[RFC PATCH 0/5] TUN/VirtioNet USO features support.

[RFC PATCH 0/5] TUN/VirtioNet USO features support.

[Andrew Melnychenko]

Added new offloads for TUN devices TUN_F_USO4 and TUN_F_USO6.
Technically they enable NETIF_F_GSO_UDP_L4
(and only if USO4 & USO6 are set simultaneously).
It allows to transmission of large UDP packets.

Different features USO4 and USO6 are required for qemu where Windows guests can
enable disable USO receives for IPv4 and IPv6 separately.
On the other side, Linux can't really differentiate USO4 and USO6, for now.
For now, to enable USO for TUN it requires enabling USO4 and USO6 together.
In the future, there would be a mechanism to control UDP_L4 GSO separately.

Test it WIP Qemu https://github.com/daynix/qemu/tree/Dev_USOv2

New types for VirtioNet already on mailing:
https://lists.oasis-open.org/archives/virtio-comment/202110/msg00010.html

Also, there is a known issue with transmitting packages between two guests.
Without hacks with skb's GSO - packages are still segmented on the host's postrouting.

Andrew Melnychenko (5):
  uapi/linux/if_tun.h: Added new ioctl for tun/tap.
  driver/net/tun: Added features for USO.
  uapi/linux/virtio_net.h: Added USO types.
  linux/virtio_net.h: Added Support for GSO_UDP_L4 offload.
  drivers/net/virtio_net.c: Added USO support.

 drivers/net/tap.c               | 18 ++++++++++++++++--
 drivers/net/tun.c               | 15 ++++++++++++++-
 drivers/net/virtio_net.c        | 22 ++++++++++++++++++----
 include/linux/virtio_net.h      | 11 +++++++++++
 include/uapi/linux/if_tun.h     |  3 +++
 include/uapi/linux/virtio_net.h |  4 ++++
 6 files changed, 66 insertions(+), 7 deletions(-)

-- 
2.34.1

[RFC PATCH 1/5] uapi/linux/if_tun.h: Added new ioctl for tun/tap.

[Andrew Melnychenko]

Added TUNGETSUPPORTEDOFFLOADS that should allow
to get bits of supported offloads.
Added 2 additional offlloads for USO(IPv4 & IPv6).
Separate offloads are required for Windows VM guests,
g.e. Windows may set USO rx only for IPv4.

Signed-off-by: Andrew Melnychenko <andrew@daynix.com>
---
 include/uapi/linux/if_tun.h | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/include/uapi/linux/if_tun.h b/include/uapi/linux/if_tun.h
index 454ae31b93c7..07680fae6e18 100644
--- a/include/uapi/linux/if_tun.h
+++ b/include/uapi/linux/if_tun.h
@@ -61,6 +61,7 @@
 #define TUNSETFILTEREBPF _IOR('T', 225, int)
 #define TUNSETCARRIER _IOW('T', 226, int)
 #define TUNGETDEVNETNS _IO('T', 227)
+#define TUNGETSUPPORTEDOFFLOADS _IOR('T', 228, unsigned int)
 
 /* TUNSETIFF ifr flags */
 #define IFF_TUN		0x0001
@@ -88,6 +89,8 @@
 #define TUN_F_TSO6	0x04	/* I can handle TSO for IPv6 packets */
 #define TUN_F_TSO_ECN	0x08	/* I can handle TSO with ECN bits. */
 #define TUN_F_UFO	0x10	/* I can handle UFO packets */
+#define TUN_F_USO4	0x20	/* I can handle USO for IPv4 packets */
+#define TUN_F_USO6	0x40	/* I can handle USO for IPv6 packets */
 
 /* Protocol info prepended to the packets (when IFF_NO_PI is not set) */
 #define TUN_PKT_STRIP	0x0001
-- 
2.34.1

Re: [RFC PATCH 1/5] uapi/linux/if_tun.h: Added new ioctl for tun/tap.

[Jason Wang]

在 2022/1/25 下午4:46, Andrew Melnychenko 写道:
> Added TUNGETSUPPORTEDOFFLOADS that should allow
> to get bits of supported offloads.


So we don't use dedicated ioctls in the past, instead, we just probing 
by checking the return value of TUNSETOFFLOADS.

E.g qemu has the following codes:

int tap_probe_has_ufo(int fd)
{
     unsigned offload;

     offload = TUN_F_CSUM | TUN_F_UFO;

     if (ioctl(fd, TUNSETOFFLOAD, offload) < 0)
         return 0;

     return 1;
}

Any reason we can't keep using that?

Thanks


> Added 2 additional offlloads for USO(IPv4 & IPv6).
> Separate offloads are required for Windows VM guests,
> g.e. Windows may set USO rx only for IPv4.
>
> Signed-off-by: Andrew Melnychenko <andrew@daynix.com>
> ---
>   include/uapi/linux/if_tun.h | 3 +++
>   1 file changed, 3 insertions(+)
>
> diff --git a/include/uapi/linux/if_tun.h b/include/uapi/linux/if_tun.h
> index 454ae31b93c7..07680fae6e18 100644
> --- a/include/uapi/linux/if_tun.h
> +++ b/include/uapi/linux/if_tun.h
> @@ -61,6 +61,7 @@
>   #define TUNSETFILTEREBPF _IOR('T', 225, int)
>   #define TUNSETCARRIER _IOW('T', 226, int)
>   #define TUNGETDEVNETNS _IO('T', 227)
> +#define TUNGETSUPPORTEDOFFLOADS _IOR('T', 228, unsigned int)
>   
>   /* TUNSETIFF ifr flags */
>   #define IFF_TUN		0x0001
> @@ -88,6 +89,8 @@
>   #define TUN_F_TSO6	0x04	/* I can handle TSO for IPv6 packets */
>   #define TUN_F_TSO_ECN	0x08	/* I can handle TSO with ECN bits. */
>   #define TUN_F_UFO	0x10	/* I can handle UFO packets */
> +#define TUN_F_USO4	0x20	/* I can handle USO for IPv4 packets */
> +#define TUN_F_USO6	0x40	/* I can handle USO for IPv6 packets */
>   
>   /* Protocol info prepended to the packets (when IFF_NO_PI is not set) */
>   #define TUN_PKT_STRIP	0x0001

Re: [RFC PATCH 1/5] uapi/linux/if_tun.h: Added new ioctl for tun/tap.

[Andrew Melnichenko]

Hi all,

On Wed, Feb 9, 2022 at 6:26 AM Jason Wang <jasowang@redhat.com> wrote:
>
>
> 在 2022/1/25 下午4:46, Andrew Melnychenko 写道:
> > Added TUNGETSUPPORTEDOFFLOADS that should allow
> > to get bits of supported offloads.
>
>
> So we don't use dedicated ioctls in the past, instead, we just probing
> by checking the return value of TUNSETOFFLOADS.
>
> E.g qemu has the following codes:
>
> int tap_probe_has_ufo(int fd)
> {
>      unsigned offload;
>
>      offload = TUN_F_CSUM | TUN_F_UFO;
>
>      if (ioctl(fd, TUNSETOFFLOAD, offload) < 0)
>          return 0;
>
>      return 1;
> }
>
> Any reason we can't keep using that?
>
> Thanks
>

Well, even in this example. To check the ufo feature, we trying to set it.
What if we don't need to "enable" UFO and/or do not change its state?
I think it's a good idea to have the ability to get supported offloads
without changing device behavior.

>
> > Added 2 additional offlloads for USO(IPv4 & IPv6).
> > Separate offloads are required for Windows VM guests,
> > g.e. Windows may set USO rx only for IPv4.
> >
> > Signed-off-by: Andrew Melnychenko <andrew@daynix.com>
> > ---
> >   include/uapi/linux/if_tun.h | 3 +++
> >   1 file changed, 3 insertions(+)
> >
> > diff --git a/include/uapi/linux/if_tun.h b/include/uapi/linux/if_tun.h
> > index 454ae31b93c7..07680fae6e18 100644
> > --- a/include/uapi/linux/if_tun.h
> > +++ b/include/uapi/linux/if_tun.h
> > @@ -61,6 +61,7 @@
> >   #define TUNSETFILTEREBPF _IOR('T', 225, int)
> >   #define TUNSETCARRIER _IOW('T', 226, int)
> >   #define TUNGETDEVNETNS _IO('T', 227)
> > +#define TUNGETSUPPORTEDOFFLOADS _IOR('T', 228, unsigned int)
> >
> >   /* TUNSETIFF ifr flags */
> >   #define IFF_TUN             0x0001
> > @@ -88,6 +89,8 @@
> >   #define TUN_F_TSO6  0x04    /* I can handle TSO for IPv6 packets */
> >   #define TUN_F_TSO_ECN       0x08    /* I can handle TSO with ECN bits. */
> >   #define TUN_F_UFO   0x10    /* I can handle UFO packets */
> > +#define TUN_F_USO4   0x20    /* I can handle USO for IPv4 packets */
> > +#define TUN_F_USO6   0x40    /* I can handle USO for IPv6 packets */
> >
> >   /* Protocol info prepended to the packets (when IFF_NO_PI is not set) */
> >   #define TUN_PKT_STRIP       0x0001
>

Re: [RFC PATCH 1/5] uapi/linux/if_tun.h: Added new ioctl for tun/tap.

[Jason Wang]

On Tue, Feb 22, 2022 at 9:28 PM Andrew Melnichenko <andrew@daynix.com> wrote:
>
> Hi all,
>
> On Wed, Feb 9, 2022 at 6:26 AM Jason Wang <jasowang@redhat.com> wrote:
> >
> >
> > 在 2022/1/25 下午4:46, Andrew Melnychenko 写道:
> > > Added TUNGETSUPPORTEDOFFLOADS that should allow
> > > to get bits of supported offloads.
> >
> >
> > So we don't use dedicated ioctls in the past, instead, we just probing
> > by checking the return value of TUNSETOFFLOADS.
> >
> > E.g qemu has the following codes:
> >
> > int tap_probe_has_ufo(int fd)
> > {
> >      unsigned offload;
> >
> >      offload = TUN_F_CSUM | TUN_F_UFO;
> >
> >      if (ioctl(fd, TUNSETOFFLOAD, offload) < 0)
> >          return 0;
> >
> >      return 1;
> > }
> >
> > Any reason we can't keep using that?
> >
> > Thanks
> >
>
> Well, even in this example. To check the ufo feature, we trying to set it.
> What if we don't need to "enable" UFO and/or do not change its state?

So at least Qemu doesn't have such a requirement since during the
probe the virtual networking backend is not even started.

> I think it's a good idea to have the ability to get supported offloads
> without changing device behavior.

Do you see a real user for this?

Btw, we still need to probe this new ioctl anyway.

Thanks

>
> >
> > > Added 2 additional offlloads for USO(IPv4 & IPv6).
> > > Separate offloads are required for Windows VM guests,
> > > g.e. Windows may set USO rx only for IPv4.
> > >
> > > Signed-off-by: Andrew Melnychenko <andrew@daynix.com>
> > > ---
> > >   include/uapi/linux/if_tun.h | 3 +++
> > >   1 file changed, 3 insertions(+)
> > >
> > > diff --git a/include/uapi/linux/if_tun.h b/include/uapi/linux/if_tun.h
> > > index 454ae31b93c7..07680fae6e18 100644
> > > --- a/include/uapi/linux/if_tun.h
> > > +++ b/include/uapi/linux/if_tun.h
> > > @@ -61,6 +61,7 @@
> > >   #define TUNSETFILTEREBPF _IOR('T', 225, int)
> > >   #define TUNSETCARRIER _IOW('T', 226, int)
> > >   #define TUNGETDEVNETNS _IO('T', 227)
> > > +#define TUNGETSUPPORTEDOFFLOADS _IOR('T', 228, unsigned int)
> > >
> > >   /* TUNSETIFF ifr flags */
> > >   #define IFF_TUN             0x0001
> > > @@ -88,6 +89,8 @@
> > >   #define TUN_F_TSO6  0x04    /* I can handle TSO for IPv6 packets */
> > >   #define TUN_F_TSO_ECN       0x08    /* I can handle TSO with ECN bits. */
> > >   #define TUN_F_UFO   0x10    /* I can handle UFO packets */
> > > +#define TUN_F_USO4   0x20    /* I can handle USO for IPv4 packets */
> > > +#define TUN_F_USO6   0x40    /* I can handle USO for IPv6 packets */
> > >
> > >   /* Protocol info prepended to the packets (when IFF_NO_PI is not set) */
> > >   #define TUN_PKT_STRIP       0x0001
> >
>

Re: [RFC PATCH 1/5] uapi/linux/if_tun.h: Added new ioctl for tun/tap.

[Yuri Benditovich]

Hi Jason,
We agree that the same can be done also using the old way, i.e. try to
set specific offload - if failed, probably it is not supported.
We think this is a little not scalable and we suggest adding the ioctl
that will allow us to query allo the supported features in a single
call.
We think this will make QEMU code more simple also in future.
Do I understand correctly that you suggest to skip this new ioctl and
use the old way of query for this (USO) feature and all future
extensions?

Thanks


On Wed, Feb 23, 2022 at 5:53 AM Jason Wang <jasowang@redhat.com> wrote:
>
> On Tue, Feb 22, 2022 at 9:28 PM Andrew Melnichenko <andrew@daynix.com> wrote:
> >
> > Hi all,
> >
> > On Wed, Feb 9, 2022 at 6:26 AM Jason Wang <jasowang@redhat.com> wrote:
> > >
> > >
> > > 在 2022/1/25 下午4:46, Andrew Melnychenko 写道:
> > > > Added TUNGETSUPPORTEDOFFLOADS that should allow
> > > > to get bits of supported offloads.
> > >
> > >
> > > So we don't use dedicated ioctls in the past, instead, we just probing
> > > by checking the return value of TUNSETOFFLOADS.
> > >
> > > E.g qemu has the following codes:
> > >
> > > int tap_probe_has_ufo(int fd)
> > > {
> > >      unsigned offload;
> > >
> > >      offload = TUN_F_CSUM | TUN_F_UFO;
> > >
> > >      if (ioctl(fd, TUNSETOFFLOAD, offload) < 0)
> > >          return 0;
> > >
> > >      return 1;
> > > }
> > >
> > > Any reason we can't keep using that?
> > >
> > > Thanks
> > >
> >
> > Well, even in this example. To check the ufo feature, we trying to set it.
> > What if we don't need to "enable" UFO and/or do not change its state?
>
> So at least Qemu doesn't have such a requirement since during the
> probe the virtual networking backend is not even started.
>
> > I think it's a good idea to have the ability to get supported offloads
> > without changing device behavior.
>
> Do you see a real user for this?
>
> Btw, we still need to probe this new ioctl anyway.
>
> Thanks
>
> >
> > >
> > > > Added 2 additional offlloads for USO(IPv4 & IPv6).
> > > > Separate offloads are required for Windows VM guests,
> > > > g.e. Windows may set USO rx only for IPv4.
> > > >
> > > > Signed-off-by: Andrew Melnychenko <andrew@daynix.com>
> > > > ---
> > > >   include/uapi/linux/if_tun.h | 3 +++
> > > >   1 file changed, 3 insertions(+)
> > > >
> > > > diff --git a/include/uapi/linux/if_tun.h b/include/uapi/linux/if_tun.h
> > > > index 454ae31b93c7..07680fae6e18 100644
> > > > --- a/include/uapi/linux/if_tun.h
> > > > +++ b/include/uapi/linux/if_tun.h
> > > > @@ -61,6 +61,7 @@
> > > >   #define TUNSETFILTEREBPF _IOR('T', 225, int)
> > > >   #define TUNSETCARRIER _IOW('T', 226, int)
> > > >   #define TUNGETDEVNETNS _IO('T', 227)
> > > > +#define TUNGETSUPPORTEDOFFLOADS _IOR('T', 228, unsigned int)
> > > >
> > > >   /* TUNSETIFF ifr flags */
> > > >   #define IFF_TUN             0x0001
> > > > @@ -88,6 +89,8 @@
> > > >   #define TUN_F_TSO6  0x04    /* I can handle TSO for IPv6 packets */
> > > >   #define TUN_F_TSO_ECN       0x08    /* I can handle TSO with ECN bits. */
> > > >   #define TUN_F_UFO   0x10    /* I can handle UFO packets */
> > > > +#define TUN_F_USO4   0x20    /* I can handle USO for IPv4 packets */
> > > > +#define TUN_F_USO6   0x40    /* I can handle USO for IPv6 packets */
> > > >
> > > >   /* Protocol info prepended to the packets (when IFF_NO_PI is not set) */
> > > >   #define TUN_PKT_STRIP       0x0001
> > >
> >
>

Re: [RFC PATCH 1/5] uapi/linux/if_tun.h: Added new ioctl for tun/tap.

[Jason Wang]

On Wed, Feb 23, 2022 at 9:31 PM Yuri Benditovich
<yuri.benditovich@daynix.com> wrote:
>
> Hi Jason,
> We agree that the same can be done also using the old way, i.e. try to
> set specific offload - if failed, probably it is not supported.
> We think this is a little not scalable and we suggest adding the ioctl
> that will allow us to query allo the supported features in a single
> call.

Possibly but then you need some kind of probing. E.g we need endup
with probing TUNGETSUPPORTEDOFFLOADS iotctl itself.

> We think this will make QEMU code more simple also in future.

We can discuss this when qemu patches were sent.

> Do I understand correctly that you suggest to skip this new ioctl and
> use the old way of query for this (USO) feature and all future
> extensions?

Yes, since it's not a must. And we can do the TUNGETSUPPORTEDOFFLOADS
in a separate series.

Thanks

>
> Thanks
>
>
> On Wed, Feb 23, 2022 at 5:53 AM Jason Wang <jasowang@redhat.com> wrote:
> >
> > On Tue, Feb 22, 2022 at 9:28 PM Andrew Melnichenko <andrew@daynix.com> wrote:
> > >
> > > Hi all,
> > >
> > > On Wed, Feb 9, 2022 at 6:26 AM Jason Wang <jasowang@redhat.com> wrote:
> > > >
> > > >
> > > > 在 2022/1/25 下午4:46, Andrew Melnychenko 写道:
> > > > > Added TUNGETSUPPORTEDOFFLOADS that should allow
> > > > > to get bits of supported offloads.
> > > >
> > > >
> > > > So we don't use dedicated ioctls in the past, instead, we just probing
> > > > by checking the return value of TUNSETOFFLOADS.
> > > >
> > > > E.g qemu has the following codes:
> > > >
> > > > int tap_probe_has_ufo(int fd)
> > > > {
> > > >      unsigned offload;
> > > >
> > > >      offload = TUN_F_CSUM | TUN_F_UFO;
> > > >
> > > >      if (ioctl(fd, TUNSETOFFLOAD, offload) < 0)
> > > >          return 0;
> > > >
> > > >      return 1;
> > > > }
> > > >
> > > > Any reason we can't keep using that?
> > > >
> > > > Thanks
> > > >
> > >
> > > Well, even in this example. To check the ufo feature, we trying to set it.
> > > What if we don't need to "enable" UFO and/or do not change its state?
> >
> > So at least Qemu doesn't have such a requirement since during the
> > probe the virtual networking backend is not even started.
> >
> > > I think it's a good idea to have the ability to get supported offloads
> > > without changing device behavior.
> >
> > Do you see a real user for this?
> >
> > Btw, we still need to probe this new ioctl anyway.
> >
> > Thanks
> >
> > >
> > > >
> > > > > Added 2 additional offlloads for USO(IPv4 & IPv6).
> > > > > Separate offloads are required for Windows VM guests,
> > > > > g.e. Windows may set USO rx only for IPv4.
> > > > >
> > > > > Signed-off-by: Andrew Melnychenko <andrew@daynix.com>
> > > > > ---
> > > > >   include/uapi/linux/if_tun.h | 3 +++
> > > > >   1 file changed, 3 insertions(+)
> > > > >
> > > > > diff --git a/include/uapi/linux/if_tun.h b/include/uapi/linux/if_tun.h
> > > > > index 454ae31b93c7..07680fae6e18 100644
> > > > > --- a/include/uapi/linux/if_tun.h
> > > > > +++ b/include/uapi/linux/if_tun.h
> > > > > @@ -61,6 +61,7 @@
> > > > >   #define TUNSETFILTEREBPF _IOR('T', 225, int)
> > > > >   #define TUNSETCARRIER _IOW('T', 226, int)
> > > > >   #define TUNGETDEVNETNS _IO('T', 227)
> > > > > +#define TUNGETSUPPORTEDOFFLOADS _IOR('T', 228, unsigned int)
> > > > >
> > > > >   /* TUNSETIFF ifr flags */
> > > > >   #define IFF_TUN             0x0001
> > > > > @@ -88,6 +89,8 @@
> > > > >   #define TUN_F_TSO6  0x04    /* I can handle TSO for IPv6 packets */
> > > > >   #define TUN_F_TSO_ECN       0x08    /* I can handle TSO with ECN bits. */
> > > > >   #define TUN_F_UFO   0x10    /* I can handle UFO packets */
> > > > > +#define TUN_F_USO4   0x20    /* I can handle USO for IPv4 packets */
> > > > > +#define TUN_F_USO6   0x40    /* I can handle USO for IPv6 packets */
> > > > >
> > > > >   /* Protocol info prepended to the packets (when IFF_NO_PI is not set) */
> > > > >   #define TUN_PKT_STRIP       0x0001
> > > >
> > >
> >
>

[RFC PATCH 2/5] driver/net/tun: Added features for USO.

[Andrew Melnychenko]

Added support for USO4 and USO6, also added code for new ioctl TUNGETSUPPORTEDOFFLOADS.
For now, to "enable" USO, it's required to set both USO4 and USO6 simultaneously.
USO enables NETIF_F_GSO_UDP_L4.

Signed-off-by: Andrew Melnychenko <andrew@daynix.com>
---
 drivers/net/tap.c | 18 ++++++++++++++++--
 drivers/net/tun.c | 15 ++++++++++++++-
 2 files changed, 30 insertions(+), 3 deletions(-)

diff --git a/drivers/net/tap.c b/drivers/net/tap.c
index 8e3a28ba6b28..82d742ba78b1 100644
--- a/drivers/net/tap.c
+++ b/drivers/net/tap.c
@@ -940,6 +940,10 @@ static int set_offload(struct tap_queue *q, unsigned long arg)
 			if (arg & TUN_F_TSO6)
 				feature_mask |= NETIF_F_TSO6;
 		}
+
+		/* TODO: for now USO4 and USO6 should work simultaneously */
+		if (arg & (TUN_F_USO4 | TUN_F_USO6) == (TUN_F_USO4 | TUN_F_USO6))
+			features |= NETIF_F_GSO_UDP_L4;
 	}
 
 	/* tun/tap driver inverts the usage for TSO offloads, where
@@ -950,7 +954,8 @@ static int set_offload(struct tap_queue *q, unsigned long arg)
 	 * When user space turns off TSO, we turn off GSO/LRO so that
 	 * user-space will not receive TSO frames.
 	 */
-	if (feature_mask & (NETIF_F_TSO | NETIF_F_TSO6))
+	if (feature_mask & (NETIF_F_TSO | NETIF_F_TSO6) ||
+	    feature_mask & (TUN_F_USO4 | TUN_F_USO6) == (TUN_F_USO4 | TUN_F_USO6))
 		features |= RX_OFFLOADS;
 	else
 		features &= ~RX_OFFLOADS;
@@ -979,6 +984,7 @@ static long tap_ioctl(struct file *file, unsigned int cmd,
 	unsigned short u;
 	int __user *sp = argp;
 	struct sockaddr sa;
+	unsigned int supported_offloads;
 	int s;
 	int ret;
 
@@ -1074,7 +1080,8 @@ static long tap_ioctl(struct file *file, unsigned int cmd,
 	case TUNSETOFFLOAD:
 		/* let the user check for future flags */
 		if (arg & ~(TUN_F_CSUM | TUN_F_TSO4 | TUN_F_TSO6 |
-			    TUN_F_TSO_ECN | TUN_F_UFO))
+			    TUN_F_TSO_ECN | TUN_F_UFO |
+			    TUN_F_USO4 | TUN_F_USO6))
 			return -EINVAL;
 
 		rtnl_lock();
@@ -1082,6 +1089,13 @@ static long tap_ioctl(struct file *file, unsigned int cmd,
 		rtnl_unlock();
 		return ret;
 
+	case TUNGETSUPPORTEDOFFLOADS:
+		supported_offloads = TUN_F_CSUM | TUN_F_TSO4 | TUN_F_TSO6 |
+						TUN_F_TSO_ECN | TUN_F_UFO | TUN_F_USO4 | TUN_F_USO6;
+		if (copy_to_user(&arg, &supported_offloads, sizeof(supported_offloads)))
+			return -EFAULT;
+		return 0;
+
 	case SIOCGIFHWADDR:
 		rtnl_lock();
 		tap = tap_get_tap_dev(q);
diff --git a/drivers/net/tun.c b/drivers/net/tun.c
index fed85447701a..4f2105d1e6f1 100644
--- a/drivers/net/tun.c
+++ b/drivers/net/tun.c
@@ -185,7 +185,7 @@ struct tun_struct {
 	struct net_device	*dev;
 	netdev_features_t	set_features;
 #define TUN_USER_FEATURES (NETIF_F_HW_CSUM|NETIF_F_TSO_ECN|NETIF_F_TSO| \
-			  NETIF_F_TSO6)
+			  NETIF_F_TSO6 | NETIF_F_GSO_UDP_L4)
 
 	int			align;
 	int			vnet_hdr_sz;
@@ -2821,6 +2821,12 @@ static int set_offload(struct tun_struct *tun, unsigned long arg)
 		}
 
 		arg &= ~TUN_F_UFO;
+
+		/* TODO: for now USO4 and USO6 should work simultaneously */
+		if (arg & TUN_F_USO4 && arg & TUN_F_USO6) {
+			features |= NETIF_F_GSO_UDP_L4;
+			arg &= ~(TUN_F_USO4 | TUN_F_USO6);
+		}
 	}
 
 	/* This gives the user a way to test for new features in future by
@@ -2991,6 +2997,7 @@ static long __tun_chr_ioctl(struct file *file, unsigned int cmd,
 	int sndbuf;
 	int vnet_hdr_sz;
 	int le;
+	unsigned int supported_offloads;
 	int ret;
 	bool do_notify = false;
 
@@ -3154,6 +3161,12 @@ static long __tun_chr_ioctl(struct file *file, unsigned int cmd,
 	case TUNSETOFFLOAD:
 		ret = set_offload(tun, arg);
 		break;
+	case TUNGETSUPPORTEDOFFLOADS:
+		supported_offloads = TUN_F_CSUM | TUN_F_TSO4 | TUN_F_TSO6 |
+				TUN_F_TSO_ECN | TUN_F_UFO | TUN_F_USO4 | TUN_F_USO6;
+		if (copy_to_user(&arg, &supported_offloads, sizeof(supported_offloads)))
+			ret = -EFAULT;
+		break;
 
 	case TUNSETTXFILTER:
 		/* Can be set only for TAPs */
-- 
2.34.1

Re: [RFC PATCH 2/5] driver/net/tun: Added features for USO.

[Jason Wang]

在 2022/1/25 下午4:46, Andrew Melnychenko 写道:
> Added support for USO4 and USO6, also added code for new ioctl TUNGETSUPPORTEDOFFLOADS.
> For now, to "enable" USO, it's required to set both USO4 and USO6 simultaneously.
> USO enables NETIF_F_GSO_UDP_L4.
>
> Signed-off-by: Andrew Melnychenko <andrew@daynix.com>
> ---
>   drivers/net/tap.c | 18 ++++++++++++++++--
>   drivers/net/tun.c | 15 ++++++++++++++-
>   2 files changed, 30 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/net/tap.c b/drivers/net/tap.c
> index 8e3a28ba6b28..82d742ba78b1 100644
> --- a/drivers/net/tap.c
> +++ b/drivers/net/tap.c
> @@ -940,6 +940,10 @@ static int set_offload(struct tap_queue *q, unsigned long arg)
>   			if (arg & TUN_F_TSO6)
>   				feature_mask |= NETIF_F_TSO6;
>   		}
> +
> +		/* TODO: for now USO4 and USO6 should work simultaneously */
> +		if (arg & (TUN_F_USO4 | TUN_F_USO6) == (TUN_F_USO4 | TUN_F_USO6))
> +			features |= NETIF_F_GSO_UDP_L4;


If kernel doesn't want to split the GSO_UDP features, I wonder how much 
value to keep separated features for TUN and virtio.

Thanks


>   	}
>   
>   	/* tun/tap driver inverts the usage for TSO offloads, where
> @@ -950,7 +954,8 @@ static int set_offload(struct tap_queue *q, unsigned long arg)
>   	 * When user space turns off TSO, we turn off GSO/LRO so that
>   	 * user-space will not receive TSO frames.
>   	 */
> -	if (feature_mask & (NETIF_F_TSO | NETIF_F_TSO6))
> +	if (feature_mask & (NETIF_F_TSO | NETIF_F_TSO6) ||
> +	    feature_mask & (TUN_F_USO4 | TUN_F_USO6) == (TUN_F_USO4 | TUN_F_USO6))
>   		features |= RX_OFFLOADS;
>   	else
>   		features &= ~RX_OFFLOADS;
> @@ -979,6 +984,7 @@ static long tap_ioctl(struct file *file, unsigned int cmd,
>   	unsigned short u;
>   	int __user *sp = argp;
>   	struct sockaddr sa;
> +	unsigned int supported_offloads;
>   	int s;
>   	int ret;
>   
> @@ -1074,7 +1080,8 @@ static long tap_ioctl(struct file *file, unsigned int cmd,
>   	case TUNSETOFFLOAD:
>   		/* let the user check for future flags */
>   		if (arg & ~(TUN_F_CSUM | TUN_F_TSO4 | TUN_F_TSO6 |
> -			    TUN_F_TSO_ECN | TUN_F_UFO))
> +			    TUN_F_TSO_ECN | TUN_F_UFO |
> +			    TUN_F_USO4 | TUN_F_USO6))
>   			return -EINVAL;
>   
>   		rtnl_lock();
> @@ -1082,6 +1089,13 @@ static long tap_ioctl(struct file *file, unsigned int cmd,
>   		rtnl_unlock();
>   		return ret;
>   
> +	case TUNGETSUPPORTEDOFFLOADS:
> +		supported_offloads = TUN_F_CSUM | TUN_F_TSO4 | TUN_F_TSO6 |
> +						TUN_F_TSO_ECN | TUN_F_UFO | TUN_F_USO4 | TUN_F_USO6;
> +		if (copy_to_user(&arg, &supported_offloads, sizeof(supported_offloads)))
> +			return -EFAULT;
> +		return 0;
> +
>   	case SIOCGIFHWADDR:
>   		rtnl_lock();
>   		tap = tap_get_tap_dev(q);
> diff --git a/drivers/net/tun.c b/drivers/net/tun.c
> index fed85447701a..4f2105d1e6f1 100644
> --- a/drivers/net/tun.c
> +++ b/drivers/net/tun.c
> @@ -185,7 +185,7 @@ struct tun_struct {
>   	struct net_device	*dev;
>   	netdev_features_t	set_features;
>   #define TUN_USER_FEATURES (NETIF_F_HW_CSUM|NETIF_F_TSO_ECN|NETIF_F_TSO| \
> -			  NETIF_F_TSO6)
> +			  NETIF_F_TSO6 | NETIF_F_GSO_UDP_L4)
>   
>   	int			align;
>   	int			vnet_hdr_sz;
> @@ -2821,6 +2821,12 @@ static int set_offload(struct tun_struct *tun, unsigned long arg)
>   		}
>   
>   		arg &= ~TUN_F_UFO;
> +
> +		/* TODO: for now USO4 and USO6 should work simultaneously */
> +		if (arg & TUN_F_USO4 && arg & TUN_F_USO6) {
> +			features |= NETIF_F_GSO_UDP_L4;
> +			arg &= ~(TUN_F_USO4 | TUN_F_USO6);
> +		}
>   	}
>   
>   	/* This gives the user a way to test for new features in future by
> @@ -2991,6 +2997,7 @@ static long __tun_chr_ioctl(struct file *file, unsigned int cmd,
>   	int sndbuf;
>   	int vnet_hdr_sz;
>   	int le;
> +	unsigned int supported_offloads;
>   	int ret;
>   	bool do_notify = false;
>   
> @@ -3154,6 +3161,12 @@ static long __tun_chr_ioctl(struct file *file, unsigned int cmd,
>   	case TUNSETOFFLOAD:
>   		ret = set_offload(tun, arg);
>   		break;
> +	case TUNGETSUPPORTEDOFFLOADS:
> +		supported_offloads = TUN_F_CSUM | TUN_F_TSO4 | TUN_F_TSO6 |
> +				TUN_F_TSO_ECN | TUN_F_UFO | TUN_F_USO4 | TUN_F_USO6;
> +		if (copy_to_user(&arg, &supported_offloads, sizeof(supported_offloads)))
> +			ret = -EFAULT;
> +		break;
>   
>   	case TUNSETTXFILTER:
>   		/* Can be set only for TAPs */

Re: [RFC PATCH 2/5] driver/net/tun: Added features for USO.

[Andrew Melnichenko]

On Wed, Feb 9, 2022 at 6:39 AM Jason Wang <jasowang@redhat.com> wrote:
>
>
> 在 2022/1/25 下午4:46, Andrew Melnychenko 写道:
> > Added support for USO4 and USO6, also added code for new ioctl TUNGETSUPPORTEDOFFLOADS.
> > For now, to "enable" USO, it's required to set both USO4 and USO6 simultaneously.
> > USO enables NETIF_F_GSO_UDP_L4.
> >
> > Signed-off-by: Andrew Melnychenko <andrew@daynix.com>
> > ---
> >   drivers/net/tap.c | 18 ++++++++++++++++--
> >   drivers/net/tun.c | 15 ++++++++++++++-
> >   2 files changed, 30 insertions(+), 3 deletions(-)
> >
> > diff --git a/drivers/net/tap.c b/drivers/net/tap.c
> > index 8e3a28ba6b28..82d742ba78b1 100644
> > --- a/drivers/net/tap.c
> > +++ b/drivers/net/tap.c
> > @@ -940,6 +940,10 @@ static int set_offload(struct tap_queue *q, unsigned long arg)
> >                       if (arg & TUN_F_TSO6)
> >                               feature_mask |= NETIF_F_TSO6;
> >               }
> > +
> > +             /* TODO: for now USO4 and USO6 should work simultaneously */
> > +             if (arg & (TUN_F_USO4 | TUN_F_USO6) == (TUN_F_USO4 | TUN_F_USO6))
> > +                     features |= NETIF_F_GSO_UDP_L4;
>
>
> If kernel doesn't want to split the GSO_UDP features, I wonder how much
> value to keep separated features for TUN and virtio.
>
> Thanks
>

It's important for Windows guests that may request USO receive only
for IPv4 or IPv6.
Or there is possible to implement one feature and change its
"meanings" when "split" happens.
I think it's a good idea to implement an interface for iUSO4/USO6 and
do it right away.

>
> >       }
> >
> >       /* tun/tap driver inverts the usage for TSO offloads, where
> > @@ -950,7 +954,8 @@ static int set_offload(struct tap_queue *q, unsigned long arg)
> >        * When user space turns off TSO, we turn off GSO/LRO so that
> >        * user-space will not receive TSO frames.
> >        */
> > -     if (feature_mask & (NETIF_F_TSO | NETIF_F_TSO6))
> > +     if (feature_mask & (NETIF_F_TSO | NETIF_F_TSO6) ||
> > +         feature_mask & (TUN_F_USO4 | TUN_F_USO6) == (TUN_F_USO4 | TUN_F_USO6))
> >               features |= RX_OFFLOADS;
> >       else
> >               features &= ~RX_OFFLOADS;
> > @@ -979,6 +984,7 @@ static long tap_ioctl(struct file *file, unsigned int cmd,
> >       unsigned short u;
> >       int __user *sp = argp;
> >       struct sockaddr sa;
> > +     unsigned int supported_offloads;
> >       int s;
> >       int ret;
> >
> > @@ -1074,7 +1080,8 @@ static long tap_ioctl(struct file *file, unsigned int cmd,
> >       case TUNSETOFFLOAD:
> >               /* let the user check for future flags */
> >               if (arg & ~(TUN_F_CSUM | TUN_F_TSO4 | TUN_F_TSO6 |
> > -                         TUN_F_TSO_ECN | TUN_F_UFO))
> > +                         TUN_F_TSO_ECN | TUN_F_UFO |
> > +                         TUN_F_USO4 | TUN_F_USO6))
> >                       return -EINVAL;
> >
> >               rtnl_lock();
> > @@ -1082,6 +1089,13 @@ static long tap_ioctl(struct file *file, unsigned int cmd,
> >               rtnl_unlock();
> >               return ret;
> >
> > +     case TUNGETSUPPORTEDOFFLOADS:
> > +             supported_offloads = TUN_F_CSUM | TUN_F_TSO4 | TUN_F_TSO6 |
> > +                                             TUN_F_TSO_ECN | TUN_F_UFO | TUN_F_USO4 | TUN_F_USO6;
> > +             if (copy_to_user(&arg, &supported_offloads, sizeof(supported_offloads)))
> > +                     return -EFAULT;
> > +             return 0;
> > +
> >       case SIOCGIFHWADDR:
> >               rtnl_lock();
> >               tap = tap_get_tap_dev(q);
> > diff --git a/drivers/net/tun.c b/drivers/net/tun.c
> > index fed85447701a..4f2105d1e6f1 100644
> > --- a/drivers/net/tun.c
> > +++ b/drivers/net/tun.c
> > @@ -185,7 +185,7 @@ struct tun_struct {
> >       struct net_device       *dev;
> >       netdev_features_t       set_features;
> >   #define TUN_USER_FEATURES (NETIF_F_HW_CSUM|NETIF_F_TSO_ECN|NETIF_F_TSO| \
> > -                       NETIF_F_TSO6)
> > +                       NETIF_F_TSO6 | NETIF_F_GSO_UDP_L4)
> >
> >       int                     align;
> >       int                     vnet_hdr_sz;
> > @@ -2821,6 +2821,12 @@ static int set_offload(struct tun_struct *tun, unsigned long arg)
> >               }
> >
> >               arg &= ~TUN_F_UFO;
> > +
> > +             /* TODO: for now USO4 and USO6 should work simultaneously */
> > +             if (arg & TUN_F_USO4 && arg & TUN_F_USO6) {
> > +                     features |= NETIF_F_GSO_UDP_L4;
> > +                     arg &= ~(TUN_F_USO4 | TUN_F_USO6);
> > +             }
> >       }
> >
> >       /* This gives the user a way to test for new features in future by
> > @@ -2991,6 +2997,7 @@ static long __tun_chr_ioctl(struct file *file, unsigned int cmd,
> >       int sndbuf;
> >       int vnet_hdr_sz;
> >       int le;
> > +     unsigned int supported_offloads;
> >       int ret;
> >       bool do_notify = false;
> >
> > @@ -3154,6 +3161,12 @@ static long __tun_chr_ioctl(struct file *file, unsigned int cmd,
> >       case TUNSETOFFLOAD:
> >               ret = set_offload(tun, arg);
> >               break;
> > +     case TUNGETSUPPORTEDOFFLOADS:
> > +             supported_offloads = TUN_F_CSUM | TUN_F_TSO4 | TUN_F_TSO6 |
> > +                             TUN_F_TSO_ECN | TUN_F_UFO | TUN_F_USO4 | TUN_F_USO6;
> > +             if (copy_to_user(&arg, &supported_offloads, sizeof(supported_offloads)))
> > +                     ret = -EFAULT;
> > +             break;
> >
> >       case TUNSETTXFILTER:
> >               /* Can be set only for TAPs */
>

[RFC PATCH 3/5] uapi/linux/virtio_net.h: Added USO types.

[Andrew Melnychenko]

Added new GSO type for USO: VIRTIO_NET_HDR_GSO_UDP_L4.
Feature VIRTIO_NET_F_HOST_USO allows to enable NETIF_F_GSO_UDP_L4.
Separated VIRTIO_NET_F_GUEST_USO4 & VIRTIO_NET_F_GUEST_USO6 features
required for Windows guests.

Signed-off-by: Andrew Melnychenko <andrew@daynix.com>
---
 include/uapi/linux/virtio_net.h | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/include/uapi/linux/virtio_net.h b/include/uapi/linux/virtio_net.h
index 3f55a4215f11..620addc5767b 100644
--- a/include/uapi/linux/virtio_net.h
+++ b/include/uapi/linux/virtio_net.h
@@ -56,6 +56,9 @@
 #define VIRTIO_NET_F_MQ	22	/* Device supports Receive Flow
 					 * Steering */
 #define VIRTIO_NET_F_CTRL_MAC_ADDR 23	/* Set MAC address */
+#define VIRTIO_NET_F_GUEST_USO4	54	/* Guest can handle USOv4 in. */
+#define VIRTIO_NET_F_GUEST_USO6	55	/* Guest can handle USOv6 in. */
+#define VIRTIO_NET_F_HOST_USO	56	/* Host can handle USO in. */
 
 #define VIRTIO_NET_F_HASH_REPORT  57	/* Supports hash report */
 #define VIRTIO_NET_F_RSS	  60	/* Supports RSS RX steering */
@@ -130,6 +133,7 @@ struct virtio_net_hdr_v1 {
 #define VIRTIO_NET_HDR_GSO_TCPV4	1	/* GSO frame, IPv4 TCP (TSO) */
 #define VIRTIO_NET_HDR_GSO_UDP		3	/* GSO frame, IPv4 UDP (UFO) */
 #define VIRTIO_NET_HDR_GSO_TCPV6	4	/* GSO frame, IPv6 TCP */
+#define VIRTIO_NET_HDR_GSO_UDP_L4	5	/* GSO frame, IPv4 & IPv6 UDP (USO) */
 #define VIRTIO_NET_HDR_GSO_ECN		0x80	/* TCP has ECN set */
 	__u8 gso_type;
 	__virtio16 hdr_len;	/* Ethernet + IP + tcp/udp hdrs */
-- 
2.34.1

Re: [RFC PATCH 3/5] uapi/linux/virtio_net.h: Added USO types.

[Jason Wang]

在 2022/1/25 下午4:47, Andrew Melnychenko 写道:
> Added new GSO type for USO: VIRTIO_NET_HDR_GSO_UDP_L4.
> Feature VIRTIO_NET_F_HOST_USO allows to enable NETIF_F_GSO_UDP_L4.
> Separated VIRTIO_NET_F_GUEST_USO4 & VIRTIO_NET_F_GUEST_USO6 features
> required for Windows guests.
>
> Signed-off-by: Andrew Melnychenko <andrew@daynix.com>
> ---
>   include/uapi/linux/virtio_net.h | 4 ++++
>   1 file changed, 4 insertions(+)
>
> diff --git a/include/uapi/linux/virtio_net.h b/include/uapi/linux/virtio_net.h
> index 3f55a4215f11..620addc5767b 100644
> --- a/include/uapi/linux/virtio_net.h
> +++ b/include/uapi/linux/virtio_net.h
> @@ -56,6 +56,9 @@
>   #define VIRTIO_NET_F_MQ	22	/* Device supports Receive Flow
>   					 * Steering */
>   #define VIRTIO_NET_F_CTRL_MAC_ADDR 23	/* Set MAC address */
> +#define VIRTIO_NET_F_GUEST_USO4	54	/* Guest can handle USOv4 in. */
> +#define VIRTIO_NET_F_GUEST_USO6	55	/* Guest can handle USOv6 in. */
> +#define VIRTIO_NET_F_HOST_USO	56	/* Host can handle USO in. */


I think it's better to be consistent here. Either we split in both guest 
and host or not.

Thanks


>   
>   #define VIRTIO_NET_F_HASH_REPORT  57	/* Supports hash report */
>   #define VIRTIO_NET_F_RSS	  60	/* Supports RSS RX steering */
> @@ -130,6 +133,7 @@ struct virtio_net_hdr_v1 {
>   #define VIRTIO_NET_HDR_GSO_TCPV4	1	/* GSO frame, IPv4 TCP (TSO) */
>   #define VIRTIO_NET_HDR_GSO_UDP		3	/* GSO frame, IPv4 UDP (UFO) */
>   #define VIRTIO_NET_HDR_GSO_TCPV6	4	/* GSO frame, IPv6 TCP */
> +#define VIRTIO_NET_HDR_GSO_UDP_L4	5	/* GSO frame, IPv4 & IPv6 UDP (USO) */
>   #define VIRTIO_NET_HDR_GSO_ECN		0x80	/* TCP has ECN set */
>   	__u8 gso_type;
>   	__virtio16 hdr_len;	/* Ethernet + IP + tcp/udp hdrs */

Re: [RFC PATCH 3/5] uapi/linux/virtio_net.h: Added USO types.

[Andrew Melnichenko]

Hi all,



On Wed, Feb 9, 2022 at 6:41 AM Jason Wang <jasowang@redhat.com> wrote:
>
>
> 在 2022/1/25 下午4:47, Andrew Melnychenko 写道:
> > Added new GSO type for USO: VIRTIO_NET_HDR_GSO_UDP_L4.
> > Feature VIRTIO_NET_F_HOST_USO allows to enable NETIF_F_GSO_UDP_L4.
> > Separated VIRTIO_NET_F_GUEST_USO4 & VIRTIO_NET_F_GUEST_USO6 features
> > required for Windows guests.
> >
> > Signed-off-by: Andrew Melnychenko <andrew@daynix.com>
> > ---
> >   include/uapi/linux/virtio_net.h | 4 ++++
> >   1 file changed, 4 insertions(+)
> >
> > diff --git a/include/uapi/linux/virtio_net.h b/include/uapi/linux/virtio_net.h
> > index 3f55a4215f11..620addc5767b 100644
> > --- a/include/uapi/linux/virtio_net.h
> > +++ b/include/uapi/linux/virtio_net.h
> > @@ -56,6 +56,9 @@
> >   #define VIRTIO_NET_F_MQ     22      /* Device supports Receive Flow
> >                                        * Steering */
> >   #define VIRTIO_NET_F_CTRL_MAC_ADDR 23       /* Set MAC address */
> > +#define VIRTIO_NET_F_GUEST_USO4      54      /* Guest can handle USOv4 in. */
> > +#define VIRTIO_NET_F_GUEST_USO6      55      /* Guest can handle USOv6 in. */
> > +#define VIRTIO_NET_F_HOST_USO        56      /* Host can handle USO in. */
>
>
> I think it's better to be consistent here. Either we split in both guest
> and host or not.
>
> Thanks
>

The main reason that receives USO packets depends on the kernel, where
transmitting the feature that VirtIO implements.
Windows systems have the option to manipulate receive offload. That's
why there are two GUEST_USO features.
For HOST_USO - technically there is no point in "split" it, and there
is should not be any difference between IPv4/IPv6.
Technically, we either support transmitting big UDP packets or not.

>
> >
> >   #define VIRTIO_NET_F_HASH_REPORT  57        /* Supports hash report */
> >   #define VIRTIO_NET_F_RSS      60    /* Supports RSS RX steering */
> > @@ -130,6 +133,7 @@ struct virtio_net_hdr_v1 {
> >   #define VIRTIO_NET_HDR_GSO_TCPV4    1       /* GSO frame, IPv4 TCP (TSO) */
> >   #define VIRTIO_NET_HDR_GSO_UDP              3       /* GSO frame, IPv4 UDP (UFO) */
> >   #define VIRTIO_NET_HDR_GSO_TCPV6    4       /* GSO frame, IPv6 TCP */
> > +#define VIRTIO_NET_HDR_GSO_UDP_L4    5       /* GSO frame, IPv4 & IPv6 UDP (USO) */
> >   #define VIRTIO_NET_HDR_GSO_ECN              0x80    /* TCP has ECN set */
> >       __u8 gso_type;
> >       __virtio16 hdr_len;     /* Ethernet + IP + tcp/udp hdrs */
>

[RFC PATCH 4/5] linux/virtio_net.h: Added Support for GSO_UDP_L4 offload.

[Andrew Melnychenko]

Now, it's possible to convert vnet packets from/to skb.

Signed-off-by: Andrew Melnychenko <andrew@daynix.com>
---
 include/linux/virtio_net.h | 11 +++++++++++
 1 file changed, 11 insertions(+)

diff --git a/include/linux/virtio_net.h b/include/linux/virtio_net.h
index a960de68ac69..9311d41d0a81 100644
--- a/include/linux/virtio_net.h
+++ b/include/linux/virtio_net.h
@@ -17,6 +17,9 @@ static inline bool virtio_net_hdr_match_proto(__be16 protocol, __u8 gso_type)
 	case VIRTIO_NET_HDR_GSO_UDP:
 		return protocol == cpu_to_be16(ETH_P_IP) ||
 		       protocol == cpu_to_be16(ETH_P_IPV6);
+	case VIRTIO_NET_HDR_GSO_UDP_L4:
+		return protocol == cpu_to_be16(ETH_P_IP) ||
+		       protocol == cpu_to_be16(ETH_P_IPV6);
 	default:
 		return false;
 	}
@@ -31,6 +34,7 @@ static inline int virtio_net_hdr_set_proto(struct sk_buff *skb,
 	switch (hdr->gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
 	case VIRTIO_NET_HDR_GSO_TCPV4:
 	case VIRTIO_NET_HDR_GSO_UDP:
+	case VIRTIO_NET_HDR_GSO_UDP_L4:
 		skb->protocol = cpu_to_be16(ETH_P_IP);
 		break;
 	case VIRTIO_NET_HDR_GSO_TCPV6:
@@ -69,6 +73,11 @@ static inline int virtio_net_hdr_to_skb(struct sk_buff *skb,
 			ip_proto = IPPROTO_UDP;
 			thlen = sizeof(struct udphdr);
 			break;
+		case VIRTIO_NET_HDR_GSO_UDP_L4:
+			gso_type = SKB_GSO_UDP_L4;
+			ip_proto = IPPROTO_UDP;
+			thlen = sizeof(struct udphdr);
+			break;
 		default:
 			return -EINVAL;
 		}
@@ -182,6 +191,8 @@ static inline int virtio_net_hdr_from_skb(const struct sk_buff *skb,
 			hdr->gso_type = VIRTIO_NET_HDR_GSO_TCPV4;
 		else if (sinfo->gso_type & SKB_GSO_TCPV6)
 			hdr->gso_type = VIRTIO_NET_HDR_GSO_TCPV6;
+		else if (sinfo->gso_type & SKB_GSO_UDP_L4)
+			hdr->gso_type = VIRTIO_NET_HDR_GSO_UDP_L4;
 		else
 			return -EINVAL;
 		if (sinfo->gso_type & SKB_GSO_TCP_ECN)
-- 
2.34.1

Re: [RFC PATCH 4/5] linux/virtio_net.h: Added Support for GSO_UDP_L4 offload.

[Jason Wang]

在 2022/1/25 下午4:47, Andrew Melnychenko 写道:
> Now, it's possible to convert vnet packets from/to skb.


I suggest to change the title to "net: support XXX offload in vnet header"

Thanks


>
> Signed-off-by: Andrew Melnychenko <andrew@daynix.com>
> ---
>   include/linux/virtio_net.h | 11 +++++++++++
>   1 file changed, 11 insertions(+)
>
> diff --git a/include/linux/virtio_net.h b/include/linux/virtio_net.h
> index a960de68ac69..9311d41d0a81 100644
> --- a/include/linux/virtio_net.h
> +++ b/include/linux/virtio_net.h
> @@ -17,6 +17,9 @@ static inline bool virtio_net_hdr_match_proto(__be16 protocol, __u8 gso_type)
>   	case VIRTIO_NET_HDR_GSO_UDP:
>   		return protocol == cpu_to_be16(ETH_P_IP) ||
>   		       protocol == cpu_to_be16(ETH_P_IPV6);
> +	case VIRTIO_NET_HDR_GSO_UDP_L4:
> +		return protocol == cpu_to_be16(ETH_P_IP) ||
> +		       protocol == cpu_to_be16(ETH_P_IPV6);
>   	default:
>   		return false;
>   	}
> @@ -31,6 +34,7 @@ static inline int virtio_net_hdr_set_proto(struct sk_buff *skb,
>   	switch (hdr->gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
>   	case VIRTIO_NET_HDR_GSO_TCPV4:
>   	case VIRTIO_NET_HDR_GSO_UDP:
> +	case VIRTIO_NET_HDR_GSO_UDP_L4:
>   		skb->protocol = cpu_to_be16(ETH_P_IP);
>   		break;
>   	case VIRTIO_NET_HDR_GSO_TCPV6:
> @@ -69,6 +73,11 @@ static inline int virtio_net_hdr_to_skb(struct sk_buff *skb,
>   			ip_proto = IPPROTO_UDP;
>   			thlen = sizeof(struct udphdr);
>   			break;
> +		case VIRTIO_NET_HDR_GSO_UDP_L4:
> +			gso_type = SKB_GSO_UDP_L4;
> +			ip_proto = IPPROTO_UDP;
> +			thlen = sizeof(struct udphdr);
> +			break;
>   		default:
>   			return -EINVAL;
>   		}
> @@ -182,6 +191,8 @@ static inline int virtio_net_hdr_from_skb(const struct sk_buff *skb,
>   			hdr->gso_type = VIRTIO_NET_HDR_GSO_TCPV4;
>   		else if (sinfo->gso_type & SKB_GSO_TCPV6)
>   			hdr->gso_type = VIRTIO_NET_HDR_GSO_TCPV6;
> +		else if (sinfo->gso_type & SKB_GSO_UDP_L4)
> +			hdr->gso_type = VIRTIO_NET_HDR_GSO_UDP_L4;
>   		else
>   			return -EINVAL;
>   		if (sinfo->gso_type & SKB_GSO_TCP_ECN)

[RFC PATCH 5/5] drivers/net/virtio_net.c: Added USO support.

[Andrew Melnychenko]

Now, it possible to enable GSO_UDP_L4("tx-udp-segmentation") for VirtioNet.

Signed-off-by: Andrew Melnychenko <andrew@daynix.com>
---
 drivers/net/virtio_net.c | 22 ++++++++++++++++++----
 1 file changed, 18 insertions(+), 4 deletions(-)

diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index a801ea40908f..a45eee022be4 100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c
@@ -60,13 +60,17 @@ static const unsigned long guest_offloads[] = {
 	VIRTIO_NET_F_GUEST_TSO6,
 	VIRTIO_NET_F_GUEST_ECN,
 	VIRTIO_NET_F_GUEST_UFO,
-	VIRTIO_NET_F_GUEST_CSUM
+	VIRTIO_NET_F_GUEST_CSUM,
+	VIRTIO_NET_F_GUEST_USO4,
+	VIRTIO_NET_F_GUEST_USO6
 };
 
 #define GUEST_OFFLOAD_GRO_HW_MASK ((1ULL << VIRTIO_NET_F_GUEST_TSO4) | \
 				(1ULL << VIRTIO_NET_F_GUEST_TSO6) | \
 				(1ULL << VIRTIO_NET_F_GUEST_ECN)  | \
-				(1ULL << VIRTIO_NET_F_GUEST_UFO))
+				(1ULL << VIRTIO_NET_F_GUEST_UFO)  | \
+				(1ULL << VIRTIO_NET_F_GUEST_USO4) | \
+				(1ULL << VIRTIO_NET_F_GUEST_USO6))
 
 struct virtnet_stat_desc {
 	char desc[ETH_GSTRING_LEN];
@@ -2530,7 +2534,9 @@ static int virtnet_xdp_set(struct net_device *dev, struct bpf_prog *prog,
 	        virtio_has_feature(vi->vdev, VIRTIO_NET_F_GUEST_TSO6) ||
 	        virtio_has_feature(vi->vdev, VIRTIO_NET_F_GUEST_ECN) ||
 		virtio_has_feature(vi->vdev, VIRTIO_NET_F_GUEST_UFO) ||
-		virtio_has_feature(vi->vdev, VIRTIO_NET_F_GUEST_CSUM))) {
+		virtio_has_feature(vi->vdev, VIRTIO_NET_F_GUEST_CSUM) ||
+		virtio_has_feature(vi->vdev, VIRTIO_NET_F_GUEST_USO4) ||
+		virtio_has_feature(vi->vdev, VIRTIO_NET_F_GUEST_USO6))) {
 		NL_SET_ERR_MSG_MOD(extack, "Can't set XDP while host is implementing GRO_HW/CSUM, disable GRO_HW/CSUM first");
 		return -EOPNOTSUPP;
 	}
@@ -3155,6 +3161,8 @@ static int virtnet_probe(struct virtio_device *vdev)
 			dev->hw_features |= NETIF_F_TSO6;
 		if (virtio_has_feature(vdev, VIRTIO_NET_F_HOST_ECN))
 			dev->hw_features |= NETIF_F_TSO_ECN;
+		if (virtio_has_feature(vdev, VIRTIO_NET_F_HOST_USO))
+			dev->hw_features |= NETIF_F_GSO_UDP_L4;
 
 		dev->features |= NETIF_F_GSO_ROBUST;
 
@@ -3169,6 +3177,9 @@ static int virtnet_probe(struct virtio_device *vdev)
 		dev->features |= NETIF_F_GRO_HW;
 	if (virtio_has_feature(vdev, VIRTIO_NET_F_CTRL_GUEST_OFFLOADS))
 		dev->hw_features |= NETIF_F_GRO_HW;
+	if (virtio_has_feature(vdev, VIRTIO_NET_F_GUEST_USO4) ||
+	    virtio_has_feature(vdev, VIRTIO_NET_F_GUEST_USO6))
+		dev->hw_features |= NETIF_F_LRO;
 
 	dev->vlan_features = dev->features;
 
@@ -3200,7 +3211,9 @@ static int virtnet_probe(struct virtio_device *vdev)
 	if (virtio_has_feature(vdev, VIRTIO_NET_F_GUEST_TSO4) ||
 	    virtio_has_feature(vdev, VIRTIO_NET_F_GUEST_TSO6) ||
 	    virtio_has_feature(vdev, VIRTIO_NET_F_GUEST_ECN) ||
-	    virtio_has_feature(vdev, VIRTIO_NET_F_GUEST_UFO))
+	    virtio_has_feature(vdev, VIRTIO_NET_F_GUEST_UFO) ||
+	    virtio_has_feature(vdev, VIRTIO_NET_F_GUEST_USO4) ||
+	    virtio_has_feature(vdev, VIRTIO_NET_F_GUEST_USO6))
 		vi->big_packets = true;
 
 	if (virtio_has_feature(vdev, VIRTIO_NET_F_MRG_RXBUF))
@@ -3400,6 +3413,7 @@ static struct virtio_device_id id_table[] = {
 	VIRTIO_NET_F_HOST_TSO4, VIRTIO_NET_F_HOST_UFO, VIRTIO_NET_F_HOST_TSO6, \
 	VIRTIO_NET_F_HOST_ECN, VIRTIO_NET_F_GUEST_TSO4, VIRTIO_NET_F_GUEST_TSO6, \
 	VIRTIO_NET_F_GUEST_ECN, VIRTIO_NET_F_GUEST_UFO, \
+	VIRTIO_NET_F_HOST_USO, VIRTIO_NET_F_GUEST_USO4, VIRTIO_NET_F_GUEST_USO6, \
 	VIRTIO_NET_F_MRG_RXBUF, VIRTIO_NET_F_STATUS, VIRTIO_NET_F_CTRL_VQ, \
 	VIRTIO_NET_F_CTRL_RX, VIRTIO_NET_F_CTRL_VLAN, \
 	VIRTIO_NET_F_GUEST_ANNOUNCE, VIRTIO_NET_F_MQ, \
-- 
2.34.1

Re: [RFC PATCH 5/5] drivers/net/virtio_net.c: Added USO support.

[Jason Wang]

在 2022/1/25 下午4:47, Andrew Melnychenko 写道:
> Now, it possible to enable GSO_UDP_L4("tx-udp-segmentation") for VirtioNet.
>
> Signed-off-by: Andrew Melnychenko <andrew@daynix.com>
> ---
>   drivers/net/virtio_net.c | 22 ++++++++++++++++++----
>   1 file changed, 18 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> index a801ea40908f..a45eee022be4 100644
> --- a/drivers/net/virtio_net.c
> +++ b/drivers/net/virtio_net.c
> @@ -60,13 +60,17 @@ static const unsigned long guest_offloads[] = {
>   	VIRTIO_NET_F_GUEST_TSO6,
>   	VIRTIO_NET_F_GUEST_ECN,
>   	VIRTIO_NET_F_GUEST_UFO,
> -	VIRTIO_NET_F_GUEST_CSUM
> +	VIRTIO_NET_F_GUEST_CSUM,
> +	VIRTIO_NET_F_GUEST_USO4,
> +	VIRTIO_NET_F_GUEST_USO6
>   };
>   
>   #define GUEST_OFFLOAD_GRO_HW_MASK ((1ULL << VIRTIO_NET_F_GUEST_TSO4) | \
>   				(1ULL << VIRTIO_NET_F_GUEST_TSO6) | \
>   				(1ULL << VIRTIO_NET_F_GUEST_ECN)  | \
> -				(1ULL << VIRTIO_NET_F_GUEST_UFO))
> +				(1ULL << VIRTIO_NET_F_GUEST_UFO)  | \
> +				(1ULL << VIRTIO_NET_F_GUEST_USO4) | \
> +				(1ULL << VIRTIO_NET_F_GUEST_USO6))
>   
>   struct virtnet_stat_desc {
>   	char desc[ETH_GSTRING_LEN];
> @@ -2530,7 +2534,9 @@ static int virtnet_xdp_set(struct net_device *dev, struct bpf_prog *prog,
>   	        virtio_has_feature(vi->vdev, VIRTIO_NET_F_GUEST_TSO6) ||
>   	        virtio_has_feature(vi->vdev, VIRTIO_NET_F_GUEST_ECN) ||
>   		virtio_has_feature(vi->vdev, VIRTIO_NET_F_GUEST_UFO) ||
> -		virtio_has_feature(vi->vdev, VIRTIO_NET_F_GUEST_CSUM))) {
> +		virtio_has_feature(vi->vdev, VIRTIO_NET_F_GUEST_CSUM) ||
> +		virtio_has_feature(vi->vdev, VIRTIO_NET_F_GUEST_USO4) ||
> +		virtio_has_feature(vi->vdev, VIRTIO_NET_F_GUEST_USO6))) {
>   		NL_SET_ERR_MSG_MOD(extack, "Can't set XDP while host is implementing GRO_HW/CSUM, disable GRO_HW/CSUM first");
>   		return -EOPNOTSUPP;
>   	}
> @@ -3155,6 +3161,8 @@ static int virtnet_probe(struct virtio_device *vdev)
>   			dev->hw_features |= NETIF_F_TSO6;
>   		if (virtio_has_feature(vdev, VIRTIO_NET_F_HOST_ECN))
>   			dev->hw_features |= NETIF_F_TSO_ECN;
> +		if (virtio_has_feature(vdev, VIRTIO_NET_F_HOST_USO))
> +			dev->hw_features |= NETIF_F_GSO_UDP_L4;
>   
>   		dev->features |= NETIF_F_GSO_ROBUST;
>   
> @@ -3169,6 +3177,9 @@ static int virtnet_probe(struct virtio_device *vdev)
>   		dev->features |= NETIF_F_GRO_HW;
>   	if (virtio_has_feature(vdev, VIRTIO_NET_F_CTRL_GUEST_OFFLOADS))
>   		dev->hw_features |= NETIF_F_GRO_HW;
> +	if (virtio_has_feature(vdev, VIRTIO_NET_F_GUEST_USO4) ||
> +	    virtio_has_feature(vdev, VIRTIO_NET_F_GUEST_USO6))
> +		dev->hw_features |= NETIF_F_LRO;


I think we need to use GRO_HW, see dbcf24d153884 ("virtio-net: use 
NETIF_F_GRO_HW instead of NETIF_F_LRO"

Thanks


>   
>   	dev->vlan_features = dev->features;
>   
> @@ -3200,7 +3211,9 @@ static int virtnet_probe(struct virtio_device *vdev)
>   	if (virtio_has_feature(vdev, VIRTIO_NET_F_GUEST_TSO4) ||
>   	    virtio_has_feature(vdev, VIRTIO_NET_F_GUEST_TSO6) ||
>   	    virtio_has_feature(vdev, VIRTIO_NET_F_GUEST_ECN) ||
> -	    virtio_has_feature(vdev, VIRTIO_NET_F_GUEST_UFO))
> +	    virtio_has_feature(vdev, VIRTIO_NET_F_GUEST_UFO) ||
> +	    virtio_has_feature(vdev, VIRTIO_NET_F_GUEST_USO4) ||
> +	    virtio_has_feature(vdev, VIRTIO_NET_F_GUEST_USO6))
>   		vi->big_packets = true;
>   
>   	if (virtio_has_feature(vdev, VIRTIO_NET_F_MRG_RXBUF))
> @@ -3400,6 +3413,7 @@ static struct virtio_device_id id_table[] = {
>   	VIRTIO_NET_F_HOST_TSO4, VIRTIO_NET_F_HOST_UFO, VIRTIO_NET_F_HOST_TSO6, \
>   	VIRTIO_NET_F_HOST_ECN, VIRTIO_NET_F_GUEST_TSO4, VIRTIO_NET_F_GUEST_TSO6, \
>   	VIRTIO_NET_F_GUEST_ECN, VIRTIO_NET_F_GUEST_UFO, \
> +	VIRTIO_NET_F_HOST_USO, VIRTIO_NET_F_GUEST_USO4, VIRTIO_NET_F_GUEST_USO6, \
>   	VIRTIO_NET_F_MRG_RXBUF, VIRTIO_NET_F_STATUS, VIRTIO_NET_F_CTRL_VQ, \
>   	VIRTIO_NET_F_CTRL_RX, VIRTIO_NET_F_CTRL_VLAN, \
>   	VIRTIO_NET_F_GUEST_ANNOUNCE, VIRTIO_NET_F_MQ, \

Re: [RFC PATCH 0/5] TUN/VirtioNet USO features support.

[Xuan Zhuo]

On Tue, 25 Jan 2022 10:46:57 +0200, Andrew Melnychenko <andrew@daynix.com> wrote:
> Added new offloads for TUN devices TUN_F_USO4 and TUN_F_USO6.
> Technically they enable NETIF_F_GSO_UDP_L4
> (and only if USO4 & USO6 are set simultaneously).
> It allows to transmission of large UDP packets.
>
> Different features USO4 and USO6 are required for qemu where Windows guests can
> enable disable USO receives for IPv4 and IPv6 separately.
> On the other side, Linux can't really differentiate USO4 and USO6, for now.
> For now, to enable USO for TUN it requires enabling USO4 and USO6 together.
> In the future, there would be a mechanism to control UDP_L4 GSO separately.
>
> Test it WIP Qemu https://github.com/daynix/qemu/tree/Dev_USOv2
>
> New types for VirtioNet already on mailing:
> https://lists.oasis-open.org/archives/virtio-comment/202110/msg00010.html

Seems like this hasn't been upvoted yet.

	https://github.com/oasis-tcs/virtio-spec#use-of-github-issues

Thanks.

>
> Also, there is a known issue with transmitting packages between two guests.
> Without hacks with skb's GSO - packages are still segmented on the host's postrouting.
>
> Andrew Melnychenko (5):
>   uapi/linux/if_tun.h: Added new ioctl for tun/tap.
>   driver/net/tun: Added features for USO.
>   uapi/linux/virtio_net.h: Added USO types.
>   linux/virtio_net.h: Added Support for GSO_UDP_L4 offload.
>   drivers/net/virtio_net.c: Added USO support.
>
>  drivers/net/tap.c               | 18 ++++++++++++++++--
>  drivers/net/tun.c               | 15 ++++++++++++++-
>  drivers/net/virtio_net.c        | 22 ++++++++++++++++++----
>  include/linux/virtio_net.h      | 11 +++++++++++
>  include/uapi/linux/if_tun.h     |  3 +++
>  include/uapi/linux/virtio_net.h |  4 ++++
>  6 files changed, 66 insertions(+), 7 deletions(-)
>
> --
> 2.34.1
>
> _______________________________________________
> Virtualization mailing list
> Virtualization@lists.linux-foundation.org
> https://lists.linuxfoundation.org/mailman/listinfo/virtualization

Re: [RFC PATCH 0/5] TUN/VirtioNet USO features support.

[Yuri Benditovich]

On Wed, Jan 26, 2022 at 9:54 AM Xuan Zhuo <xuanzhuo@linux.alibaba.com> wrote:
>
> On Tue, 25 Jan 2022 10:46:57 +0200, Andrew Melnychenko <andrew@daynix.com> wrote:
> > Added new offloads for TUN devices TUN_F_USO4 and TUN_F_USO6.
> > Technically they enable NETIF_F_GSO_UDP_L4
> > (and only if USO4 & USO6 are set simultaneously).
> > It allows to transmission of large UDP packets.
> >
> > Different features USO4 and USO6 are required for qemu where Windows guests can
> > enable disable USO receives for IPv4 and IPv6 separately.
> > On the other side, Linux can't really differentiate USO4 and USO6, for now.
> > For now, to enable USO for TUN it requires enabling USO4 and USO6 together.
> > In the future, there would be a mechanism to control UDP_L4 GSO separately.
> >
> > Test it WIP Qemu https://github.com/daynix/qemu/tree/Dev_USOv2
> >
> > New types for VirtioNet already on mailing:
> > https://lists.oasis-open.org/archives/virtio-comment/202110/msg00010.html
>
> Seems like this hasn't been upvoted yet.
>
>         https://github.com/oasis-tcs/virtio-spec#use-of-github-issues

Yes, correct. This is a reason why this series of patches is RFC.

>
> Thanks.
>
> >
> > Also, there is a known issue with transmitting packages between two guests.
> > Without hacks with skb's GSO - packages are still segmented on the host's postrouting.
> >
> > Andrew Melnychenko (5):
> >   uapi/linux/if_tun.h: Added new ioctl for tun/tap.
> >   driver/net/tun: Added features for USO.
> >   uapi/linux/virtio_net.h: Added USO types.
> >   linux/virtio_net.h: Added Support for GSO_UDP_L4 offload.
> >   drivers/net/virtio_net.c: Added USO support.
> >
> >  drivers/net/tap.c               | 18 ++++++++++++++++--
> >  drivers/net/tun.c               | 15 ++++++++++++++-
> >  drivers/net/virtio_net.c        | 22 ++++++++++++++++++----
> >  include/linux/virtio_net.h      | 11 +++++++++++
> >  include/uapi/linux/if_tun.h     |  3 +++
> >  include/uapi/linux/virtio_net.h |  4 ++++
> >  6 files changed, 66 insertions(+), 7 deletions(-)
> >
> > --
> > 2.34.1
> >
> > _______________________________________________
> > Virtualization mailing list
> > Virtualization@lists.linux-foundation.org
> > https://lists.linuxfoundation.org/mailman/listinfo/virtualization

Re: [RFC PATCH 0/5] TUN/VirtioNet USO features support.

[Andrew Melnichenko]

Hi people,
Can you please review this series?

On Wed, Jan 26, 2022 at 10:32 AM Yuri Benditovich
<yuri.benditovich@daynix.com> wrote:
>
> On Wed, Jan 26, 2022 at 9:54 AM Xuan Zhuo <xuanzhuo@linux.alibaba.com> wrote:
> >
> > On Tue, 25 Jan 2022 10:46:57 +0200, Andrew Melnychenko <andrew@daynix.com> wrote:
> > > Added new offloads for TUN devices TUN_F_USO4 and TUN_F_USO6.
> > > Technically they enable NETIF_F_GSO_UDP_L4
> > > (and only if USO4 & USO6 are set simultaneously).
> > > It allows to transmission of large UDP packets.
> > >
> > > Different features USO4 and USO6 are required for qemu where Windows guests can
> > > enable disable USO receives for IPv4 and IPv6 separately.
> > > On the other side, Linux can't really differentiate USO4 and USO6, for now.
> > > For now, to enable USO for TUN it requires enabling USO4 and USO6 together.
> > > In the future, there would be a mechanism to control UDP_L4 GSO separately.
> > >
> > > Test it WIP Qemu https://github.com/daynix/qemu/tree/Dev_USOv2
> > >
> > > New types for VirtioNet already on mailing:
> > > https://lists.oasis-open.org/archives/virtio-comment/202110/msg00010.html
> >
> > Seems like this hasn't been upvoted yet.
> >
> >         https://github.com/oasis-tcs/virtio-spec#use-of-github-issues
>
> Yes, correct. This is a reason why this series of patches is RFC.
>
> >
> > Thanks.
> >
> > >
> > > Also, there is a known issue with transmitting packages between two guests.
> > > Without hacks with skb's GSO - packages are still segmented on the host's postrouting.
> > >
> > > Andrew Melnychenko (5):
> > >   uapi/linux/if_tun.h: Added new ioctl for tun/tap.
> > >   driver/net/tun: Added features for USO.
> > >   uapi/linux/virtio_net.h: Added USO types.
> > >   linux/virtio_net.h: Added Support for GSO_UDP_L4 offload.
> > >   drivers/net/virtio_net.c: Added USO support.
> > >
> > >  drivers/net/tap.c               | 18 ++++++++++++++++--
> > >  drivers/net/tun.c               | 15 ++++++++++++++-
> > >  drivers/net/virtio_net.c        | 22 ++++++++++++++++++----
> > >  include/linux/virtio_net.h      | 11 +++++++++++
> > >  include/uapi/linux/if_tun.h     |  3 +++
> > >  include/uapi/linux/virtio_net.h |  4 ++++
> > >  6 files changed, 66 insertions(+), 7 deletions(-)
> > >
> > > --
> > > 2.34.1
> > >
> > > _______________________________________________
> > > Virtualization mailing list
> > > Virtualization@lists.linux-foundation.org
> > > https://lists.linuxfoundation.org/mailman/listinfo/virtualization

Re: [RFC PATCH 0/5] TUN/VirtioNet USO features support.

[Jakub Kicinski]

On Tue, 8 Feb 2022 15:09:21 +0200 Andrew Melnichenko wrote:
> Hi people,
> Can you please review this series?

Adding Willem, he might be interested.

Re: [RFC PATCH 0/5] TUN/VirtioNet USO features support.

[Jason Wang]

在 2022/2/8 下午9:09, Andrew Melnichenko 写道:
> Hi people,
> Can you please review this series?


Are there any performance number to demonstrate the difference?

Thanks


>
> On Wed, Jan 26, 2022 at 10:32 AM Yuri Benditovich
> <yuri.benditovich@daynix.com> wrote:
>> On Wed, Jan 26, 2022 at 9:54 AM Xuan Zhuo <xuanzhuo@linux.alibaba.com> wrote:
>>> On Tue, 25 Jan 2022 10:46:57 +0200, Andrew Melnychenko <andrew@daynix.com> wrote:
>>>> Added new offloads for TUN devices TUN_F_USO4 and TUN_F_USO6.
>>>> Technically they enable NETIF_F_GSO_UDP_L4
>>>> (and only if USO4 & USO6 are set simultaneously).
>>>> It allows to transmission of large UDP packets.
>>>>
>>>> Different features USO4 and USO6 are required for qemu where Windows guests can
>>>> enable disable USO receives for IPv4 and IPv6 separately.
>>>> On the other side, Linux can't really differentiate USO4 and USO6, for now.
>>>> For now, to enable USO for TUN it requires enabling USO4 and USO6 together.
>>>> In the future, there would be a mechanism to control UDP_L4 GSO separately.
>>>>
>>>> Test it WIP Qemu https://github.com/daynix/qemu/tree/Dev_USOv2
>>>>
>>>> New types for VirtioNet already on mailing:
>>>> https://lists.oasis-open.org/archives/virtio-comment/202110/msg00010.html
>>> Seems like this hasn't been upvoted yet.
>>>
>>>          https://github.com/oasis-tcs/virtio-spec#use-of-github-issues
>> Yes, correct. This is a reason why this series of patches is RFC.
>>
>>> Thanks.
>>>
>>>> Also, there is a known issue with transmitting packages between two guests.
>>>> Without hacks with skb's GSO - packages are still segmented on the host's postrouting.
>>>>
>>>> Andrew Melnychenko (5):
>>>>    uapi/linux/if_tun.h: Added new ioctl for tun/tap.
>>>>    driver/net/tun: Added features for USO.
>>>>    uapi/linux/virtio_net.h: Added USO types.
>>>>    linux/virtio_net.h: Added Support for GSO_UDP_L4 offload.
>>>>    drivers/net/virtio_net.c: Added USO support.
>>>>
>>>>   drivers/net/tap.c               | 18 ++++++++++++++++--
>>>>   drivers/net/tun.c               | 15 ++++++++++++++-
>>>>   drivers/net/virtio_net.c        | 22 ++++++++++++++++++----
>>>>   include/linux/virtio_net.h      | 11 +++++++++++
>>>>   include/uapi/linux/if_tun.h     |  3 +++
>>>>   include/uapi/linux/virtio_net.h |  4 ++++
>>>>   6 files changed, 66 insertions(+), 7 deletions(-)
>>>>
>>>> --
>>>> 2.34.1
>>>>
>>>> _______________________________________________
>>>> Virtualization mailing list
>>>> Virtualization@lists.linux-foundation.org
>>>> https://lists.linuxfoundation.org/mailman/listinfo/virtualization

Re: [RFC PATCH 0/5] TUN/VirtioNet USO features support.

[Andrew Melnichenko]

Hi all,

On Wed, Feb 9, 2022 at 7:41 AM Jason Wang <jasowang@redhat.com> wrote:
>
>
> 在 2022/2/8 下午9:09, Andrew Melnichenko 写道:
> > Hi people,
> > Can you please review this series?
>
>
> Are there any performance number to demonstrate the difference?
>
> Thanks
>

Yeah, I've used udpgso_bench from Linux to test.
Here are some numbers:

Sending packets with size 10000

Without USO:
```
$ ./udpgso_bench_tx -4 -D 192.168.15.1 -s 10000 -S 1000
random: crng init done
random: 7 urandom warning(s) missed due to ratelimiting
udp tx:     36 MB/s     3863 calls/s   3863 msg/s
udp tx:     32 MB/s     3360 calls/s   3360 msg/s
udp tx:     31 MB/s     3340 calls/s   3340 msg/s
udp tx:     31 MB/s     3353 calls/s   3353 msg/s
udp tx:     32 MB/s     3359 calls/s   3359 msg/s
udp tx:     32 MB/s     3370 calls/s   3370 msg/s
```

With USO:
```
$ ./udpgso_bench_tx -4 -D 192.168.15.1 -s 10000 -S 1000
random: crng init done
random: 7 urandom warning(s) missed due to ratelimiting
udp tx:    120 MB/s    12596 calls/s  12596 msg/s
udp tx:    122 MB/s    12885 calls/s  12885 msg/s
udp tx:    120 MB/s    12667 calls/s  12667 msg/s
udp tx:    123 MB/s    12969 calls/s  12969 msg/s
udp tx:    116 MB/s    12232 calls/s  12232 msg/s
udp tx:    108 MB/s    11389 calls/s  11389 msg/s
```


>
> >
> > On Wed, Jan 26, 2022 at 10:32 AM Yuri Benditovich
> > <yuri.benditovich@daynix.com> wrote:
> >> On Wed, Jan 26, 2022 at 9:54 AM Xuan Zhuo <xuanzhuo@linux.alibaba.com> wrote:
> >>> On Tue, 25 Jan 2022 10:46:57 +0200, Andrew Melnychenko <andrew@daynix.com> wrote:
> >>>> Added new offloads for TUN devices TUN_F_USO4 and TUN_F_USO6.
> >>>> Technically they enable NETIF_F_GSO_UDP_L4
> >>>> (and only if USO4 & USO6 are set simultaneously).
> >>>> It allows to transmission of large UDP packets.
> >>>>
> >>>> Different features USO4 and USO6 are required for qemu where Windows guests can
> >>>> enable disable USO receives for IPv4 and IPv6 separately.
> >>>> On the other side, Linux can't really differentiate USO4 and USO6, for now.
> >>>> For now, to enable USO for TUN it requires enabling USO4 and USO6 together.
> >>>> In the future, there would be a mechanism to control UDP_L4 GSO separately.
> >>>>
> >>>> Test it WIP Qemu https://github.com/daynix/qemu/tree/Dev_USOv2
> >>>>
> >>>> New types for VirtioNet already on mailing:
> >>>> https://lists.oasis-open.org/archives/virtio-comment/202110/msg00010.html
> >>> Seems like this hasn't been upvoted yet.
> >>>
> >>>          https://github.com/oasis-tcs/virtio-spec#use-of-github-issues
> >> Yes, correct. This is a reason why this series of patches is RFC.
> >>
> >>> Thanks.
> >>>
> >>>> Also, there is a known issue with transmitting packages between two guests.
> >>>> Without hacks with skb's GSO - packages are still segmented on the host's postrouting.
> >>>>
> >>>> Andrew Melnychenko (5):
> >>>>    uapi/linux/if_tun.h: Added new ioctl for tun/tap.
> >>>>    driver/net/tun: Added features for USO.
> >>>>    uapi/linux/virtio_net.h: Added USO types.
> >>>>    linux/virtio_net.h: Added Support for GSO_UDP_L4 offload.
> >>>>    drivers/net/virtio_net.c: Added USO support.
> >>>>
> >>>>   drivers/net/tap.c               | 18 ++++++++++++++++--
> >>>>   drivers/net/tun.c               | 15 ++++++++++++++-
> >>>>   drivers/net/virtio_net.c        | 22 ++++++++++++++++++----
> >>>>   include/linux/virtio_net.h      | 11 +++++++++++
> >>>>   include/uapi/linux/if_tun.h     |  3 +++
> >>>>   include/uapi/linux/virtio_net.h |  4 ++++
> >>>>   6 files changed, 66 insertions(+), 7 deletions(-)
> >>>>
> >>>> --
> >>>> 2.34.1
> >>>>
> >>>> _______________________________________________
> >>>> Virtualization mailing list
> >>>> Virtualization@lists.linux-foundation.org
> >>>> https://lists.linuxfoundation.org/mailman/listinfo/virtualization
>