RE: [PATCH] hv_netvsc: support a new host capability AllowRscDisabledStatus

From: Haiyang Zhang
Date: Thu Jun 29 2023 - 08:44:39 EST




> -----Original Message-----
> From: Shradha Gupta <shradhagupta@xxxxxxxxxxxxxxxxxxx>
> Sent: Thursday, June 29, 2023 5:59 AM
> To: linux-kernel@xxxxxxxxxxxxxxx; linux-hyperv@xxxxxxxxxxxxxxx;
> netdev@xxxxxxxxxxxxxxx
> Cc: Shradha Gupta <shradhagupta@xxxxxxxxxxxxxxxxxxx>; Eric Dumazet
> <edumazet@xxxxxxxxxx>; Jakub Kicinski <kuba@xxxxxxxxxx>; Paolo Abeni
> <pabeni@xxxxxxxxxx>; KY Srinivasan <kys@xxxxxxxxxxxxx>; Haiyang Zhang
> <haiyangz@xxxxxxxxxxxxx>; Wei Liu <wei.liu@xxxxxxxxxx>; Dexuan Cui
> <decui@xxxxxxxxxxxxx>; Long Li <longli@xxxxxxxxxxxxx>; Michael Kelley
> (LINUX) <mikelley@xxxxxxxxxxxxx>; David S. Miller <davem@xxxxxxxxxxxxx>
> Subject: [PATCH] hv_netvsc: support a new host capability
> AllowRscDisabledStatus
>
> A future Azure host update has the potential to change RSC behavior
> in the VMs. To avoid this invisble change, Vswitch will check the
> netvsc version of a VM before sending its RSC capabilities, and will
> always indicate that the host performs RSC if the VM doesn't have an
> updated netvsc driver regardless of the actual host RSC capabilities.
> Netvsc now advertises a new capability: AllowRscDisabledStatus
> The host will check for this capability before sending RSC status,
> and if a VM does not have this capability it will send RSC enabled
> status regardless of host RSC settings
>
> Signed-off-by: Shradha Gupta <shradhagupta@xxxxxxxxxxxxxxxxxxx>
> ---
> drivers/net/hyperv/hyperv_net.h | 3 +++
> drivers/net/hyperv/netvsc.c | 8 ++++++++
> 2 files changed, 11 insertions(+)
>
> diff --git a/drivers/net/hyperv/hyperv_net.h b/drivers/net/hyperv/hyperv_net.h
> index dd5919ec408b..218e0f31dd66 100644
> --- a/drivers/net/hyperv/hyperv_net.h
> +++ b/drivers/net/hyperv/hyperv_net.h
> @@ -572,6 +572,9 @@ struct nvsp_2_vsc_capability {
> u64 teaming:1;
> u64 vsubnetid:1;
> u64 rsc:1;
> + u64 timestamp:1;
> + u64 reliablecorrelationid:1;
> + u64 allowrscdisabledstatus:1;
> };
> };
> } __packed;
> diff --git a/drivers/net/hyperv/netvsc.c b/drivers/net/hyperv/netvsc.c
> index da737d959e81..2eb1e85ba940 100644
> --- a/drivers/net/hyperv/netvsc.c
> +++ b/drivers/net/hyperv/netvsc.c
> @@ -619,6 +619,14 @@ static int negotiate_nvsp_ver(struct hv_device
> *device,
> init_packet->msg.v2_msg.send_ndis_config.mtu = ndev->mtu +
> ETH_HLEN;
> init_packet->msg.v2_msg.send_ndis_config.capability.ieee8021q = 1;
>
> + /* Don't need a version check while setting this bit because if we
> + * have a New VM on an old host, the VM will set the bit but the host
> + * won't check it. If we have an old VM on a new host, the host will
> + * check the bit, see its zero, and it'll know the VM has an
> + * older NetVsc
> + */
> + init_packet-
> >msg.v2_msg.send_ndis_config.capability.allowrscdisabledstatus = 1;

Have you tested on the new host to verify: Before this patch, the host shows
RSC status on, and after this patch the host shows it's off?

Thanks,
- Haiyang