hv_netvsc: Prevent packet loss during VF add/remove
v2 cover hv_netvsc: Prevent packet loss during VF add/remove
From: Long Li <[email protected]>
This patch set fixes issues with packet loss on VF add/remove.
Long Li (3):
hv_netvsc: Check VF datapath when sending traffic to VF
hv_netvsc: Wait for completion on request SWITCH_DATA_PATH
hv_netvsc: Process NETDEV_GOING_DOWN on VF hot remove
drivers/net/hyperv/netvsc.c | 37 ++++++++++++++++++++++++++++++--- drivers/net/hyperv/netvsc_drv.c | 14 ++++++++----- 2 files changed, 43 insertions(+), 8 deletions(-)
Signature · 1 line
2.27.0
PATCH v2 1/3 hv_netvsc: Check VF datapath when sending traffic to VF
From: Long Li <[email protected]>
The driver needs to check if the datapath has been switched to VF before sending traffic to VF.
--- drivers/net/hyperv/netvsc_drv.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/net/hyperv/netvsc_drv.c +2 -1
@@ -539,7 +539,8 @@ static int netvsc_xmit(struct sk_buff *skb, struct net_device *net, bool xdp_tx)*/ vf_netdev =rcu_dereference_bh (net_device_ctx->vf_netdev);if (vf_netdev &&netif_running (vf_netdev) && -netif_carrier_ok (vf_netdev) && !netpoll_tx_running (net)) +netif_carrier_ok (vf_netdev) && !netpoll_tx_running (net) && + net_device_ctx->data_path_is_vf)return netvsc_vf_xmit (net, vf_netdev, skb);/* We will atmost need two pages to describe the rndis
Signature · 1 line
2.27.0
PATCH v2 2/3 hv_netvsc: Wait for completion on request SWITCH_DATA_PATH
In reply to Long LiFrom: Long Li <[email protected]>
The completion indicates if NVSP_MSG4_TYPE_SWITCH_DATA_PATH has been processed by the VSP. The traffic is steered to VF or synthetic after we receive this completion.
---
Change from v1:
Fixed warnings from kernel test robot.
drivers/net/hyperv/netvsc.c | 37 ++++++++++++++++++++++++++++++--- drivers/net/hyperv/netvsc_drv.c | 1 - 2 files changed, 34 insertions(+), 4 deletions(-)
drivers/net/hyperv/netvsc.c +34 -3
@@ -37,6 +37,10 @@ void netvsc_switch_datapath(struct net_device *ndev, bool vf)struct netvsc_device *nv_dev =rtnl_dereference (net_device_ctx->nvdev);struct nvsp_message *init_pkt = &nv_dev->channel_init_pkt; +/* Block sending traffic to VF if it's about to be gone */ +if (!vf) + net_device_ctx->data_path_is_vf = vf; +memset (init_pkt,0 ,sizeof (struct nvsp_message )); init_pkt->hdr.msg_type =NVSP_MSG4_TYPE_SWITCH_DATA_PATH ;if (vf)
@@ -50,8 +54,11 @@ void netvsc_switch_datapath(struct net_device *ndev, bool vf)vmbus_sendpacket (dev->channel, init_pkt,sizeof (struct nvsp_message ), -VMBUS_RQST_ID_NO_RESPONSE , -VM_PKT_DATA_INBAND ,0 ); + (unsigned long )init_pkt, +VM_PKT_DATA_INBAND , +VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED ); +wait_for_completion (&nv_dev->channel_init_wait); + net_device_ctx->data_path_is_vf = vf; }/* Worker to setup sub channels on initial setup
@@ -754,8 +761,31 @@ static void netvsc_send_completion(struct net_device *ndev, const struct vmpacket_descriptor *desc, int budget) { -const struct nvsp_message *nvsp_packet =hv_pkt_data (desc); +const struct nvsp_message *nvsp_packet;u32 msglen =hv_pkt_datalen (desc); +struct nvsp_message *pkt_rqst; +u64 cmd_rqst; + +/* First check if this is a VMBUS completion without data payload */ +if (!msglen) { + cmd_rqst =vmbus_request_addr (&incoming_channel->requestor, + (u64 )desc->trans_id); +if (cmd_rqst ==VMBUS_RQST_ERROR ) { +netdev_err (ndev,"Invalid transaction id\n" ); +return ; + } + + pkt_rqst = (struct nvsp_message *)(uintptr_t )cmd_rqst; +switch (pkt_rqst->hdr.msg_type) { +case NVSP_MSG4_TYPE_SWITCH_DATA_PATH : +complete (&net_device->channel_init_wait); +break ; + +default : +netdev_err (ndev,"Unexpected VMBUS completion!!\n" ); + } +return ; + }/* Ensure packet is big enough to read header fields */ if (msglen <sizeof (struct nvsp_message_header )) {
@@ -763,6 +793,7 @@ static void netvsc_send_completion(struct net_device *ndev,return ; } + nvsp_packet =hv_pkt_data (desc); switch (nvsp_packet->hdr.msg_type) { case NVSP_MSG_TYPE_INIT_COMPLETE: if (msglen < sizeof(struct nvsp_message_header) +
drivers/net/hyperv/netvsc_drv.c +0 -1
@@ -2400,7 +2400,6 @@ static int netvsc_vf_changed(struct net_device *vf_netdev)if (net_device_ctx->data_path_is_vf == vf_is_up)return NOTIFY_OK ; - net_device_ctx->data_path_is_vf = vf_is_up;netvsc_switch_datapath (ndev, vf_is_up); netdev_info(ndev, "Data path switched %s VF: %s\n",
Signature · 1 line
2.27.0
PATCH v2 3/3 hv_netvsc: Process NETDEV_GOING_DOWN on VF hot remove
In reply to Long LiFrom: Long Li <[email protected]>
On VF hot remove, NETDEV_GOING_DOWN is sent to notify the VF is about to go down. At this time, the VF is still sending/receiving traffic and we request the VSP to switch datapath.
On completion, the datapath is switched to synthetic and we can proceed with VF hot remove.
--- drivers/net/hyperv/netvsc_drv.c | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-)
drivers/net/hyperv/netvsc_drv.c +7 -3
@@ -2382,12 +2382,15 @@ static int netvsc_register_vf(struct net_device *vf_netdev)* During hibernation, if a VF NIC driver (e.g. mlx5) preserves the network * interface, there is only the CHANGE event and no UP or DOWN event. */ -static int netvsc_vf_changed (struct net_device *vf_netdev) +static int netvsc_vf_changed (struct net_device *vf_netdev,unsigned long event) {struct net_device_context *net_device_ctx;struct netvsc_device *netvsc_dev;struct net_device *ndev; -bool vf_is_up =netif_running (vf_netdev); +bool vf_is_up = false; + +if (event !=NETDEV_GOING_DOWN ) + vf_is_up =netif_running (vf_netdev); ndev =get_netvsc_byref (vf_netdev);if (!ndev)
@@ -2716,7 +2719,8 @@ static int netvsc_netdev_event(struct notifier_block *this,case NETDEV_UP :case NETDEV_DOWN :case NETDEV_CHANGE : -return netvsc_vf_changed (event_dev); +case NETDEV_GOING_DOWN : +return netvsc_vf_changed (event_dev, event);default :return NOTIFY_DONE ; }
Signature · 1 line
2.27.0
Jakub Kicinski
maintainerQuoted 3 lines · On Fri, 8 Jan 2021 16:53:40 -0800 Long Li wrote:
From: Long Li <[email protected]>
This patch set fixes issues with packet loss on VF add/remove.
These patches are for net-next? They just optimize the amount of packet loss on switch, not fix bugs, right?
Quoted 10 lines
Subject: Re: [PATCH v2 0/3]
hv_netvsc: Prevent packet loss during VF add/removeQuoted 3 lines · On Fri, 8 Jan 2021 16:53:40 -0800 Long Li wrote:
From: Long Li <[email protected]>
This patch set fixes issues with packet loss on VF add/remove.
These patches are for net-next? They just optimize the amount of packet loss on switch, not fix bugs, right?
Yes, those patches are for net-next.
They eliminate the packet loss introduced from Linux side during VF changes. They can be seen as optimizations.
This series was applied to netdev/net-next.git (refs/heads/master):
Hello:
This series was applied to netdev/net-next.git (refs/heads/master):
Quoted 10 lines · On Fri, 8 Jan 2021 16:53:40 -0800 you wrote:
From: Long Li <[email protected]>
This patch set fixes issues with packet loss on VF add/remove.
Long Li (3):
hv_netvsc: Check VF datapath when sending traffic to VF
hv_netvsc: Wait for completion on requestSWITCH_DATA_PATH
hv_netvsc: ProcessNETDEV_GOING_DOWNon VF hot remove[...]
Here is the summary with links:
- [v2,1/3] hv_netvsc: Check VF datapath when sending traffic to VF
https://git.kernel.org/netdev/net-next/c/69d25a6cf4ca - [v2,2/3] hv_netvsc: Wait for completion on request SWITCH_DATA_PATH https://git.kernel.org/netdev/net-next/c/8b31f8c982b7 - [v2,3/3] hv_netvsc: Process NETDEV_GOING_DOWN on VF hot remove https://git.kernel.org/netdev/net-next/c/34b06a2eee44
You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html