This patch set fixes an issue in SG tuning, and sync
offload settings from synthetic NIC to VF NIC.
Haiyang Zhang (2):
hv_netvsc: hv_netvsc: Allow scatter-gather feature to be tunable
hv_netvsc: Sync offloading features to VF NIC
drivers/net/hyperv/hyperv_net.h | 2 +-
drivers/net/hyperv/netvsc_drv.c | 26 ++++++++++++++++++++++----
drivers/net/hyperv/rndis_filter.c | 1 +
3 files changed, 24 insertions(+), 5 deletions(-)
--
1.8.3.1
In a previous patch, the NETIF_F_SG was missing after the code changes.
That caused the SG feature to be "fixed". This patch includes it into
hw_features, so it is tunable again.
Fixes: 23312a3be999 ("netvsc: negotiate checksum and segmentation parameters")
Signed-off-by: Haiyang Zhang <[email protected]>
---
drivers/net/hyperv/hyperv_net.h | 2 +-
drivers/net/hyperv/netvsc_drv.c | 4 ++--
drivers/net/hyperv/rndis_filter.c | 1 +
3 files changed, 4 insertions(+), 3 deletions(-)
diff --git a/drivers/net/hyperv/hyperv_net.h b/drivers/net/hyperv/hyperv_net.h
index ecc9af0..670ef68 100644
--- a/drivers/net/hyperv/hyperv_net.h
+++ b/drivers/net/hyperv/hyperv_net.h
@@ -822,7 +822,7 @@ struct nvsp_message {
#define NETVSC_SUPPORTED_HW_FEATURES (NETIF_F_RXCSUM | NETIF_F_IP_CSUM | \
NETIF_F_TSO | NETIF_F_IPV6_CSUM | \
- NETIF_F_TSO6 | NETIF_F_LRO)
+ NETIF_F_TSO6 | NETIF_F_LRO | NETIF_F_SG)
#define VRSS_SEND_TAB_SIZE 16 /* must be power of 2 */
#define VRSS_CHANNEL_MAX 64
diff --git a/drivers/net/hyperv/netvsc_drv.c b/drivers/net/hyperv/netvsc_drv.c
index 0a6cd2f..1f1192e 100644
--- a/drivers/net/hyperv/netvsc_drv.c
+++ b/drivers/net/hyperv/netvsc_drv.c
@@ -2313,8 +2313,8 @@ static int netvsc_probe(struct hv_device *dev,
/* hw_features computed in rndis_netdev_set_hwcaps() */
net->features = net->hw_features |
- NETIF_F_HIGHDMA | NETIF_F_SG |
- NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX;
+ NETIF_F_HIGHDMA | NETIF_F_HW_VLAN_CTAG_TX |
+ NETIF_F_HW_VLAN_CTAG_RX;
net->vlan_features = net->features;
netdev_lockdep_set_classes(net);
diff --git a/drivers/net/hyperv/rndis_filter.c b/drivers/net/hyperv/rndis_filter.c
index 317dbe9..abaf815 100644
--- a/drivers/net/hyperv/rndis_filter.c
+++ b/drivers/net/hyperv/rndis_filter.c
@@ -1207,6 +1207,7 @@ static int rndis_netdev_set_hwcaps(struct rndis_device *rndis_device,
/* Compute tx offload settings based on hw capabilities */
net->hw_features |= NETIF_F_RXCSUM;
+ net->hw_features |= NETIF_F_SG;
if ((hwcaps.csum.ip4_txcsum & NDIS_TXCSUM_ALL_TCP4) == NDIS_TXCSUM_ALL_TCP4) {
/* Can checksum TCP */
--
1.8.3.1
VF NIC may go down then come up during host servicing events. This
causes the VF NIC offloading feature settings to roll back to the
defaults. This patch can synchronize features from synthetic NIC to
the VF NIC during ndo_set_features (ethtool -K),
and netvsc_register_vf when VF comes back after host events.
Signed-off-by: Haiyang Zhang <[email protected]>
Cc: Mark Bloch <[email protected]>
---
drivers/net/hyperv/netvsc_drv.c | 22 ++++++++++++++++++++--
1 file changed, 20 insertions(+), 2 deletions(-)
diff --git a/drivers/net/hyperv/netvsc_drv.c b/drivers/net/hyperv/netvsc_drv.c
index 1f1192e..39dddcd 100644
--- a/drivers/net/hyperv/netvsc_drv.c
+++ b/drivers/net/hyperv/netvsc_drv.c
@@ -1785,13 +1785,15 @@ static int netvsc_set_features(struct net_device *ndev,
netdev_features_t change = features ^ ndev->features;
struct net_device_context *ndevctx = netdev_priv(ndev);
struct netvsc_device *nvdev = rtnl_dereference(ndevctx->nvdev);
+ struct net_device *vf_netdev = rtnl_dereference(ndevctx->vf_netdev);
struct ndis_offload_params offloads;
+ int ret = 0;
if (!nvdev || nvdev->destroy)
return -ENODEV;
if (!(change & NETIF_F_LRO))
- return 0;
+ goto syncvf;
memset(&offloads, 0, sizeof(struct ndis_offload_params));
@@ -1803,7 +1805,19 @@ static int netvsc_set_features(struct net_device *ndev,
offloads.rsc_ip_v6 = NDIS_OFFLOAD_PARAMETERS_RSC_DISABLED;
}
- return rndis_filter_set_offload_params(ndev, nvdev, &offloads);
+ ret = rndis_filter_set_offload_params(ndev, nvdev, &offloads);
+
+ if (ret)
+ features ^= NETIF_F_LRO;
+
+syncvf:
+ if (!vf_netdev)
+ return ret;
+
+ vf_netdev->wanted_features = features;
+ netdev_update_features(vf_netdev);
+
+ return ret;
}
static u32 netvsc_get_msglevel(struct net_device *ndev)
@@ -2181,6 +2195,10 @@ static int netvsc_register_vf(struct net_device *vf_netdev)
dev_hold(vf_netdev);
rcu_assign_pointer(net_device_ctx->vf_netdev, vf_netdev);
+
+ vf_netdev->wanted_features = ndev->features;
+ netdev_update_features(vf_netdev);
+
return NOTIFY_OK;
}
--
1.8.3.1
On Fri, 30 Aug 2019 03:45:38 +0000, Haiyang Zhang wrote:
> VF NIC may go down then come up during host servicing events. This
> causes the VF NIC offloading feature settings to roll back to the
> defaults. This patch can synchronize features from synthetic NIC to
> the VF NIC during ndo_set_features (ethtool -K),
> and netvsc_register_vf when VF comes back after host events.
>
> Signed-off-by: Haiyang Zhang <[email protected]>
> Cc: Mark Bloch <[email protected]>
If we want to make this change in behaviour we should change
net_failover at the same time.
On Fri, 30 Aug 2019 03:45:24 +0000, Haiyang Zhang wrote:
> In a previous patch, the NETIF_F_SG was missing after the code changes.
> That caused the SG feature to be "fixed". This patch includes it into
> hw_features, so it is tunable again.
>
> Fixes: 23312a3be999 ("netvsc: negotiate checksum and segmentation parameters")
^
Looks like a tab sneaked in there.
> Signed-off-by: Haiyang Zhang <[email protected]>
> -----Original Message-----
> From: Jakub Kicinski <[email protected]>
> Sent: Friday, August 30, 2019 4:05 PM
> To: Haiyang Zhang <[email protected]>
> Cc: [email protected]; [email protected];
> [email protected]; KY Srinivasan <[email protected]>; Stephen
> Hemminger <[email protected]>; [email protected]; vkuznets
> <[email protected]>; [email protected]; linux-
> [email protected]; Mark Bloch <[email protected]>
> Subject: Re: [PATCH net-next, 2/2] hv_netvsc: Sync offloading features to VF
> NIC
>
> On Fri, 30 Aug 2019 03:45:38 +0000, Haiyang Zhang wrote:
> > VF NIC may go down then come up during host servicing events. This
> > causes the VF NIC offloading feature settings to roll back to the
> > defaults. This patch can synchronize features from synthetic NIC to
> > the VF NIC during ndo_set_features (ethtool -K), and
> > netvsc_register_vf when VF comes back after host events.
> >
> > Signed-off-by: Haiyang Zhang <[email protected]>
> > Cc: Mark Bloch <[email protected]>
>
> If we want to make this change in behaviour we should change net_failover
> at the same time.
I will check net_failover. Thanks.
> -----Original Message-----
> From: Jakub Kicinski <[email protected]>
> Sent: Friday, August 30, 2019 7:05 PM
> To: Haiyang Zhang <[email protected]>
> Cc: [email protected]; [email protected];
> [email protected]; KY Srinivasan <[email protected]>; Stephen
> Hemminger <[email protected]>; [email protected]; vkuznets
> <[email protected]>; [email protected]; linux-
> [email protected]; Mark Bloch <[email protected]>
> Subject: Re: [PATCH net-next, 2/2] hv_netvsc: Sync offloading features to VF
> NIC
>
> On Fri, 30 Aug 2019 03:45:38 +0000, Haiyang Zhang wrote:
> > VF NIC may go down then come up during host servicing events. This
> > causes the VF NIC offloading feature settings to roll back to the
> > defaults. This patch can synchronize features from synthetic NIC to
> > the VF NIC during ndo_set_features (ethtool -K), and
> > netvsc_register_vf when VF comes back after host events.
> >
> > Signed-off-by: Haiyang Zhang <[email protected]>
> > Cc: Mark Bloch <[email protected]>
>
> If we want to make this change in behaviour we should change net_failover
> at the same time.
After checking the net_failover, I found it's for virtio based SRIOV, and very
different from what we did for Hyper-V based SRIOV.
We let the netvsc driver acts as both the synthetic (PV) driver and the transparent
bonding master for the VF NIC. But net_failover acts as a master device on top
of both virtio PV NIC, and VF NIC. And the net_failover doesn't implemented
operations, like ndo_set_features.
So the code change for our netvsc driver cannot be applied to net_failover driver.
I will re-submit my two patches (fixing the extra tab in the 1st one as you pointed
out). Thanks!
- Haiyang
On Thu, 5 Sep 2019 23:07:32 +0000, Haiyang Zhang wrote:
> > On Fri, 30 Aug 2019 03:45:38 +0000, Haiyang Zhang wrote:
> > > VF NIC may go down then come up during host servicing events. This
> > > causes the VF NIC offloading feature settings to roll back to the
> > > defaults. This patch can synchronize features from synthetic NIC to
> > > the VF NIC during ndo_set_features (ethtool -K), and
> > > netvsc_register_vf when VF comes back after host events.
> > >
> > > Signed-off-by: Haiyang Zhang <[email protected]>
> > > Cc: Mark Bloch <[email protected]>
> >
> > If we want to make this change in behaviour we should change net_failover
> > at the same time.
>
> After checking the net_failover, I found it's for virtio based SRIOV, and very
> different from what we did for Hyper-V based SRIOV.
>
> We let the netvsc driver acts as both the synthetic (PV) driver and the transparent
> bonding master for the VF NIC. But net_failover acts as a master device on top
> of both virtio PV NIC, and VF NIC. And the net_failover doesn't implemented
> operations, like ndo_set_features.
> So the code change for our netvsc driver cannot be applied to net_failover driver.
>
> I will re-submit my two patches (fixing the extra tab in the 1st one as you pointed
> out). Thanks!
I think it stands to reason that two modules which implement the same
functionality behave the same.