> 
> diff --git a/drivers/infiniband/hw/hns/hns_roce_device.h b/drivers/infiniband/hw/hns/hns_roce_device.h
> index 05d375a..47c209f 100644
> +++ b/drivers/infiniband/hw/hns/hns_roce_device.h
> @@ -711,6 +711,7 @@ struct hns_roce_ib_iboe {
>  	struct net_device      *netdevs[HNS_ROCE_MAX_PORTS];
>  	struct notifier_block	nb;
>  	u8			phy_port[HNS_ROCE_MAX_PORTS];
> +	enum ib_port_state	port_state[HNS_ROCE_MAX_PORTS];
>  };
>  
>  enum {
> @@ -1135,6 +1136,17 @@ static inline void *hns_roce_buf_offset(struct hns_roce_buf *buf, int offset)
>  		       (offset & (page_size - 1));
>  }
>  
> +static inline u8 to_rdma_port_num(u8 phy_port_num)
> +{
> +	return phy_port_num + 1;
> +}
> +
> +static inline enum ib_port_state get_port_state(struct net_device *net_dev)
> +{
> +	return (netif_running(net_dev) && netif_carrier_ok(net_dev)) ?
> +		IB_PORT_ACTIVE : IB_PORT_DOWN;
> +}
> +
>  int hns_roce_init_uar_table(struct hns_roce_dev *dev);
>  int hns_roce_uar_alloc(struct hns_roce_dev *dev, struct hns_roce_uar *uar);
>  void hns_roce_uar_free(struct hns_roce_dev *dev, struct hns_roce_uar *uar);
> diff --git a/drivers/infiniband/hw/hns/hns_roce_main.c b/drivers/infiniband/hw/hns/hns_roce_main.c
> index 665ce24..742a36e 100644
> +++ b/drivers/infiniband/hw/hns/hns_roce_main.c
> @@ -103,10 +103,13 @@ static int hns_roce_del_gid(const struct ib_gid_attr *attr, void **context)
>  }
>  
>  static int handle_en_event(struct hns_roce_dev *hr_dev, u8 port,
> -			   unsigned long event)
> +			   unsigned long dev_event)
>  {
>  	struct device *dev = hr_dev->dev;
> +	enum ib_port_state port_state;
>  	struct net_device *netdev;
> +	struct ib_event event;
> +	unsigned long flags;
>  	int ret = 0;
>  
>  	netdev = hr_dev->iboe.netdevs[port];
> @@ -115,20 +118,38 @@ static int handle_en_event(struct hns_roce_dev *hr_dev, u8 port,
>  		return -ENODEV;
>  	}
>  
> -	switch (event) {
> -	case NETDEV_UP:
> -	case NETDEV_CHANGE:
> +	switch (dev_event) {
>  	case NETDEV_REGISTER:
>  	case NETDEV_CHANGEADDR:
>  		ret = hns_roce_set_mac(hr_dev, port, netdev->dev_addr);
>  		break;
> +	case NETDEV_UP:
> +	case NETDEV_CHANGE:
> +		ret = hns_roce_set_mac(hr_dev, port, netdev->dev_addr);
> +		if (ret)
> +			return ret;
> +		/* port up/down events need send ib events */
>  	case NETDEV_DOWN:
> -		/*
> -		 * In v1 engine, only support all ports closed together.
> -		 */
> +		port_state = get_port_state(netdev);
> +
> +		spin_lock_irqsave(&hr_dev->iboe.lock, flags);
> +		if (hr_dev->iboe.port_state[port] == port_state) {
> +			spin_unlock_irqrestore(&hr_dev->iboe.lock, flags);
> +			return NOTIFY_DONE;
> +		}
> +		hr_dev->iboe.port_state[port] = port_state;
> +		spin_unlock_irqrestore(&hr_dev->iboe.lock, flags);
> +
> +		event.device = &hr_dev->ib_dev;
> +		event.event = (port_state == IB_PORT_ACTIVE) ?
> +			      IB_EVENT_PORT_ACTIVE : IB_EVENT_PORT_ERR;
> +		event.element.port_num = to_rdma_port_num(port);
> +		ib_dispatch_event(&event);
> +		break;

Every roce driver shouldn't have this kind of stuff, please try to
make a version to add this to the core code. Now that we have the
ib_device_set_netdev() it should be possible

Also, this driver should be using ib_device_set_netdev and the sketchy
code in hns_roce_netdev_event() removed

Jason
diff mbox series

Patch

diff --git a/drivers/infiniband/hw/hns/hns_roce_device.h b/drivers/infiniband/hw/hns/hns_roce_device.h
index 05d375a..47c209f 100644
--- a/drivers/infiniband/hw/hns/hns_roce_device.h
+++ b/drivers/infiniband/hw/hns/hns_roce_device.h
@@ -711,6 +711,7 @@  struct hns_roce_ib_iboe {
 	struct net_device      *netdevs[HNS_ROCE_MAX_PORTS];
 	struct notifier_block	nb;
 	u8			phy_port[HNS_ROCE_MAX_PORTS];
+	enum ib_port_state	port_state[HNS_ROCE_MAX_PORTS];
 };
 
 enum {
@@ -1135,6 +1136,17 @@  static inline void *hns_roce_buf_offset(struct hns_roce_buf *buf, int offset)
 		       (offset & (page_size - 1));
 }
 
+static inline u8 to_rdma_port_num(u8 phy_port_num)
+{
+	return phy_port_num + 1;
+}
+
+static inline enum ib_port_state get_port_state(struct net_device *net_dev)
+{
+	return (netif_running(net_dev) && netif_carrier_ok(net_dev)) ?
+		IB_PORT_ACTIVE : IB_PORT_DOWN;
+}
+
 int hns_roce_init_uar_table(struct hns_roce_dev *dev);
 int hns_roce_uar_alloc(struct hns_roce_dev *dev, struct hns_roce_uar *uar);
 void hns_roce_uar_free(struct hns_roce_dev *dev, struct hns_roce_uar *uar);
diff --git a/drivers/infiniband/hw/hns/hns_roce_main.c b/drivers/infiniband/hw/hns/hns_roce_main.c
index 665ce24..742a36e 100644
--- a/drivers/infiniband/hw/hns/hns_roce_main.c
+++ b/drivers/infiniband/hw/hns/hns_roce_main.c
@@ -103,10 +103,13 @@  static int hns_roce_del_gid(const struct ib_gid_attr *attr, void **context)
 }
 
 static int handle_en_event(struct hns_roce_dev *hr_dev, u8 port,
-			   unsigned long event)
+			   unsigned long dev_event)
 {
 	struct device *dev = hr_dev->dev;
+	enum ib_port_state port_state;
 	struct net_device *netdev;
+	struct ib_event event;
+	unsigned long flags;
 	int ret = 0;
 
 	netdev = hr_dev->iboe.netdevs[port];
@@ -115,20 +118,38 @@  static int handle_en_event(struct hns_roce_dev *hr_dev, u8 port,
 		return -ENODEV;
 	}
 
-	switch (event) {
-	case NETDEV_UP:
-	case NETDEV_CHANGE:
+	switch (dev_event) {
 	case NETDEV_REGISTER:
 	case NETDEV_CHANGEADDR:
 		ret = hns_roce_set_mac(hr_dev, port, netdev->dev_addr);
 		break;
+	case NETDEV_UP:
+	case NETDEV_CHANGE:
+		ret = hns_roce_set_mac(hr_dev, port, netdev->dev_addr);
+		if (ret)
+			return ret;
+		/* port up/down events need send ib events */
 	case NETDEV_DOWN:
-		/*
-		 * In v1 engine, only support all ports closed together.
-		 */
+		port_state = get_port_state(netdev);
+
+		spin_lock_irqsave(&hr_dev->iboe.lock, flags);
+		if (hr_dev->iboe.port_state[port] == port_state) {
+			spin_unlock_irqrestore(&hr_dev->iboe.lock, flags);
+			return NOTIFY_DONE;
+		}
+		hr_dev->iboe.port_state[port] = port_state;
+		spin_unlock_irqrestore(&hr_dev->iboe.lock, flags);
+
+		event.device = &hr_dev->ib_dev;
+		event.event = (port_state == IB_PORT_ACTIVE) ?
+			      IB_EVENT_PORT_ACTIVE : IB_EVENT_PORT_ERR;
+		event.element.port_num = to_rdma_port_num(port);
+		ib_dispatch_event(&event);
+		break;
+	case NETDEV_UNREGISTER:
 		break;
 	default:
-		dev_dbg(dev, "NETDEV event = 0x%x!\n", (u32)(event));
+		dev_dbg(dev, "NETDEV event = 0x%x!\n", (u32)(dev_event));
 		break;
 	}
 
@@ -259,8 +280,7 @@  static int hns_roce_query_port(struct ib_device *ib_dev, u8 port_num,
 
 	mtu = iboe_get_mtu(net_dev->mtu);
 	props->active_mtu = mtu ? min(props->max_mtu, mtu) : IB_MTU_256;
-	props->state = (netif_running(net_dev) && netif_carrier_ok(net_dev)) ?
-			IB_PORT_ACTIVE : IB_PORT_DOWN;
+	props->state = get_port_state(net_dev);
 	props->phys_state = (props->state == IB_PORT_ACTIVE) ?
 			     IB_PORT_PHYS_STATE_LINK_UP :
 			     IB_PORT_PHYS_STATE_DISABLED;