[v1] virtio-net: avoid unnecessary checksum calculation on guest RX

[PATCH net-next] virtio-net: avoid unnecessary checksum calculation on guest RX

Posted by Jon Kohler 6 days, 5 hours ago

Commit a2fb4bc4e2a6 ("net: implement virtio helpers to handle UDP
GSO tunneling.") inadvertently altered checksum offload behavior
for guests not using UDP GSO tunneling.

Before, tun_put_user called tun_vnet_hdr_from_skb, which passed
has_data_valid = true to virtio_net_hdr_from_skb.

After, tun_put_user began calling tun_vnet_hdr_tnl_from_skb instead,
which passes has_data_valid = false into both call sites.

This caused virtio hdr flags to not include VIRTIO_NET_HDR_F_DATA_VALID
for SKBs where skb->ip_summed == CHECKSUM_UNNECESSARY. As a result,
guests are forced to recalculate checksums unnecessarily.

Restore the previous behavior by ensuring has_data_valid = true is
passed in the !tnl_gso_type case.

Cc: Paolo Abeni <pabeni@redhat.com>
Fixes: a2fb4bc4e2a6 ("net: implement virtio helpers to handle UDP GSO tunneling.")
Signed-off-by: Jon Kohler <jon@nutanix.com>
---
 include/linux/virtio_net.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/include/linux/virtio_net.h b/include/linux/virtio_net.h
index b673c31569f3..570c6dd1666d 100644
--- a/include/linux/virtio_net.h
+++ b/include/linux/virtio_net.h
@@ -394,7 +394,7 @@ virtio_net_hdr_tnl_from_skb(const struct sk_buff *skb,
 	tnl_gso_type = skb_shinfo(skb)->gso_type & (SKB_GSO_UDP_TUNNEL |
 						    SKB_GSO_UDP_TUNNEL_CSUM);
 	if (!tnl_gso_type)
-		return virtio_net_hdr_from_skb(skb, hdr, little_endian, false,
+		return virtio_net_hdr_from_skb(skb, hdr, little_endian, true,
 					       vlan_hlen);
 
 	/* Tunnel support not negotiated but skb ask for it. */
-- 
2.43.0

Re: [PATCH net-next] virtio-net: avoid unnecessary checksum calculation on guest RX

Posted by Paolo Abeni 6 days, 5 hours ago

CC netdev

On 11/25/25 6:51 PM, Jon Kohler wrote:
> Commit a2fb4bc4e2a6 ("net: implement virtio helpers to handle UDP
> GSO tunneling.") inadvertently altered checksum offload behavior
> for guests not using UDP GSO tunneling.
> 
> Before, tun_put_user called tun_vnet_hdr_from_skb, which passed
> has_data_valid = true to virtio_net_hdr_from_skb.
> 
> After, tun_put_user began calling tun_vnet_hdr_tnl_from_skb instead,
> which passes has_data_valid = false into both call sites.
> 
> This caused virtio hdr flags to not include VIRTIO_NET_HDR_F_DATA_VALID
> for SKBs where skb->ip_summed == CHECKSUM_UNNECESSARY. As a result,
> guests are forced to recalculate checksums unnecessarily.
> 
> Restore the previous behavior by ensuring has_data_valid = true is
> passed in the !tnl_gso_type case.
> 
> Cc: Paolo Abeni <pabeni@redhat.com>
> Fixes: a2fb4bc4e2a6 ("net: implement virtio helpers to handle UDP GSO tunneling.")
> Signed-off-by: Jon Kohler <jon@nutanix.com>
> ---
>  include/linux/virtio_net.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/include/linux/virtio_net.h b/include/linux/virtio_net.h
> index b673c31569f3..570c6dd1666d 100644
> --- a/include/linux/virtio_net.h
> +++ b/include/linux/virtio_net.h
> @@ -394,7 +394,7 @@ virtio_net_hdr_tnl_from_skb(const struct sk_buff *skb,
>  	tnl_gso_type = skb_shinfo(skb)->gso_type & (SKB_GSO_UDP_TUNNEL |
>  						    SKB_GSO_UDP_TUNNEL_CSUM);
>  	if (!tnl_gso_type)
> -		return virtio_net_hdr_from_skb(skb, hdr, little_endian, false,
> +		return virtio_net_hdr_from_skb(skb, hdr, little_endian, true,
>  					       vlan_hlen);
>  
>  	/* Tunnel support not negotiated but skb ask for it. */

virtio_net_hdr_tnl_from_skb() is used also by the virtio_net driver,
which in turn must not use VIRTIO_NET_HDR_F_DATA_VALID on tx.

I think you need to add another argument to
virtio_net_hdr_tnl_from_skb(), or possibly implement a separate helper
to take care of csum offload - the symmetric of
virtio_net_handle_csum_offload().

Also you need to CC netdev, otherwise the patch will not be processed by
patchwork.

/P

Re: [PATCH net-next] virtio-net: avoid unnecessary checksum calculation on guest RX

Posted by Jon Kohler 6 days, 3 hours ago


> On Nov 25, 2025, at 12:57 PM, Paolo Abeni <pabeni@redhat.com> wrote:
> 
> CC netdev

Thats odd, I used git send-email --to-cmd='./scripts/get_maintainer.pl,
but it looks like in MAINTAINERS, that only would have hit
VIRTIO CORE AND NET DRIVERS, which does not include netdev@

Should that have ?
L: netdev@vger.kernel.org <mailto:netdev@vger.kernel.org>

Said another way, should all changes to include/linux/virtio_net.h
be cc’d to netdev DL?

I suspect the answer is yes, I’ll send a patch for that in the
interest of not having this issue again :)

> 
> On 11/25/25 6:51 PM, Jon Kohler wrote:
>> Commit a2fb4bc4e2a6 ("net: implement virtio helpers to handle UDP
>> GSO tunneling.") inadvertently altered checksum offload behavior
>> for guests not using UDP GSO tunneling.
>> 
>> Before, tun_put_user called tun_vnet_hdr_from_skb, which passed
>> has_data_valid = true to virtio_net_hdr_from_skb.
>> 
>> After, tun_put_user began calling tun_vnet_hdr_tnl_from_skb instead,
>> which passes has_data_valid = false into both call sites.
>> 
>> This caused virtio hdr flags to not include VIRTIO_NET_HDR_F_DATA_VALID
>> for SKBs where skb->ip_summed == CHECKSUM_UNNECESSARY. As a result,
>> guests are forced to recalculate checksums unnecessarily.
>> 
>> Restore the previous behavior by ensuring has_data_valid = true is
>> passed in the !tnl_gso_type case.
>> 
>> Cc: Paolo Abeni <pabeni@redhat.com>
>> Fixes: a2fb4bc4e2a6 ("net: implement virtio helpers to handle UDP GSO tunneling.")
>> Signed-off-by: Jon Kohler <jon@nutanix.com>
>> ---
>> include/linux/virtio_net.h | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>> 
>> diff --git a/include/linux/virtio_net.h b/include/linux/virtio_net.h
>> index b673c31569f3..570c6dd1666d 100644
>> --- a/include/linux/virtio_net.h
>> +++ b/include/linux/virtio_net.h
>> @@ -394,7 +394,7 @@ virtio_net_hdr_tnl_from_skb(const struct sk_buff *skb,
>> tnl_gso_type = skb_shinfo(skb)->gso_type & (SKB_GSO_UDP_TUNNEL |
>>     SKB_GSO_UDP_TUNNEL_CSUM);
>> if (!tnl_gso_type)
>> - return virtio_net_hdr_from_skb(skb, hdr, little_endian, false,
>> + return virtio_net_hdr_from_skb(skb, hdr, little_endian, true,
>>        vlan_hlen);
>> 
>> /* Tunnel support not negotiated but skb ask for it. */
> 
> virtio_net_hdr_tnl_from_skb() is used also by the virtio_net driver,
> which in turn must not use VIRTIO_NET_HDR_F_DATA_VALID on tx.

Ah! Good eye, I’ll see what trouble I can get into and send a v2
> 
> I think you need to add another argument to
> virtio_net_hdr_tnl_from_skb(), or possibly implement a separate helper
> to take care of csum offload - the symmetric of
> virtio_net_handle_csum_offload().
> 
> Also you need to CC netdev, otherwise the patch will not be processed by
> patchwork.
> 
> /P

No problems on cc’ing netdev, just didn’t realize this one header didn’t
auto cc the list. Will keep an eye on that, and happy to send a patch to
MAINTAINERS file for discussion.

Re: [PATCH net-next] virtio-net: avoid unnecessary checksum calculation on guest RX

Posted by Paolo Abeni 5 days, 12 hours ago

On 11/25/25 9:00 PM, Jon Kohler wrote:
>> On Nov 25, 2025, at 12:57 PM, Paolo Abeni <pabeni@redhat.com> wrote:
>>
>> CC netdev
> 
> Thats odd, I used git send-email --to-cmd='./scripts/get_maintainer.pl,
> but it looks like in MAINTAINERS, that only would have hit
> VIRTIO CORE AND NET DRIVERS, 

?!? not here:

./scripts/get_maintainer.pl drivers/net/virtio_net.c
"Michael S. Tsirkin" <mst@redhat.com> (maintainer:VIRTIO CORE AND NET
DRIVERS)
Jason Wang <jasowang@redhat.com> (maintainer:VIRTIO CORE AND NET DRIVERS)
Xuan Zhuo <xuanzhuo@linux.alibaba.com> (reviewer:VIRTIO CORE AND NET
DRIVERS)
"Eugenio Pérez" <eperezma@redhat.com> (reviewer:VIRTIO CORE AND NET DRIVERS)
Andrew Lunn <andrew+netdev@lunn.ch> (maintainer:NETWORKING DRIVERS)
"David S. Miller" <davem@davemloft.net> (maintainer:NETWORKING DRIVERS)
Eric Dumazet <edumazet@google.com> (maintainer:NETWORKING DRIVERS)
Jakub Kicinski <kuba@kernel.org> (maintainer:NETWORKING DRIVERS)
Paolo Abeni <pabeni@redhat.com> (maintainer:NETWORKING DRIVERS)
virtualization@lists.linux.dev (open list:VIRTIO CORE AND NET DRIVERS)
netdev@vger.kernel.org (open list:NETWORKING DRIVERS)
linux-kernel@vger.kernel.org (open list)

The "NETWORKING DRIVER" entry should catch even virtio_net. Something
odd in your setup?!?

BTW, this is a bit too late, but you should wait at least 24h before
reposting on netdev:

https://elixir.bootlin.com/linux/v6.17.9/source/Documentation/process/maintainer-netdev.rst#L422

/P

Re: [PATCH net-next] virtio-net: avoid unnecessary checksum calculation on guest RX

Posted by Jon Kohler 5 days, 4 hours ago


> On Nov 26, 2025, at 6:16 AM, Paolo Abeni <pabeni@redhat.com> wrote:
> 
> On 11/25/25 9:00 PM, Jon Kohler wrote:
>>> On Nov 25, 2025, at 12:57 PM, Paolo Abeni <pabeni@redhat.com> wrote:
>>> 
>>> CC netdev
>> 
>> Thats odd, I used git send-email --to-cmd='./scripts/get_maintainer.pl,
>> but it looks like in MAINTAINERS, that only would have hit
>> VIRTIO CORE AND NET DRIVERS,
> 
> ?!? not here:
> 
> ./scripts/get_maintainer.pl drivers/net/virtio_net.c
> "Michael S. Tsirkin" <mst@redhat.com> (maintainer:VIRTIO CORE AND NET
> DRIVERS)
> Jason Wang <jasowang@redhat.com> (maintainer:VIRTIO CORE AND NET DRIVERS)
> Xuan Zhuo <xuanzhuo@linux.alibaba.com> (reviewer:VIRTIO CORE AND NET
> DRIVERS)
> "Eugenio Pérez" <eperezma@redhat.com> (reviewer:VIRTIO CORE AND NET DRIVERS)
> Andrew Lunn <andrew+netdev@lunn.ch> (maintainer:NETWORKING DRIVERS)
> "David S. Miller" <davem@davemloft.net> (maintainer:NETWORKING DRIVERS)
> Eric Dumazet <edumazet@google.com> (maintainer:NETWORKING DRIVERS)
> Jakub Kicinski <kuba@kernel.org> (maintainer:NETWORKING DRIVERS)
> Paolo Abeni <pabeni@redhat.com> (maintainer:NETWORKING DRIVERS)
> virtualization@lists.linux.dev (open list:VIRTIO CORE AND NET DRIVERS)
> netdev@vger.kernel.org (open list:NETWORKING DRIVERS)
> linux-kernel@vger.kernel.org (open list)
> 
> The "NETWORKING DRIVER" entry should catch even virtio_net. Something
> odd in your setup?!?

Literally all I’ve got is a clone of net-next and using get_maintainer.pl
but yea the issue is the header-only change. I have a knack for falling
exactly into these sorts of process holes :) It’ll be a good cleanup
anyhow

> 
> BTW, this is a bit too late, but you should wait at least 24h before
> reposting on netdev:

Thanks for the guidance again, I’ll be sure to pace
things out if/when I find myself in this situation
again! I appreciate the help

Re: [PATCH net-next] virtio-net: avoid unnecessary checksum calculation on guest RX

Posted by Michael S. Tsirkin 5 days, 11 hours ago

On Wed, Nov 26, 2025 at 12:16:56PM +0100, Paolo Abeni wrote:
> On 11/25/25 9:00 PM, Jon Kohler wrote:
> >> On Nov 25, 2025, at 12:57 PM, Paolo Abeni <pabeni@redhat.com> wrote:
> >>
> >> CC netdev
> > 
> > Thats odd, I used git send-email --to-cmd='./scripts/get_maintainer.pl,
> > but it looks like in MAINTAINERS, that only would have hit
> > VIRTIO CORE AND NET DRIVERS, 
> 
> ?!? not here:
> 
> ./scripts/get_maintainer.pl drivers/net/virtio_net.c

yes but not include/linux/virtio_net.h


> "Michael S. Tsirkin" <mst@redhat.com> (maintainer:VIRTIO CORE AND NET
> DRIVERS)
> Jason Wang <jasowang@redhat.com> (maintainer:VIRTIO CORE AND NET DRIVERS)
> Xuan Zhuo <xuanzhuo@linux.alibaba.com> (reviewer:VIRTIO CORE AND NET
> DRIVERS)
> "Eugenio Pérez" <eperezma@redhat.com> (reviewer:VIRTIO CORE AND NET DRIVERS)
> Andrew Lunn <andrew+netdev@lunn.ch> (maintainer:NETWORKING DRIVERS)
> "David S. Miller" <davem@davemloft.net> (maintainer:NETWORKING DRIVERS)
> Eric Dumazet <edumazet@google.com> (maintainer:NETWORKING DRIVERS)
> Jakub Kicinski <kuba@kernel.org> (maintainer:NETWORKING DRIVERS)
> Paolo Abeni <pabeni@redhat.com> (maintainer:NETWORKING DRIVERS)
> virtualization@lists.linux.dev (open list:VIRTIO CORE AND NET DRIVERS)
> netdev@vger.kernel.org (open list:NETWORKING DRIVERS)
> linux-kernel@vger.kernel.org (open list)
> 
> The "NETWORKING DRIVER" entry should catch even virtio_net. Something
> odd in your setup?!?
> 
> BTW, this is a bit too late, but you should wait at least 24h before
> reposting on netdev:
> 
> https://elixir.bootlin.com/linux/v6.17.9/source/Documentation/process/maintainer-netdev.rst#L422
> 
> /P

Re: [PATCH net-next] virtio-net: avoid unnecessary checksum calculation on guest RX

Posted by Jon Kohler 5 days, 4 hours ago


> On Nov 26, 2025, at 7:30 AM, Michael S. Tsirkin <mst@redhat.com> wrote:
> 
> On Wed, Nov 26, 2025 at 12:16:56PM +0100, Paolo Abeni wrote:
>> On 11/25/25 9:00 PM, Jon Kohler wrote:
>>>> On Nov 25, 2025, at 12:57 PM, Paolo Abeni <pabeni@redhat.com> wrote:
>>>> 
>>>> CC netdev
>>> 
>>> Thats odd, I used git send-email --to-cmd='./scripts/get_maintainer.pl,
>>> but it looks like in MAINTAINERS, that only would have hit
>>> VIRTIO CORE AND NET DRIVERS,
>> 
>> ?!? not here:
>> 
>> ./scripts/get_maintainer.pl drivers/net/virtio_net.c
> 
> yes but not include/linux/virtio_net.h

Indeed! This fell thru a very small gap, will be fixed now in
the MAINTAINERS update I sent.

Small side note: the virtio_blk.c is a duplicate within core,
it isn’t hurting anything, but could be cleaned up. I left it
as-is for now.

Re: [PATCH net-next] virtio-net: avoid unnecessary checksum calculation on guest RX

Posted by Michael S. Tsirkin 5 days, 23 hours ago

On Tue, Nov 25, 2025 at 08:00:55PM +0000, Jon Kohler wrote:
> 
> 
> > On Nov 25, 2025, at 12:57 PM, Paolo Abeni <pabeni@redhat.com> wrote:
> > 
> > CC netdev
> 
> Thats odd, I used git send-email --to-cmd='./scripts/get_maintainer.pl,
> but it looks like in MAINTAINERS, that only would have hit
> VIRTIO CORE AND NET DRIVERS, which does not include netdev@
> 
> Should that have ?
> L: netdev@vger.kernel.org <mailto:netdev@vger.kernel.org>
> 
> Said another way, should all changes to include/linux/virtio_net.h
> be cc’d to netdev DL?
> 
> I suspect the answer is yes, I’ll send a patch for that in the
> interest of not having this issue again :)

I think yes. But only virtio net not core. I guess we should
split net from core then.

> > 
> > On 11/25/25 6:51 PM, Jon Kohler wrote:
> >> Commit a2fb4bc4e2a6 ("net: implement virtio helpers to handle UDP
> >> GSO tunneling.") inadvertently altered checksum offload behavior
> >> for guests not using UDP GSO tunneling.
> >> 
> >> Before, tun_put_user called tun_vnet_hdr_from_skb, which passed
> >> has_data_valid = true to virtio_net_hdr_from_skb.
> >> 
> >> After, tun_put_user began calling tun_vnet_hdr_tnl_from_skb instead,
> >> which passes has_data_valid = false into both call sites.
> >> 
> >> This caused virtio hdr flags to not include VIRTIO_NET_HDR_F_DATA_VALID
> >> for SKBs where skb->ip_summed == CHECKSUM_UNNECESSARY. As a result,
> >> guests are forced to recalculate checksums unnecessarily.
> >> 
> >> Restore the previous behavior by ensuring has_data_valid = true is
> >> passed in the !tnl_gso_type case.
> >> 
> >> Cc: Paolo Abeni <pabeni@redhat.com>
> >> Fixes: a2fb4bc4e2a6 ("net: implement virtio helpers to handle UDP GSO tunneling.")
> >> Signed-off-by: Jon Kohler <jon@nutanix.com>
> >> ---
> >> include/linux/virtio_net.h | 2 +-
> >> 1 file changed, 1 insertion(+), 1 deletion(-)
> >> 
> >> diff --git a/include/linux/virtio_net.h b/include/linux/virtio_net.h
> >> index b673c31569f3..570c6dd1666d 100644
> >> --- a/include/linux/virtio_net.h
> >> +++ b/include/linux/virtio_net.h
> >> @@ -394,7 +394,7 @@ virtio_net_hdr_tnl_from_skb(const struct sk_buff *skb,
> >> tnl_gso_type = skb_shinfo(skb)->gso_type & (SKB_GSO_UDP_TUNNEL |
> >>     SKB_GSO_UDP_TUNNEL_CSUM);
> >> if (!tnl_gso_type)
> >> - return virtio_net_hdr_from_skb(skb, hdr, little_endian, false,
> >> + return virtio_net_hdr_from_skb(skb, hdr, little_endian, true,
> >>        vlan_hlen);
> >> 
> >> /* Tunnel support not negotiated but skb ask for it. */
> > 
> > virtio_net_hdr_tnl_from_skb() is used also by the virtio_net driver,
> > which in turn must not use VIRTIO_NET_HDR_F_DATA_VALID on tx.
> 
> Ah! Good eye, I’ll see what trouble I can get into and send a v2
> > 
> > I think you need to add another argument to
> > virtio_net_hdr_tnl_from_skb(), or possibly implement a separate helper
> > to take care of csum offload - the symmetric of
> > virtio_net_handle_csum_offload().
> > 
> > Also you need to CC netdev, otherwise the patch will not be processed by
> > patchwork.
> > 
> > /P
> 
> No problems on cc’ing netdev, just didn’t realize this one header didn’t
> auto cc the list. Will keep an eye on that, and happy to send a patch to
> MAINTAINERS file for discussion.

Re: [PATCH net-next] virtio-net: avoid unnecessary checksum calculation on guest RX

Posted by Paolo Abeni 6 days, 5 hours ago

On 11/25/25 6:57 PM, Paolo Abeni wrote:
> On 11/25/25 6:51 PM, Jon Kohler wrote:
>> Commit a2fb4bc4e2a6 ("net: implement virtio helpers to handle UDP
>> GSO tunneling.") inadvertently altered checksum offload behavior
>> for guests not using UDP GSO tunneling.
>>
>> Before, tun_put_user called tun_vnet_hdr_from_skb, which passed
>> has_data_valid = true to virtio_net_hdr_from_skb.
>>
>> After, tun_put_user began calling tun_vnet_hdr_tnl_from_skb instead,
>> which passes has_data_valid = false into both call sites.
>>
>> This caused virtio hdr flags to not include VIRTIO_NET_HDR_F_DATA_VALID
>> for SKBs where skb->ip_summed == CHECKSUM_UNNECESSARY. As a result,
>> guests are forced to recalculate checksums unnecessarily.
>>
>> Restore the previous behavior by ensuring has_data_valid = true is
>> passed in the !tnl_gso_type case.
>>
>> Cc: Paolo Abeni <pabeni@redhat.com>
>> Fixes: a2fb4bc4e2a6 ("net: implement virtio helpers to handle UDP GSO tunneling.")
>> Signed-off-by: Jon Kohler <jon@nutanix.com>
>> ---
>>  include/linux/virtio_net.h | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/include/linux/virtio_net.h b/include/linux/virtio_net.h
>> index b673c31569f3..570c6dd1666d 100644
>> --- a/include/linux/virtio_net.h
>> +++ b/include/linux/virtio_net.h
>> @@ -394,7 +394,7 @@ virtio_net_hdr_tnl_from_skb(const struct sk_buff *skb,
>>  	tnl_gso_type = skb_shinfo(skb)->gso_type & (SKB_GSO_UDP_TUNNEL |
>>  						    SKB_GSO_UDP_TUNNEL_CSUM);
>>  	if (!tnl_gso_type)
>> -		return virtio_net_hdr_from_skb(skb, hdr, little_endian, false,
>> +		return virtio_net_hdr_from_skb(skb, hdr, little_endian, true,
>>  					       vlan_hlen);
>>  
>>  	/* Tunnel support not negotiated but skb ask for it. */
> 
> virtio_net_hdr_tnl_from_skb() is used also by the virtio_net driver,
> which in turn must not use VIRTIO_NET_HDR_F_DATA_VALID on tx.
> 
> I think you need to add another argument to
> virtio_net_hdr_tnl_from_skb(), or possibly implement a separate helper
> to take care of csum offload - the symmetric of
> virtio_net_handle_csum_offload().
> 
> Also you need to CC netdev, otherwise the patch will not be processed by
> patchwork.

whoops... I almost forgot... this is 'net' material, the subj prefix
should be adjusted accordingly.

Thanks,

Paolo