Skip to contentSkip to conversation

virtio-net: Introduce a new field to indicate outer network header offset

PATCH Superseded Kommula Shiva Shankar 2 participants 4 messages Jan 14, 2025 – Jan 22, 2025

v1 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 virtio-net: Introduce a new field to indicate outer network header offset

Hi Everyone,
It has been some time since this patch was sent. I am resending it for your review and look forward to your feedback.

This patch introduces a new field in the virtio_net_header called out_nh_offset, along with a new net device feature, VIRTIO_NET_F_OUT_NET_HEADER.

Currently, there is no field available to directly read the outer network header offset in case of segmentation offload. This requires reading packet data, which significantly affects performance in datapath.
Additionally, some hardware implementations requrie knowledege of the outer L3 offset (aka L2 length) for inline IPSec hardware acceleration.

To address this limitation, we propose splitting the csum_offset field into two 8-bit fields named csum_offset and out_nh_offset. The csum_offset indicates the offset value from the csum_start and may not exceed 256B bits(2^8) for protocols that use a 16-bit one's complement checksum

Following table lists such protocols and their checksum offset fields within their headers

+-----+--------+
|Proto|csum_off|
+-----+--------+
| IPV4| 10B    |
| ICMP| 2B     |
| IGMP| 2B     |
|  TCP| 16B    |
|  UDP| 6B     |
+-----+--------+

The out_nh_offset represents the start offset of the outer network header from the beginning of the packet data

This issue was briefly discussed on the mailing list in a different thread, which can be found here
https://lore.kernel.org/all/DM4PR18MB4269FAAC3CFC7E57E25DFBD2DF8B2@DM4PR18MB4269.namprd18.prod.outlook.com/

---
 device-types/net/description.tex | 26 +++++++++++++++++++++++++-
 1 file changed, 25 insertions(+), 1 deletion(-)
device-types/net/description.tex +25 -1
index 76585b0..48c081b 100644
@@ -88,6 +88,10 @@ \subsection{Feature bits}\label{sec:Device Types / Network Device / Feature bits
 \item[VIRTIO_NET_F_CTRL_MAC_ADDR(23)] Set MAC address through control
     channel.
 
+\item[VIRTIO_NET_F_OUT_NET_HEADER(50)] Driver can provide the start of \field{out_nh_offset}
+    value. Device gains advantage by not reading packet data to calculate outer network
+    header offset.
+
 \item[VIRTIO_NET_F_HASH_TUNNEL(51)] Device supports inner header hash for encapsulated packets.
 
 \item[VIRTIO_NET_F_VQ_NOTF_COAL(52)] Device supports virtqueue notification coalescing.
@@ -418,7 +422,13 @@ \subsection{Device Operation}\label{sec:Device Types / Network Device / Device O
         le16 hdr_len;
         le16 gso_size;
         le16 csum_start;
-        le16 csum_offset;
+        union {
+                le16 csum_offset;
+                struct { (Only if VIRTIO_NET_F_OUT_NET_HEADER negotiated)
+                        le8 csum_offset;
+                        le8 out_nh_offset;
+                };
+        };
         le16 num_buffers;
         le32 hash_value;        (Only if VIRTIO_NET_F_HASH_REPORT negotiated)
         le16 hash_report;       (Only if VIRTIO_NET_F_HASH_REPORT negotiated)
@@ -457,6 +467,10 @@ \subsubsection{Packet Transmission}\label{sec:Device Types / Network Device / De
 \item The driver can send a completely checksummed packet.  In this case,
   \field{flags} will be zero, and \field{gso_type} will be VIRTIO_NET_HDR_GSO_NONE.
 
+\item If the driver negotiated VIRTIO_NET_F_OUT_NET_HEADER,
+   \field{out_nh_offset} indicates the outer network header offset in bytes
+    from the beginning of the packet.
+
 \item If the driver negotiated VIRTIO_NET_F_CSUM, it can skip
   checksumming the packet:
   \begin{itemize}
@@ -531,6 +545,10 @@ \subsubsection{Packet Transmission}\label{sec:Device Types / Network Device / De
 \field{flags} to zero and SHOULD supply a fully checksummed
 packet to the device.
 
+If the VIRTIO_NET_F_OUT_NET_HEADER feature has been negotiated,
+the driver MUST set \field{out_nh_offset} to point outer
+network header start, otherwise to zero.
+
 If VIRTIO_NET_F_HOST_TSO4 is negotiated, the driver MAY set
 \field{gso_type} to VIRTIO_NET_HDR_GSO_TCPV4 to request TCPv4
 segmentation, otherwise the driver MUST NOT set
@@ -596,6 +614,7 @@ \subsubsection{Packet Transmission}\label{sec:Device Types / Network Device / De
 	the driver SHOULD set \field{hdr_len} to a value
 	not less than the length of the headers, including the transport
 	header.
+
 \end{itemize}
 
 The driver SHOULD accept the VIRTIO_NET_F_GUEST_HDRLEN feature if it has
@@ -610,6 +629,10 @@ \subsubsection{Packet Transmission}\label{sec:Device Types / Network Device / De
 If VIRTIO_NET_HDR_F_NEEDS_CSUM bit in \field{flags} is not set, the
 device MUST NOT use the \field{csum_start} and \field{csum_offset}.
 
+If the VIRTIO_NET_F_OUT_NET_HEADER feature has been negotiated,
+the device MAY use \field{out_nh_offset} as the outer network header
+offset, otherwise device MUST NOT use the \field{out_nh_offset} value.
+
 If one of the VIRTIO_NET_F_HOST_TSO4, TSO6, USO or UFO options have
 been negotiated:
 \begin{itemize}
@@ -631,6 +654,7 @@ \subsubsection{Packet Transmission}\label{sec:Device Types / Network Device / De
 	\begin{note}
 	This is due to various bugs in implementations.
 	\end{note}
+
 \end{itemize}
 
 If VIRTIO_NET_HDR_F_NEEDS_CSUM is not set, the device MUST NOT
Signature · 1 line
2.43.0
#1 lore raw
Quoted 67 lines

From: Kommula Shiva Shankar <[email protected]>
Sent: Tuesday, January 14, 2025 10:47 PM

Hi Everyone,
It has been some time since this patch was sent. I am resending it for your review and look forward to your feedback.

This patch introduces a new field in the virtio_net_header called out_nh_offset, along with a new net device feature, VIRTIO_NET_F_OUT_NET_HEADER.

Currently, there is no field available to directly read the outer network header offset in case of segmentation offload.
This requires reading packet data, which significantly affects performance in datapath.
Additionally, some hardware implementations requrie knowledege of the outer L3 offset (aka L2 length) for inline IPSec hardware acceleration.

To address this limitation, we propose splitting the csum_offset field into two 8-bit fields named csum_offset and out_nh_offset.
The csum_offset indicates the offset value from the csum_start and may not exceed 256B bits(2^8) for protocols that use a 16-bit one's complement checksum

Following table lists such protocols and their checksum offset fields within their headers

+-----+--------+
|Proto|csum_off|
+-----+--------+
| IPV4| 10B    |
| ICMP| 2B     |
| IGMP| 2B     |
|  TCP| 16B    |
|  UDP| 6B     |
+-----+--------+

The out_nh_offset represents the start offset of the outer network header from the beginning of the packet data

This issue was briefly discussed on the mailing list in a different thread, which can be found here
https://lore.k/
ernel.org%2Fall%2FDM4PR18MB4269FAAC3CFC7E57E25DFBD2DF8B2%40DM 4PR18MB4269.namprd18.prod.outlook.com%2F&data=05%7C02%7Cparav% 40nvidia.com%7Ce81096db39a14378627708dd34bf3909%7C43083d1572734 0c1b7db39efd9ccc17a%7C0%7C0%7C638724718122375976%7CUnknown%7 CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJ XaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=TH SOD9fclucwxtnizGLo8pIpjsTs1YoRww4VA%2FvDwgI%3D&reserved=0

---
 device-types/net/description.tex | 26 +++++++++++++++++++++++++-
 1 file changed, 25 insertions(+), 1 deletion(-)
device- renamed from device-types/net/description.tex +0 -0

types/net/description.tex
index 76585b0..48c081b 100644

device-types/net/description.tex +0 -0
@@ -88,6 +88,10 @@ \subsection{Feature bits}\label{sec:Device Types /

Network Device / Feature bits \item[VIRTIO_NET_F_CTRL_MAC_ADDR(23)] Set MAC address through control
channel.

Quoted 2 lines

+\item[VIRTIO_NET_F_OUT_NET_HEADER(50)] Driver can provide the start of \field{out_nh_offset}

The driver

device-types/net/description.tex \subsection{Feature bits}\label{sec:Device Types / Network Device / Feature bits · line 94

Quoted 4 lines

+ value. Device gains advantage by not reading packet data to calculate
outer network

device-types/net/description.tex +2 -0
+    header offset.
+

The device

May be just say 'packet' instead of 'packet data', because its slightly confusing between packet L2 header and packet data in this context.

Quoted 28 lines

\item[VIRTIO_NET_F_HASH_TUNNEL(51)] Device supports inner header
hash for encapsulated packets.

\item[VIRTIO_NET_F_VQ_NOTF_COAL(52)] Device supports virtqueue
notification coalescing.

device-types/net/description.tex +0 -0
@@ -418,7 +422,13 @@ \subsection{Device Operation}\label{sec:Device

Types / Network Device / Device O

         le16 hdr_len;
         le16 gso_size;
         le16 csum_start;
device-types/net/description.tex +7 -1
-        le16 csum_offset;
+        union {
+                le16 csum_offset;
+                struct { (Only if VIRTIO_NET_F_OUT_NET_HEADER negotiated)
+                        le8 csum_offset;
+                        le8 out_nh_offset;
+                };
+        };
         le16 num_buffers;
         le32 hash_value;        (Only if VIRTIO_NET_F_HASH_REPORT negotiated)
         le16 hash_report;       (Only if VIRTIO_NET_F_HASH_REPORT negotiated)
@@ -457,6 +467,10 @@ \subsubsection{Packet

Transmission}\label{sec:Device Types / Network Device / De \item The driver can send a completely checksummed packet. In this case,
\field{flags} will be zero, and \field{gso_type} will be
VIRTIO_NET_HDR_GSO_NONE.

+\item If the driver negotiated VIRTIO_NET_F_OUT_NET_HEADER,

device-types/net/description.tex \subsubsection{Packet Transmission}\label{sec:Device Types / Network Device / De · line 473

Quoted 3 lines
device-types/net/description.tex +3 -0
+   \field{out_nh_offset} indicates the outer network header offset in bytes
+    from the beginning of the packet.
+

If the driver has negotiated

device-types/net/description.tex

Quoted 11 lines

\item If the driver negotiated VIRTIO_NET_F_CSUM, it can skip
checksumming the packet:

   \begin{itemize}
device-types/net/description.tex +0 -0
@@ -531,6 +545,10 @@ \subsubsection{Packet

Transmission}\label{sec:Device Types / Network Device / De \field{flags} to zero and SHOULD supply a fully checksummed packet to the device.

device-types/net/description.tex +4 -0
+If the VIRTIO_NET_F_OUT_NET_HEADER feature has been negotiated, the
+driver MUST set \field{out_nh_offset} to point outer network header
+start, otherwise to zero.
+

Need to make it more verbose that, only when there is L3 header.
If it is some packet that does not need L3 header offset, the offset can be omitted. I believe we need a flag bit in the vnet_hdr.flags to indicate that header offset is valid.

Quoted 26 lines

If VIRTIO_NET_F_HOST_TSO4 is negotiated, the driver MAY set
\field{gso_type} to VIRTIO_NET_HDR_GSO_TCPV4 to request TCPv4 segmentation, otherwise the driver MUST NOT set @@ -596,6 +614,7 @@ \subsubsection{Packet Transmission}\label{sec:Device Types / Network Device / De
the driver SHOULD set \field{hdr_len} to a value
not less than the length of the headers, including the transport header.
+
\end{itemize}

The driver SHOULD accept the VIRTIO_NET_F_GUEST_HDRLEN feature if it
has @@ -610,6 +629,10 @@ \subsubsection{Packet Transmission}\label{sec:Device Types / Network Device / De If VIRTIO_NET_HDR_F_NEEDS_CSUM bit in \field{flags} is not set, the device MUST NOT use the \field{csum_start} and \field{csum_offset}.

device-types/net/description.tex +4 -0
+If the VIRTIO_NET_F_OUT_NET_HEADER feature has been negotiated, the
+device MAY use \field{out_nh_offset} as the outer network header
+offset, otherwise device MUST NOT use the \field{out_nh_offset} value.
+
 If one of the VIRTIO_NET_F_HOST_TSO4, TSO6, USO or UFO options have

been negotiated:
\begin{itemize}

device-types/net/description.tex +0 -0
@@ -631,6 +654,7 @@ \subsubsection{Packet

Transmission}\label{sec:Device Types / Network Device / De

Quoted 4 lines
      \begin{note}
      This is due to various bugs in implementations.
      \end{note}

+

Please remove this blank line.

Quoted 5 lines

\end{itemize}

If VIRTIO_NET_HDR_F_NEEDS_CSUM is not set, the device MUST NOT
--
2.43.0

#2 lore raw
In reply to Kommula Shiva Shankar
Quoted 103 lines

From: Kommula Shiva Shankar <[email protected]>
Sent: Tuesday, January 14, 2025 10:47 PM

Hi Everyone,
It has been some time since this patch was sent. I am resending it for your review and look forward to your feedback.

This patch introduces a new field in the virtio_net_header called out_nh_offset, along with a new net device feature, VIRTIO_NET_F_OUT_NET_HEADER.

Currently, there is no field available to directly read the outer network header offset in case of segmentation offload.
This requires reading packet data, which significantly affects performance in datapath.
Additionally, some hardware implementations requrie knowledege of the outer L3 offset (aka L2 length) for inline IPSec hardware acceleration.

To address this limitation, we propose splitting the csum_offset field into two 8-bit fields named csum_offset and out_nh_offset.
The csum_offset indicates the offset value from the csum_start and may not exceed 256B bits(2^8) for protocols that use a 16-bit one's complement checksum

Following table lists such protocols and their checksum offset fields within their headers

+-----+--------+
|Proto|csum_off|
+-----+--------+
| IPV4| 10B    |
| ICMP| 2B     |
| IGMP| 2B     |
|  TCP| 16B    |
|  UDP| 6B     |
+-----+--------+

The out_nh_offset represents the start offset of the outer network header from the beginning of the packet data

This issue was briefly discussed on the mailing list in a different thread, which can be found here
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flore.
kernel.org%2Fall%2FDM4PR18MB4269FAAC3CFC7E57E25DFBD2DF8B2%40 DM4PR18MB4269.namprd18.prod.outlook.com%2F&data=05%7C02%7Cpa rav%40nvidia.com%7Ce81096db39a14378627708dd34bf3909%7C43083d 15727340c1b7db39efd9ccc17a%7C0%7C0%7C638724718122375976%7C Unknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMD AwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7 C%7C&sdata=THSOD9fclucwxtnizGLo8pIpjsTs1YoRww4VA%2FvDwgI%3D&re served=0

---
 device-types/net/description.tex | 26 +++++++++++++++++++++++++-
 1 file changed, 25 insertions(+), 1 deletion(-)
device- renamed from device-types/net/description.tex +0 -0

types/net/description.tex
index 76585b0..48c081b 100644

device-types/net/description.tex +0 -0
@@ -88,6 +88,10 @@ \subsection{Feature bits}\label{sec:Device Types /

Network Device / Feature bits \item[VIRTIO_NET_F_CTRL_MAC_ADDR(23)] Set MAC address through control
channel.

+\item[VIRTIO_NET_F_OUT_NET_HEADER(50)] Driver can provide the start of \field{out_nh_offset}
+ value. Device gains advantage by not reading packet data to calculate outer
network

device-types/net/description.tex +2 -0
+    header offset.
+
 \item[VIRTIO_NET_F_HASH_TUNNEL(51)] Device supports inner header hash

for encapsulated packets.

\item[VIRTIO_NET_F_VQ_NOTF_COAL(52)] Device supports virtqueue
notification coalescing.

device-types/net/description.tex +0 -0
@@ -418,7 +422,13 @@ \subsection{Device Operation}\label{sec:Device

Types / Network Device / Device O

         le16 hdr_len;
         le16 gso_size;
         le16 csum_start;
device-types/net/description.tex +7 -1
-        le16 csum_offset;
+        union {
+                le16 csum_offset;
+                struct { (Only if VIRTIO_NET_F_OUT_NET_HEADER negotiated)
+                        le8 csum_offset;
+                        le8 out_nh_offset;
+                };
+        };
         le16 num_buffers;
         le32 hash_value;        (Only if VIRTIO_NET_F_HASH_REPORT negotiated)
         le16 hash_report;       (Only if VIRTIO_NET_F_HASH_REPORT negotiated)
@@ -457,6 +467,10 @@ \subsubsection{Packet

Transmission}\label{sec:Device Types / Network Device / De \item The driver can send a completely checksummed packet. In this case,
\field{flags} will be zero, and \field{gso_type} will be
VIRTIO_NET_HDR_GSO_NONE.

+\item If the driver negotiated VIRTIO_NET_F_OUT_NET_HEADER,

device-types/net/description.tex \subsubsection{Packet Transmission}\label{sec:Device Types / Network Device / De · line 473

Quoted 3 lines
device-types/net/description.tex +3 -0
+   \field{out_nh_offset} indicates the outer network header offset in bytes
+    from the beginning of the packet.
+

This field is mostly useful to device when there is GSO or checksum offload. So this should be set optionally by the driver.
When feature is negotiated, it means that csum_offset is 8 bits in width. A nw header offset may/may not be present.
When offset = 0, its not present (because 0 is impossible value for nw header). When its non zero, it is valid value that device should use.

Can you please update the proposal to cover these aspects?

Also, this may find its use even on the rx side too, so that sw does not need to parse it packet. So describe the text for it similarly.

#3 lore raw
Quoted 140 lines

-----Original Message-----
From: Parav Pandit <[email protected]>
Sent: Sunday, January 19, 2025 3:08 PM
To: Shiva Shankar Kommula <[email protected]>; [email protected]; [email protected]; [email protected]; virtio- [email protected]
Cc: Nithin Kumar Dabilpuram <[email protected]>; Jerin Jacob <[email protected]>; Rahul Bhansali <[email protected]>
Subject: [EXTERNAL] RE: [PATCH] virtio-net: Introduce a new field to indicate outer network header offset

Quoted 44 lines

From: Kommula Shiva Shankar <kshankar@ marvell. com> > Sent: Tuesday, January 14, 2025 10: 47 PM > > Hi Everyone, > It has been some time since this patch was sent. I am resending it for your > review and look forward

From: Kommula Shiva Shankar <[email protected]>
Sent: Tuesday, January 14, 2025 10:47 PM

Hi Everyone,
It has been some time since this patch was sent. I am resending it for your review and look forward to your feedback.

This patch introduces a new field in the virtio_net_header called out_nh_offset, along with a new net device feature, VIRTIO_NET_F_OUT_NET_HEADER.

Currently, there is no field available to directly read the outer network header offset in case of segmentation offload.
This requires reading packet data, which significantly affects performance in datapath.
Additionally, some hardware implementations requrie knowledege of the outer L3 offset (aka L2 length) for inline IPSec hardware acceleration.

To address this limitation, we propose splitting the csum_offset field into two 8-bit fields named csum_offset and out_nh_offset. The csum_offset indicates the offset value from the csum_start and may not exceed 256B bits(2^8) for protocols that use a 16-bit one's complement checksum

Following table lists such protocols and their checksum offset fields within their headers

+-----+--------+
|Proto|csum_off|
+-----+--------+
| IPV4| 10B    |
| ICMP| 2B     |
| IGMP| 2B     |
|  TCP| 16B    |
|  UDP| 6B     |
+-----+--------+
Quoted 6 lines

The out_nh_offset represents the start offset of the outer network header from the beginning of the packet data

This issue was briefly discussed on the mailing list in a different thread, which can be found here
https://urldefense.proofpoint.com/v2/url?u=https-

3A__nam11.safelinks.protection.outlook.com_-3Furl-3Dhttps-253A-252F- 252Flore&d=DwIFAg&c=nKjWec2b6R0mOyPaz7xtfQ&r=XP728rR_yI3bCCyFq0 MEWPS2176D_oSryc7cD-0TRPs&m=TIEwJZ_xRsOSm7KLONVdffWXkShROf- Sdyg3IKBNiKzIV8csp8rKVRdL0RlLde13&s=6cECQ7qUASslb35Xab87t1Kd3kFJyqK 9iT9KzbPMN0Y&e=.

Quoted 29 lines

kernel.org%2Fall%2FDM4PR18MB4269FAAC3CFC7E57E25DFBD2DF8B2%40 DM4PR18MB4269.namprd18.prod.outlook.com%2F&data=05%7C02%7Cpa rav%40nvidia.com%7Ce81096db39a14378627708dd34bf3909%7C43083d 15727340c1b7db39efd9ccc17a%7C0%7C0%7C638724718122375976%7C Unknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMD AwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7 C%7C&sdata=THSOD9fclucwxtnizGLo8pIpjsTs1YoRww4VA%2FvDwgI%3D&re served=0

---
 device-types/net/description.tex | 26 +++++++++++++++++++++++++-
 1 file changed, 25 insertions(+), 1 deletion(-)
device- renamed from device-types/net/description.tex +0 -0

types/net/description.tex index 76585b0..48c081b 100644

device-types/net/description.tex +0 -0
@@ -88,6 +88,10 @@ \subsection{Feature bits}\label{sec:Device Types /

Network Device / Feature bits \item[VIRTIO_NET_F_CTRL_MAC_ADDR(23)] Set MAC address through control
channel.

+\item[VIRTIO_NET_F_OUT_NET_HEADER(50)] Driver can provide the start +of
\field{out_nh_offset}
+ value. Device gains advantage by not reading packet data to
+ calculate outer
network

device-types/net/description.tex \subsection{Feature bits}\label{sec:Device Types / Network Device / Feature bits

Quoted 3 lines
device-types/net/description.tex +2 -0
+    header offset.
+
 \item[VIRTIO_NET_F_HASH_TUNNEL(51)] Device supports inner header

hash

Quoted 9 lines

for encapsulated packets.

\item[VIRTIO_NET_F_VQ_NOTF_COAL(52)] Device supports virtqueue
notification coalescing.

device-types/net/description.tex +0 -0
@@ -418,7 +422,13 @@ \subsection{Device Operation}\label{sec:Device

Types / Network Device / Device O

         le16 hdr_len;
         le16 gso_size;
         le16 csum_start;

device-types/net/description.tex \subsection{Device Operation}\label{sec:Device Types / Network Device / Device O

Quoted 10 lines
device-types/net/description.tex +7 -1
-        le16 csum_offset;
+        union {
+                le16 csum_offset;
+                struct { (Only if VIRTIO_NET_F_OUT_NET_HEADER negotiated)
+                        le8 csum_offset;
+                        le8 out_nh_offset;
+                };
+        };
         le16 num_buffers;
         le32 hash_value;        (Only if VIRTIO_NET_F_HASH_REPORT

negotiated)

Quoted 1 line

le16 hash_report; (Only if VIRTIO_NET_F_HASH_REPORT

negotiated)

device-types/net/description.tex \subsubsection{Packet Transmission}\label{sec:Device Types / Network Device / De · line 473

Quoted 10 lines
device-types/net/description.tex +0 -0
@@ -457,6 +467,10 @@ \subsubsection{Packet

Transmission}\label{sec:Device Types / Network Device / De \item The driver can send a completely checksummed packet. In this case,
\field{flags} will be zero, and \field{gso_type} will be
VIRTIO_NET_HDR_GSO_NONE.

+\item If the driver negotiated VIRTIO_NET_F_OUT_NET_HEADER,

device-types/net/description.tex +3 -0
+   \field{out_nh_offset} indicates the outer network header offset in bytes
+    from the beginning of the packet.
+

This field is mostly useful to device when there is GSO or checksum offload. So this should be set optionally by the driver.
When feature is negotiated, it means that csum_offset is 8 bits in width. A nw header offset may/may not be present.
When offset = 0, its not present (because 0 is impossible value for nw header).
When its non zero, it is valid value that device should use.

Quoted 5 lines

Can you please update the proposal to cover these aspects?

Also, this may find its use even on the rx side too, so that sw does not need to parse it packet.
So describe the text for it similarly.

FWIW, while processing incoming packets, when the VIRTIO_NET_HDR_F_NEEDS_CSUM flag is set by device, the csum_start and csum_offset fields indicate how to calculate the checksum, which still involves reading packet data in software.
I'm unsure how this benefit in the RX path. Could you please clarify ?