Skip to contentSkip to conversation

virtio-net: Fix to avoid using reserved feature bits

PATCH In review Parav Pandit 6 participants 36 messages Jan 26, 2025 – May 19, 2025

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 virtio-net: Fix to avoid using reserved feature bits

Listed patches in the fixes tag, incorrectly used the reserved feature bits. Fix them to use the well defined device specific range.

Fixes 8cd457d8aa82 ("virtio-net: define UDP tunnel segmentation offload feature")
Fixes 3fea589bd7c6 ("virtio-net: define UDP tunnel checksum offload feature")

---
changelog:
v0->v1:
- added Cornelia's reviewed-by
- added github issue number
- skipped Paolo's suggestion to modify 'le64 offloads' as its only 64-bits and new feature bits are in > 64 bits range

---
 device-types/net/description.tex | 24 ++++++++++++------------
 1 file changed, 12 insertions(+), 12 deletions(-)
device-types/net/description.tex +12 -12
index efb543f..b546cd7 100644
@@ -88,18 +88,6 @@ \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_GUEST_UDP_TUNNEL_GSO (46)] Driver can receive GSO packets
-  carried by a UDP tunnel.
-
-\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_CSUM (47)] Driver handles packets
-  carried by a UDP tunnel with partial csum for the outer header.
-
-\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO (48)] Device can receive GSO packets
-  carried by a UDP tunnel.
-
-\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO_CSUM (49)] Device handles packets
-  carried by a UDP tunnel with partial csum for the outer header.
-
 \item[VIRTIO_NET_F_DEVICE_STATS(50)] Device can provide device-level statistics
     to the driver through the control virtqueue.
 
@@ -136,6 +124,18 @@ \subsection{Feature bits}\label{sec:Device Types / Network Device / Feature bits
 \item[VIRTIO_NET_F_SPEED_DUPLEX(63)] Device reports speed and duplex.
 
 \item[VIRTIO_NET_F_RSS_CONTEXT(64)] Device supports multiple RSS contexts.
+
+\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO (65)] Driver can receive GSO packets
+  carried by a UDP tunnel.
+
+\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_CSUM (66)] Driver handles packets
+  carried by a UDP tunnel with partial csum for the outer header.
+
+\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO (67)] Device can receive GSO packets
+  carried by a UDP tunnel.
+
+\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO_CSUM (68)] Device handles packets
+  carried by a UDP tunnel with partial csum for the outer header.
 \end{description}
 
 \subsubsection{Feature bit requirements}\label{sec:Device Types / Network Device / Feature bits / Feature bit requirements}
Signature · 1 line
2.34.1
#1 lore raw
Quoted 7 lines · Parav Pandit

Listed patches in the fixes tag, incorrectly used the reserved feature bits. Fix them to use the well defined device specific range.

Fixes 8cd457d8aa82 ("virtio-net: define UDP tunnel segmentation offload feature")
Fixes 3fea589bd7c6 ("virtio-net: define UDP tunnel checksum offload feature")

Well the hashes will be wrong I think.
So, we can do the ballot for this fixup, and wait with applying the patches until that new ballot completes. Or we can do a completely new ballot.

Quoted 58 lines

---
changelog:
v0->v1:
- added Cornelia's reviewed-by
- added github issue number
- skipped Paolo's suggestion to modify 'le64 offloads' as its only 64-bits and new feature bits are in > 64 bits range

---
 device-types/net/description.tex | 24 ++++++++++++------------
 1 file changed, 12 insertions(+), 12 deletions(-)
device-types/net/description.tex +12 -12
index efb543f..b546cd7 100644
@@ -88,18 +88,6 @@ \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_GUEST_UDP_TUNNEL_GSO (46)] Driver can receive GSO packets
-  carried by a UDP tunnel.
-
-\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_CSUM (47)] Driver handles packets
-  carried by a UDP tunnel with partial csum for the outer header.
-
-\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO (48)] Device can receive GSO packets
-  carried by a UDP tunnel.
-
-\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO_CSUM (49)] Device handles packets
-  carried by a UDP tunnel with partial csum for the outer header.
-
 \item[VIRTIO_NET_F_DEVICE_STATS(50)] Device can provide device-level statistics
     to the driver through the control virtqueue.
 
@@ -136,6 +124,18 @@ \subsection{Feature bits}\label{sec:Device Types / Network Device / Feature bits
 \item[VIRTIO_NET_F_SPEED_DUPLEX(63)] Device reports speed and duplex.
 
 \item[VIRTIO_NET_F_RSS_CONTEXT(64)] Device supports multiple RSS contexts.
+
+\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO (65)] Driver can receive GSO packets
+  carried by a UDP tunnel.
+
+\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_CSUM (66)] Driver handles packets
+  carried by a UDP tunnel with partial csum for the outer header.
+
+\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO (67)] Device can receive GSO packets
+  carried by a UDP tunnel.
+
+\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO_CSUM (68)] Device handles packets
+  carried by a UDP tunnel with partial csum for the outer header.
 \end{description}
 
 \subsubsection{Feature bit requirements}\label{sec:Device Types / Network Device / Feature bits / Feature bit requirements}
Signature · 1 line
2.34.1
#2 lore raw
Quoted 17 lines

From: Michael S. Tsirkin <[email protected]>
Sent: Sunday, January 26, 2025 2:50 PM

Quoted 9 lines · Parav Pandit

Listed patches in the fixes tag, incorrectly used the reserved feature bits. Fix them to use the well defined device specific range.

Fixes 8cd457d8aa82 ("virtio-net: define UDP tunnel segmentation

offload feature")

Fixes 3fea589bd7c6 ("virtio-net: define UDP tunnel checksum offload

feature")

Quoted 1 line

Well the hashes will be wrong I think.

I pulled the patches from the mailing list to local repo using b4.
Once the new fix is voted, I will merge both the series together so that hashes stays the same.

Quoted 2 lines

So, we can do the ballot for this fixup, and wait with applying the patches until that new ballot completes.

Yes. The small fix is reviewed by Paolo and Cornelia, so if it looks ok, we can wait for this week for any additional review comments. And merge by Feb-10 or so after new ballot completes.

Quoted 82 lines

Or we can do a completely new ballot.

Quoted 20 lines

---
changelog:
v0->v1:
- added Cornelia's reviewed-by
- added github issue number
- skipped Paolo's suggestion to modify 'le64 offloads' as its only 64-bits and new feature bits are in > 64 bits range

---
 device-types/net/description.tex | 24 ++++++++++++------------
 1 file changed, 12 insertions(+), 12 deletions(-)
device-types/net/description.tex +0 -0
index efb543f..b546cd7 100644
@@ -88,18 +88,6 @@ \subsection{Feature bits}\label{sec:Device Types /
Quoted 1 line

Network Device / Feature bits \item[VIRTIO_NET_F_CTRL_MAC_ADDR(23)]

Set MAC address through control

Quoted 3 lines

channel.

-\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO (46)] Driver can receive

GSO

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

Quoted 4 lines

packets

device-types/net/description.tex +0 -3
-  carried by a UDP tunnel.
-
-\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_CSUM (47)] Driver

handles

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

Quoted 4 lines

packets

device-types/net/description.tex +0 -3
-  carried by a UDP tunnel with partial csum for the outer header.
-
-\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO (48)] Device can receive

GSO

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

Quoted 4 lines

packets

device-types/net/description.tex +0 -3
-  carried by a UDP tunnel.
-
-\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO_CSUM (49)] Device

handles

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

Quoted 4 lines

packets

device-types/net/description.tex +0 -2
-  carried by a UDP tunnel with partial csum for the outer header.
-
 \item[VIRTIO_NET_F_DEVICE_STATS(50)] Device can provide device-level

statistics

Quoted 4 lines

to the driver through the control virtqueue.

device-types/net/description.tex +0 -0
@@ -136,6 +124,18 @@ \subsection{Feature bits}\label{sec:Device Types

/ Network Device / Feature bits \item[VIRTIO_NET_F_SPEED_DUPLEX(63)]

Device reports speed and duplex.

Quoted 2 lines

\item[VIRTIO_NET_F_RSS_CONTEXT(64)] Device supports multiple RSS

contexts.

Quoted 2 lines

+
+\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO (65)] Driver can receive

GSO

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

Quoted 4 lines
device-types/net/description.tex +4 -0
+packets
+  carried by a UDP tunnel.
+
+\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_CSUM (66)] Driver

handles

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

Quoted 4 lines
device-types/net/description.tex +4 -0
+packets
+  carried by a UDP tunnel with partial csum for the outer header.
+
+\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO (67)] Device can receive

GSO

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

Quoted 4 lines
device-types/net/description.tex +4 -0
+packets
+  carried by a UDP tunnel.
+
+\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO_CSUM (68)] Device

handles

Quoted 8 lines
device-types/net/description.tex +2 -0
+packets
+  carried by a UDP tunnel with partial csum for the outer header.
 \end{description}
 
 \subsubsection{Feature bit requirements}\label{sec:Device Types /

Network Device / Feature bits / Feature bit requirements} --
2.34.1

#3 lore raw
Quoted 17 lines · Parav Pandit
Quoted 17 lines

From: Michael S. Tsirkin <[email protected]>
Sent: Sunday, January 26, 2025 2:50 PM

Quoted 9 lines · Parav Pandit

Listed patches in the fixes tag, incorrectly used the reserved feature bits. Fix them to use the well defined device specific range.

Fixes 8cd457d8aa82 ("virtio-net: define UDP tunnel segmentation

offload feature")

Fixes 3fea589bd7c6 ("virtio-net: define UDP tunnel checksum offload

feature")

Quoted 3 lines
Quoted 1 line

Well the hashes will be wrong I think.

I pulled the patches from the mailing list to local repo using b4.
Once the new fix is voted, I will merge both the series together so that hashes stays the same.

ok. just make sure to verify them before you push.

Quoted 4 lines
Quoted 2 lines

So, we can do the ballot for this fixup, and wait with applying the patches until that new ballot completes.

Yes. The small fix is reviewed by Paolo and Cornelia, so if it looks ok, we can wait for this week for any additional review comments. And merge by Feb-10 or so after new ballot completes.

Quoted 82 lines
Quoted 82 lines

Or we can do a completely new ballot.

Quoted 20 lines

---
changelog:
v0->v1:
- added Cornelia's reviewed-by
- added github issue number
- skipped Paolo's suggestion to modify 'le64 offloads' as its only 64-bits and new feature bits are in > 64 bits range

---
 device-types/net/description.tex | 24 ++++++++++++------------
 1 file changed, 12 insertions(+), 12 deletions(-)
device-types/net/description.tex +0 -0
index efb543f..b546cd7 100644
@@ -88,18 +88,6 @@ \subsection{Feature bits}\label{sec:Device Types /
Quoted 1 line

Network Device / Feature bits \item[VIRTIO_NET_F_CTRL_MAC_ADDR(23)]

Set MAC address through control

Quoted 3 lines

channel.

-\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO (46)] Driver can receive

GSO

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

Quoted 4 lines

packets

device-types/net/description.tex +0 -3
-  carried by a UDP tunnel.
-
-\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_CSUM (47)] Driver

handles

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

Quoted 4 lines

packets

device-types/net/description.tex +0 -3
-  carried by a UDP tunnel with partial csum for the outer header.
-
-\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO (48)] Device can receive

GSO

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

Quoted 4 lines

packets

device-types/net/description.tex +0 -3
-  carried by a UDP tunnel.
-
-\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO_CSUM (49)] Device

handles

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

Quoted 4 lines

packets

device-types/net/description.tex +0 -2
-  carried by a UDP tunnel with partial csum for the outer header.
-
 \item[VIRTIO_NET_F_DEVICE_STATS(50)] Device can provide device-level

statistics

Quoted 4 lines

to the driver through the control virtqueue.

device-types/net/description.tex +0 -0
@@ -136,6 +124,18 @@ \subsection{Feature bits}\label{sec:Device Types

/ Network Device / Feature bits \item[VIRTIO_NET_F_SPEED_DUPLEX(63)]

Device reports speed and duplex.

Quoted 2 lines

\item[VIRTIO_NET_F_RSS_CONTEXT(64)] Device supports multiple RSS

contexts.

Quoted 2 lines

+
+\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO (65)] Driver can receive

GSO

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

Quoted 4 lines
device-types/net/description.tex +4 -0
+packets
+  carried by a UDP tunnel.
+
+\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_CSUM (66)] Driver

handles

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

Quoted 4 lines
device-types/net/description.tex +4 -0
+packets
+  carried by a UDP tunnel with partial csum for the outer header.
+
+\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO (67)] Device can receive

GSO

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

Quoted 4 lines
device-types/net/description.tex +4 -0
+packets
+  carried by a UDP tunnel.
+
+\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO_CSUM (68)] Device

handles

Quoted 8 lines
device-types/net/description.tex +2 -0
+packets
+  carried by a UDP tunnel with partial csum for the outer header.
 \end{description}
 
 \subsubsection{Feature bit requirements}\label{sec:Device Types /

Network Device / Feature bits / Feature bit requirements} --
2.34.1

#4 lore raw
Quoted 24 lines · Michael S. Tsirkin
Quoted 17 lines · Parav Pandit
Quoted 17 lines

From: Michael S. Tsirkin <[email protected]>
Sent: Sunday, January 26, 2025 2:50 PM

Quoted 9 lines · Parav Pandit

Listed patches in the fixes tag, incorrectly used the reserved feature bits. Fix them to use the well defined device specific range.

Fixes 8cd457d8aa82 ("virtio-net: define UDP tunnel segmentation

offload feature")

Fixes 3fea589bd7c6 ("virtio-net: define UDP tunnel checksum offload

feature")

Quoted 3 lines
Quoted 1 line

Well the hashes will be wrong I think.

I pulled the patches from the mailing list to local repo using b4.
Once the new fix is voted, I will merge both the series together so that hashes stays the same.

ok. just make sure to verify them before you push.

Quoted 7 lines
Quoted 4 lines
Quoted 2 lines

So, we can do the ballot for this fixup, and wait with applying the patches until that new ballot completes.

Yes. The small fix is reviewed by Paolo and Cornelia, so if it looks ok, we can wait for this week for any additional review comments. And merge by Feb-10 or so after new ballot completes.

As I don't see this change as particularly contentious, I'd suggest to just go ahead and start the voting process -- there's plenty of time to voice concerns while the ballot is open.

#5 lore raw

Hi Matias, Michael, Cornelia,

Quoted 39 lines

From: Cornelia Huck <[email protected]>
Sent: Monday, January 27, 2025 2:52 PM
To: Michael S. Tsirkin <[email protected]>; Parav Pandit <[email protected]>
Cc: [email protected]; [email protected]; Shahaf Shuler <[email protected]>
Subject: Re: [PATCH v1] virtio-net: Fix to avoid using reserved feature bits

Quoted 19 lines · Michael S. Tsirkin
Quoted 18 lines · Parav Pandit
Quoted 18 lines

From: Michael S. Tsirkin <[email protected]>
Sent: Sunday, January 26, 2025 2:50 PM

Quoted 10 lines · Parav Pandit

Listed patches in the fixes tag, incorrectly used the reserved feature bits. Fix them to use the well defined device specific range.

Fixes 8cd457d8aa82 ("virtio-net: define UDP tunnel segmentation

offload feature")

Fixes 3fea589bd7c6 ("virtio-net: define UDP tunnel checksum

offload
feature")

Quoted 3 lines
Quoted 3 lines
Quoted 1 line

Well the hashes will be wrong I think.

I pulled the patches from the mailing list to local repo using b4. Once the new fix is voted, I will merge both the series together so that

hashes stays the same.

Quoted 6 lines

ok. just make sure to verify them before you push.

Quoted 3 lines
Quoted 2 lines

So, we can do the ballot for this fixup, and wait with applying the patches until that new ballot completes.

Yes. The small fix is reviewed by Paolo and Cornelia, so if it looks ok, we can

wait for this week for any additional review comments.

Quoted 8 lines
Quoted 4 lines
Quoted 1 line

And merge by Feb-10 or so after new ballot completes.

As I don't see this change as particularly contentious, I'd suggest to just go ahead and start the voting process -- there's plenty of time to voice concerns while the ballot is open.

Can you please open the voting ballot for https://github.com/oasis-tcs/virtio-spec/issues/213 ? The proposed patch is: https://lore.kernel.org/virtio-comment/[email protected]/T/#u

#6 lore raw
In reply to Parav Pandit
Quoted 7 lines · Parav Pandit
device-types/net/description.tex +3 -0
@@ -136,6 +124,18 @@ \subsection{Feature bits}\label{sec:Device Types / Network Device / Feature bits
 \item[VIRTIO_NET_F_SPEED_DUPLEX(63)] Device reports speed and duplex.
 
 \item[VIRTIO_NET_F_RSS_CONTEXT(64)] Device supports multiple RSS contexts.
+
+\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO (65)] Driver can receive GSO packets
+  carried by a UDP tunnel.

device-types/net/description.tex \subsection{Feature bits}\label{sec:Device Types / Network Device / Feature bits · line 139

Quoted 10 lines
device-types/net/description.tex +9 -0
+
+\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_CSUM (66)] Driver handles packets
+  carried by a UDP tunnel with partial csum for the outer header.
+
+\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO (67)] Device can receive GSO packets
+  carried by a UDP tunnel.
+
+\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO_CSUM (68)] Device handles packets
+  carried by a UDP tunnel with partial csum for the outer header.
 \end{description}

I'm finally trying to finalize the implementation and I must admit I underlooked the implication of using features bit >= 64.

AFAICS all the virtio drivers in the Linux kernel and qemu assume that the features can be represented with a single u64.

Supporting the above requires major reworks in both user-space and kernel space. At least on the kernel side a similar rework (extending the _netdev_ features field) has proven to be complex enough to defeat any implementation attempts for years and in the end it was dismissed in favor of lower bits reuse.

I'm wondering if we could reconsider using reserved bits here?

Thanks!

Paolo

#7 lore raw
Quoted 31 lines · Paolo Abeni
Quoted 7 lines · Parav Pandit
device-types/net/description.tex +3 -0
@@ -136,6 +124,18 @@ \subsection{Feature bits}\label{sec:Device Types / Network Device / Feature bits
 \item[VIRTIO_NET_F_SPEED_DUPLEX(63)] Device reports speed and duplex.
 
 \item[VIRTIO_NET_F_RSS_CONTEXT(64)] Device supports multiple RSS contexts.
+
+\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO (65)] Driver can receive GSO packets
+  carried by a UDP tunnel.

device-types/net/description.tex \subsection{Feature bits}\label{sec:Device Types / Network Device / Feature bits · line 139

Quoted 10 lines
device-types/net/description.tex +9 -0
+
+\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_CSUM (66)] Driver handles packets
+  carried by a UDP tunnel with partial csum for the outer header.
+
+\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO (67)] Device can receive GSO packets
+  carried by a UDP tunnel.
+
+\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO_CSUM (68)] Device handles packets
+  carried by a UDP tunnel with partial csum for the outer header.
 \end{description}

I'm finally trying to finalize the implementation and I must admit I underlooked the implication of using features bit >= 64.

AFAICS all the virtio drivers in the Linux kernel and qemu assume that the features can be represented with a single u64.

Supporting the above requires major reworks in both user-space and kernel space. At least on the kernel side a similar rework (extending the _netdev_ features field) has proven to be complex enough to defeat any implementation attempts for years and in the end it was dismissed in favor of lower bits reuse.

Quoted 5 lines

I'm wondering if we could reconsider using reserved bits here?

Thanks!

Paolo

:(

It's been there for 2 months now.

And if you want to avoid using all bits > 63 then it's even more of a mess, you need to move VIRTIO_NET_F_RSS_CONTEXT which has been there for a year and a half.

I'm afraid we'll have to bite the bullet.

Signature · 1 line
MST
#8 lore raw
Quoted 34 lines · Michael S. Tsirkin
Quoted 33 lines · Paolo Abeni
Quoted 7 lines · Parav Pandit
device-types/net/description.tex +3 -0
@@ -136,6 +124,18 @@ \subsection{Feature bits}\label{sec:Device Types / Network Device / Feature bits
 \item[VIRTIO_NET_F_SPEED_DUPLEX(63)] Device reports speed and duplex.
 
 \item[VIRTIO_NET_F_RSS_CONTEXT(64)] Device supports multiple RSS contexts.
+
+\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO (65)] Driver can receive GSO packets
+  carried by a UDP tunnel.

device-types/net/description.tex \subsection{Feature bits}\label{sec:Device Types / Network Device / Feature bits · line 139

Quoted 10 lines
device-types/net/description.tex +9 -0
+
+\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_CSUM (66)] Driver handles packets
+  carried by a UDP tunnel with partial csum for the outer header.
+
+\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO (67)] Device can receive GSO packets
+  carried by a UDP tunnel.
+
+\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO_CSUM (68)] Device handles packets
+  carried by a UDP tunnel with partial csum for the outer header.
 \end{description}

I'm finally trying to finalize the implementation and I must admit I underlooked the implication of using features bit >= 64.

AFAICS all the virtio drivers in the Linux kernel and qemu assume that the features can be represented with a single u64.

Supporting the above requires major reworks in both user-space and kernel space. At least on the kernel side a similar rework (extending the _netdev_ features field) has proven to be complex enough to defeat any implementation attempts for years and in the end it was dismissed in favor of lower bits reuse.

I'm wondering if we could reconsider using reserved bits here?

Quoted 7 lines
Quoted 3 lines

Thanks!

Paolo

:(

It's been there for 2 months now.

Unfortunately I was not able to get back earlier.

Quoted 3 lines

And if you want to avoid using all bits > 63 then it's even more of a mess, you need to move VIRTIO_NET_F_RSS_CONTEXT which has been there for a year and a half.

I started the 'support [arbitrary] larger features bitmap' exercise and I think there is a difference between the VIRTIO_NET_F_RSS_CONTEXT bit and UDP_TUNNEL_GSO related ones: AFAICS the latter go through the VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET command, which in turn limit the offload to 64 bit:

"""
The class VIRTIO_NET_CTRL_GUEST_OFFLOADS has one command:
VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET applies the new offloads configuration.

le64 value passed as command data is a bitmask, bits set define offloads to be enabled, bits cleared - offloads to be disabled. """

while I don't see similar constraints for the VIRTIO_NET_F_RSS_CONTEXT feature.

The current specification is inconsistent, as it does not allow enabling the UDP_TUNNEL guest offloads.

AFAICS we have to either add the specification for another VIRTIO_NET_CTRL_GUEST_OFFLOADS command to allow passing the additional data or reconsider using reserved bits just for the UDP_TUNNEL offload (I'm sorry for reiterating on this, it still looks like a very tempting path).

Thanks,

Paolo

#9 lore raw
Quoted 63 lines · Paolo Abeni
Quoted 34 lines · Michael S. Tsirkin
Quoted 33 lines · Paolo Abeni
Quoted 7 lines · Parav Pandit
device-types/net/description.tex +3 -0
@@ -136,6 +124,18 @@ \subsection{Feature bits}\label{sec:Device Types / Network Device / Feature bits
 \item[VIRTIO_NET_F_SPEED_DUPLEX(63)] Device reports speed and duplex.
 
 \item[VIRTIO_NET_F_RSS_CONTEXT(64)] Device supports multiple RSS contexts.
+
+\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO (65)] Driver can receive GSO packets
+  carried by a UDP tunnel.

device-types/net/description.tex \subsection{Feature bits}\label{sec:Device Types / Network Device / Feature bits · line 139

Quoted 10 lines
device-types/net/description.tex +9 -0
+
+\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_CSUM (66)] Driver handles packets
+  carried by a UDP tunnel with partial csum for the outer header.
+
+\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO (67)] Device can receive GSO packets
+  carried by a UDP tunnel.
+
+\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO_CSUM (68)] Device handles packets
+  carried by a UDP tunnel with partial csum for the outer header.
 \end{description}

I'm finally trying to finalize the implementation and I must admit I underlooked the implication of using features bit >= 64.

AFAICS all the virtio drivers in the Linux kernel and qemu assume that the features can be represented with a single u64.

Supporting the above requires major reworks in both user-space and kernel space. At least on the kernel side a similar rework (extending the _netdev_ features field) has proven to be complex enough to defeat any implementation attempts for years and in the end it was dismissed in favor of lower bits reuse.

I'm wondering if we could reconsider using reserved bits here?

Quoted 7 lines
Quoted 3 lines

Thanks!

Paolo

:(

It's been there for 2 months now.

Unfortunately I was not able to get back earlier.

Quoted 3 lines

And if you want to avoid using all bits > 63 then it's even more of a mess, you need to move VIRTIO_NET_F_RSS_CONTEXT which has been there for a year and a half.

I started the 'support [arbitrary] larger features bitmap' exercise and I think there is a difference between the VIRTIO_NET_F_RSS_CONTEXT bit and UDP_TUNNEL_GSO related ones: AFAICS the latter go through the VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET command, which in turn limit the offload to 64 bit:

"""
The class VIRTIO_NET_CTRL_GUEST_OFFLOADS has one command:
VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET applies the new offloads configuration.

le64 value passed as command data is a bitmask, bits set define offloads to be enabled, bits cleared - offloads to be disabled. """

Quoted 5 lines

while I don't see similar constraints for the VIRTIO_NET_F_RSS_CONTEXT feature.

The current specification is inconsistent, as it does not allow enabling the UDP_TUNNEL guest offloads.

Through the command, yes.

Quoted 3 lines

AFAICS we have to either add the specification for another VIRTIO_NET_CTRL_GUEST_OFFLOADS command to allow passing the additional data

Given features are negotiated first, we can just say the value can be 64 or 128 bits? In any case, I agree it's buggy as defined.

Quoted 7 lines

or reconsider using reserved bits just for the UDP_TUNNEL offload (I'm sorry for reiterating on this, it still looks like a very tempting path).

Thanks,

Paolo

Parav what's your take on this one?

Signature · 1 line
MST
#10 lore raw

Hi,

I'll try to consolidate here the current discussion.

My understanding is that the we are on same WRT that there is a problem with the current spec.

Quoted 2 lines · Michael S. Tsirkin

Given features are negotiated first, we can just say the value can be 64 or 128 bits? In any case, I agree it's buggy as defined.

Do you mean that the 'offload' field should be 64 or 128 bits long depending on the negotiated features?

Could that be problematic with VM migrations and/or backward compatibility? I can't see a specific faulty scenario, but the 'implied' field size sounds fragile to me.

/P

#11 lore raw
Quoted 11 lines · Paolo Abeni

Hi,

I'll try to consolidate here the current discussion.

My understanding is that the we are on same WRT that there is a problem with the current spec.

Quoted 2 lines · Michael S. Tsirkin

Given features are negotiated first, we can just say the value can be 64 or 128 bits? In any case, I agree it's buggy as defined.

Quoted 8 lines

Do you mean that the 'offload' field should be 64 or 128 bits long depending on the negotiated features?

Could that be problematic with VM migrations and/or backward compatibility? I can't see a specific faulty scenario, but the 'implied' field size sounds fragile to me.

/P

My idea is to say any number of le64 values, and any bits excluded must be 0.

So looking at qemu for example, if you send it more than 64 bit it will just ignore the top 64 bit. But it does not currently support any low bits, so it's fine.

Again any change will break compat technicall but then any change to reserved bits does, too.

Signature · 1 line
MST
#12 lore raw
In reply to Michael S. Tsirkin
Quoted 41 lines · Michael S. Tsirkin
Quoted 31 lines · Paolo Abeni
Quoted 7 lines · Parav Pandit
device-types/net/description.tex +3 -0
@@ -136,6 +124,18 @@ \subsection{Feature bits}\label{sec:Device Types / Network Device / Feature bits
 \item[VIRTIO_NET_F_SPEED_DUPLEX(63)] Device reports speed and duplex.
 
 \item[VIRTIO_NET_F_RSS_CONTEXT(64)] Device supports multiple RSS contexts.
+
+\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO (65)] Driver can receive GSO packets
+  carried by a UDP tunnel.

device-types/net/description.tex \subsection{Feature bits}\label{sec:Device Types / Network Device / Feature bits · line 139

Quoted 10 lines
device-types/net/description.tex +9 -0
+
+\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_CSUM (66)] Driver handles packets
+  carried by a UDP tunnel with partial csum for the outer header.
+
+\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO (67)] Device can receive GSO packets
+  carried by a UDP tunnel.
+
+\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO_CSUM (68)] Device handles packets
+  carried by a UDP tunnel with partial csum for the outer header.
 \end{description}

I'm finally trying to finalize the implementation and I must admit I underlooked the implication of using features bit >= 64.

AFAICS all the virtio drivers in the Linux kernel and qemu assume that the features can be represented with a single u64.

Supporting the above requires major reworks in both user-space and kernel space. At least on the kernel side a similar rework (extending the _netdev_ features field) has proven to be complex enough to defeat any implementation attempts for years and in the end it was dismissed in favor of lower bits reuse.

Quoted 5 lines

I'm wondering if we could reconsider using reserved bits here?

Thanks!

Paolo

:(

Quoted 7 lines

It's been there for 2 months now.

And if you want to avoid using all bits > 63 then it's even more of a mess, you need to move VIRTIO_NET_F_RSS_CONTEXT which has been there for a year and a half.

I'm afraid we'll have to bite the bullet.

One other issue with bits > 63 is that the vhost-user protocol VHOST_USER_GET_FEATURES and VHOST_USER_SET_FEATURES messages use u64 to represent the features, so vhost-user-net devices can't query or enable these features. vhost-user is outside the scope of the virtio spec, though, and I think it's reasonable to extend the protocol to enable high feature bits rather than avoiding them forever.

Thanks,
-- Daniel

#13 lore raw
Quoted 57 lines · Daniel Verkamp
Quoted 41 lines · Michael S. Tsirkin
Quoted 31 lines · Paolo Abeni
Quoted 7 lines · Parav Pandit
device-types/net/description.tex +3 -0
@@ -136,6 +124,18 @@ \subsection{Feature bits}\label{sec:Device Types / Network Device / Feature bits
 \item[VIRTIO_NET_F_SPEED_DUPLEX(63)] Device reports speed and duplex.
 
 \item[VIRTIO_NET_F_RSS_CONTEXT(64)] Device supports multiple RSS contexts.
+
+\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO (65)] Driver can receive GSO packets
+  carried by a UDP tunnel.

device-types/net/description.tex \subsection{Feature bits}\label{sec:Device Types / Network Device / Feature bits · line 139

Quoted 10 lines
device-types/net/description.tex +9 -0
+
+\item[VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_CSUM (66)] Driver handles packets
+  carried by a UDP tunnel with partial csum for the outer header.
+
+\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO (67)] Device can receive GSO packets
+  carried by a UDP tunnel.
+
+\item[VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO_CSUM (68)] Device handles packets
+  carried by a UDP tunnel with partial csum for the outer header.
 \end{description}

I'm finally trying to finalize the implementation and I must admit I underlooked the implication of using features bit >= 64.

AFAICS all the virtio drivers in the Linux kernel and qemu assume that the features can be represented with a single u64.

Supporting the above requires major reworks in both user-space and kernel space. At least on the kernel side a similar rework (extending the _netdev_ features field) has proven to be complex enough to defeat any implementation attempts for years and in the end it was dismissed in favor of lower bits reuse.

Quoted 5 lines

I'm wondering if we could reconsider using reserved bits here?

Thanks!

Paolo

:(

Quoted 7 lines

It's been there for 2 months now.

And if you want to avoid using all bits > 63 then it's even more of a mess, you need to move VIRTIO_NET_F_RSS_CONTEXT which has been there for a year and a half.

I'm afraid we'll have to bite the bullet.

One other issue with bits > 63 is that the vhost-user protocol VHOST_USER_GET_FEATURES and VHOST_USER_SET_FEATURES messages use u64 to represent the features, so vhost-user-net devices can't query or enable these features. vhost-user is outside the scope of the virtio spec, though, and I think it's reasonable to extend the protocol to enable high feature bits rather than avoiding them forever.

Quoted 2 lines

Thanks,
-- Daniel

Yes you would use VHOST_USER_SET_PROTOCOL_FEATURES to make VHOST_USER_GET_FEATURES return two u64s, or even a new message returning an array.

#14 lore raw
Quoted 11 lines · Michael S. Tsirkin
Quoted 9 lines · Daniel Verkamp
Quoted 1 line · Michael S. Tsirkin

I'm afraid we'll have to bite the bullet.

One other issue with bits > 63 is that the vhost-user protocol VHOST_USER_GET_FEATURES and VHOST_USER_SET_FEATURES messages use u64 to represent the features, so vhost-user-net devices can't query or enable these features. vhost-user is outside the scope of the virtio spec, though, and I think it's reasonable to extend the protocol to enable high feature bits rather than avoiding them forever.

Quoted 2 lines

Yes you would use VHOST_USER_SET_PROTOCOL_FEATURES to make VHOST_USER_GET_FEATURES return two u64s, or even a new message returning an array.

I think that additionally the VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET command will need some clarification, as in the current text looks a bit inconsistent:

"""
// in Offloads State Configuration / Setting Offloads State:

#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO 46

// ...

The class VIRTIO_NET_CTRL_GUEST_OFFLOADS has one command:
VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET applies the new offloads configuration.

le64 value passed as command data is a bitmask, bits set define offloads to be enabled, bits cleared - offloads to be disabled.

There is a corresponding device feature for each offload. Upon feature negotiation corresponding offload gets enabled to preserve backward compatibility
"""

The "corresponding device feature" has the same numerical value of the selected offloads, except for UDP tunnels related one (which are mapped to bits corresponding to reserved features).

It's unclear to me which should be the better way to address this inconsistency.

/P

#15 lore raw
Quoted 37 lines · Paolo Abeni
Quoted 11 lines · Michael S. Tsirkin
Quoted 9 lines · Daniel Verkamp
Quoted 1 line · Michael S. Tsirkin

I'm afraid we'll have to bite the bullet.

One other issue with bits > 63 is that the vhost-user protocol VHOST_USER_GET_FEATURES and VHOST_USER_SET_FEATURES messages use u64 to represent the features, so vhost-user-net devices can't query or enable these features. vhost-user is outside the scope of the virtio spec, though, and I think it's reasonable to extend the protocol to enable high feature bits rather than avoiding them forever.

Quoted 2 lines

Yes you would use VHOST_USER_SET_PROTOCOL_FEATURES to make VHOST_USER_GET_FEATURES return two u64s, or even a new message returning an array.

I think that additionally the VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET command will need some clarification, as in the current text looks a bit inconsistent:

"""
// in Offloads State Configuration / Setting Offloads State:

#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO 46

// ...

The class VIRTIO_NET_CTRL_GUEST_OFFLOADS has one command:
VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET applies the new offloads configuration.

le64 value passed as command data is a bitmask, bits set define offloads to be enabled, bits cleared - offloads to be disabled.

There is a corresponding device feature for each offload. Upon feature negotiation corresponding offload gets enabled to preserve backward compatibility
"""

Quoted 8 lines

The "corresponding device feature" has the same numerical value of the selected offloads, except for UDP tunnels related one (which are mapped to bits corresponding to reserved features).

It's unclear to me which should be the better way to address this inconsistency.

/P

Could you pls explain what is inconsistent?

Signature · 1 line
MST
#16 lore raw
In reply to Paolo Abeni
Quoted 37 lines · Paolo Abeni
Quoted 11 lines · Michael S. Tsirkin
Quoted 9 lines · Daniel Verkamp
Quoted 1 line · Michael S. Tsirkin

I'm afraid we'll have to bite the bullet.

One other issue with bits > 63 is that the vhost-user protocol VHOST_USER_GET_FEATURES and VHOST_USER_SET_FEATURES messages use u64 to represent the features, so vhost-user-net devices can't query or enable these features. vhost-user is outside the scope of the virtio spec, though, and I think it's reasonable to extend the protocol to enable high feature bits rather than avoiding them forever.

Quoted 2 lines

Yes you would use VHOST_USER_SET_PROTOCOL_FEATURES to make VHOST_USER_GET_FEATURES return two u64s, or even a new message returning an array.

I think that additionally the VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET command will need some clarification, as in the current text looks a bit inconsistent:

"""
// in Offloads State Configuration / Setting Offloads State:

#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO 46

// ...

The class VIRTIO_NET_CTRL_GUEST_OFFLOADS has one command:
VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET applies the new offloads configuration.

le64 value passed as command data is a bitmask, bits set define offloads to be enabled, bits cleared - offloads to be disabled.

There is a corresponding device feature for each offload. Upon feature negotiation corresponding offload gets enabled to preserve backward compatibility
"""

Quoted 8 lines

The "corresponding device feature" has the same numerical value of the selected offloads, except for UDP tunnels related one (which are mapped to bits corresponding to reserved features).

It's unclear to me which should be the better way to address this inconsistency.

/P

Parav, what's your take here? Given your change broke VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET, do you want to revert it?

Signature · 1 line
MST
#17 lore raw
Quoted 51 lines

From: Michael S. Tsirkin <[email protected]>
Sent: Wednesday, April 30, 2025 2:14 AM

Quoted 3 lines · Paolo Abeni
Quoted 2 lines · Michael S. Tsirkin
Quoted 1 line · Daniel Verkamp

On Tue, Apr 22, 2025 at 10:46 PM Michael S. Tsirkin <[email protected]>

Quoted 4 lines · wrote:
Quoted 4 lines
Quoted 4 lines
Quoted 1 line

I'm afraid we'll have to bite the bullet.

One other issue with bits > 63 is that the vhost-user protocol VHOST_USER_GET_FEATURES and VHOST_USER_SET_FEATURES

messages use

Quoted 7 lines
Quoted 7 lines
Quoted 4 lines

u64 to represent the features, so vhost-user-net devices can't query or enable these features. vhost-user is outside the scope of the virtio spec, though, and I think it's reasonable to extend the protocol to enable high feature bits rather than avoiding them forever.

Yes you would use VHOST_USER_SET_PROTOCOL_FEATURES to make VHOST_USER_GET_FEATURES return two u64s, or even a new message

returning an array.

Quoted 9 lines

I think that additionally the VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET command will need some clarification, as in the current text looks a bit
inconsistent:

"""
// in Offloads State Configuration / Setting Offloads State:

Quoted 6 lines

#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO 46

// ...

The class VIRTIO_NET_CTRL_GUEST_OFFLOADS has one command:
VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET applies the new offloads

configuration.

Quoted 12 lines

le64 value passed as command data is a bitmask, bits set define offloads to be enabled, bits cleared - offloads to be disabled.

There is a corresponding device feature for each offload. Upon feature negotiation corresponding offload gets enabled to preserve backward compatibility """

The "corresponding device feature" has the same numerical value of the selected offloads, except for UDP tunnels related one (which are mapped to bits corresponding to reserved features).

Quoted 8 lines
Quoted 4 lines

It's unclear to me which should be the better way to address this inconsistency.

/P

Parav, what's your take here? Given your change broke VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET, do you want to revert it?

I see two options.
Opt_1:
Open source Linux kernel driver and DPDK PMD has not used RSS_CONTEXT yet.
If Heng from Alibaba acks that they do not have any internal implementation either, it may be safe to shift _all_ feature > 63 to lower position. We can get Yuri's feedback, if at all windows driver has used RSS context.

And once for all we mark it that feature bits are limited to 0-63.
There is enough infrastructure in place in virtio spec to not try to squeeze things in feature bits.
And these 4 bits are good example of it already, which could have been negotiated/communicated at later phase of driver at runtime. Only bit required was a bit to expand vnet header size at early stage.

Advantage: brings the good practice to adapt to the modern and efficient driver->device interface.
Risk: May break RSS_CONTEXT (risk looks low)

Opt_2. VIRTIO_NET_CTRL_GUEST_OFFLOADS command text to be updated to indicate that,

Below defines corresponds to respective feature bits 65 to 68. There is still one to one mapping, its just position is different inside the class. This is clarification text to be added and sw can adjust for it.

#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO 46
#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_CSUM 47
#define VIRTIO_NET_F_GUEST_USO4 54
#define VIRTIO_NET_F_GUEST_USO6 55

Advantages of 2nd option are:
a. featuring bits remain open upto 127.
b. Does not break RSS_CONTEXT.

Both options are practical to me.
I prefer #1, if Heng acks it, but also ok for #2.

#18 lore raw

Windows driver does not use RSS_CONTEXT yet. But this is not the last addition of an offload feature, so we'll face the same problem later. I also do not see any risks in option 2.

Quoted 100 lines · Parav Pandit
Quoted 8 lines

From: Michael S. Tsirkin <[email protected]>
Sent: Wednesday, April 30, 2025 2:14 AM

Quoted 2 lines · Paolo Abeni
Quoted 1 line · Michael S. Tsirkin

On Wed, Apr 23, 2025 at 09:29:11AM -0700, Daniel Verkamp wrote:

Quoted 12 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line

On Tue, Apr 22, 2025 at 10:46 PM Michael S. Tsirkin <[email protected]

Quoted 4 lines · wrote:
Quoted 4 lines
Quoted 4 lines
Quoted 1 line

I'm afraid we'll have to bite the bullet.

One other issue with bits > 63 is that the vhost-user protocol VHOST_USER_GET_FEATURES and VHOST_USER_SET_FEATURES

messages use

Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

u64 to represent the features, so vhost-user-net devices can't query or enable these features. vhost-user is outside the scope of the virtio spec, though, and I think it's reasonable to extend the protocol to enable high feature bits rather than avoiding them

forever.

Quoted 32 lines
Quoted 3 lines
Quoted 3 lines

Yes you would use VHOST_USER_SET_PROTOCOL_FEATURES to make VHOST_USER_GET_FEATURES return two u64s, or even a new message

returning an array.

Quoted 9 lines

I think that additionally the VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET command will need some clarification, as in the current text looks a bit
inconsistent:

"""
// in Offloads State Configuration / Setting Offloads State:

Quoted 6 lines

#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO 46

// ...

The class VIRTIO_NET_CTRL_GUEST_OFFLOADS has one command:
VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET applies the new offloads

configuration.

Quoted 12 lines

le64 value passed as command data is a bitmask, bits set define offloads to be enabled, bits cleared - offloads to be disabled.

There is a corresponding device feature for each offload. Upon feature negotiation corresponding offload gets enabled to preserve backward compatibility """

The "corresponding device feature" has the same numerical value of the selected offloads, except for UDP tunnels related one (which are mapped to bits corresponding to reserved features).

Quoted 8 lines
Quoted 4 lines

It's unclear to me which should be the better way to address this inconsistency.

/P

Parav, what's your take here? Given your change broke VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET, do you want to revert it?

I see two options.
Opt_1:
Open source Linux kernel driver and DPDK PMD has not used RSS_CONTEXT yet. If Heng from Alibaba acks that they do not have any internal implementation either, it may be safe to shift _all_ feature > 63 to lower position.
We can get Yuri's feedback, if at all windows driver has used RSS context.

And once for all we mark it that feature bits are limited to 0-63. There is enough infrastructure in place in virtio spec to not try to squeeze things in feature bits.
And these 4 bits are good example of it already, which could have been negotiated/communicated at later phase of driver at runtime. Only bit required was a bit to expand vnet header size at early stage.

Advantage: brings the good practice to adapt to the modern and efficient driver->device interface.
Risk: May break RSS_CONTEXT (risk looks low)

Opt_2. VIRTIO_NET_CTRL_GUEST_OFFLOADS command text to be updated to indicate that,

Below defines corresponds to respective feature bits 65 to 68. There is still one to one mapping, its just position is different inside the class. This is clarification text to be added and sw can adjust for it.

#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO 46
#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_CSUM 47
#define VIRTIO_NET_F_GUEST_USO4 54
#define VIRTIO_NET_F_GUEST_USO6 55

Advantages of 2nd option are:
a. featuring bits remain open upto 127.
b. Does not break RSS_CONTEXT.

Both options are practical to me.
I prefer #1, if Heng acks it, but also ok for #2.

#19 lore raw

This Is not last, but most offload features including this one does not need to tell the device in driver boot flow. User must be able to enable/disable them dynamically as/when needed.
So we will not face the same problem when they are done dynamically without feature bit.

From: Yuri Benditovich <[email protected]>
Sent: Wednesday, April 30, 2025 10:55 AM
To: Parav Pandit <[email protected]>
Cc: Michael S. Tsirkin <[email protected]>; Paolo Abeni <[email protected]>; [email protected]; [email protected]; [email protected]; [email protected]; Jason Wang <[email protected]>; Shahaf Shuler <[email protected]>; Willem de Bruijn <[email protected]>; Daniel Verkamp <[email protected]>
Subject: Re: [PATCH v1] virtio-net: Fix to avoid using reserved feature bits

Windows driver does not use RSS_CONTEXT yet. But this is not the last addition of an offload feature, so we'll face the same problem later. I also do not see any risks in option 2.

Quoted 49 lines · On Wed, Apr 30, 2025 at 7:44 AM Parav Pandit…

From: Michael S. Tsirkin <[email protected]<mailto:[email protected]>>
Sent: Wednesday, April 30, 2025 2:14 AM

Quoted 3 lines · Paolo Abeni
Quoted 2 lines · Michael S. Tsirkin
Quoted 1 line · Daniel Verkamp

On Tue, Apr 22, 2025 at 10:46 PM Michael S. Tsirkin <[email protected]<mailto:[email protected]>>

Quoted 4 lines · wrote:
Quoted 4 lines
Quoted 4 lines
Quoted 1 line

I'm afraid we'll have to bite the bullet.

One other issue with bits > 63 is that the vhost-user protocol VHOST_USER_GET_FEATURES and VHOST_USER_SET_FEATURES

messages use

Quoted 7 lines
Quoted 7 lines
Quoted 4 lines

u64 to represent the features, so vhost-user-net devices can't query or enable these features. vhost-user is outside the scope of the virtio spec, though, and I think it's reasonable to extend the protocol to enable high feature bits rather than avoiding them forever.

Yes you would use VHOST_USER_SET_PROTOCOL_FEATURES to make VHOST_USER_GET_FEATURES return two u64s, or even a new message

returning an array.

Quoted 9 lines

I think that additionally the VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET command will need some clarification, as in the current text looks a bit
inconsistent:

"""
// in Offloads State Configuration / Setting Offloads State:

Quoted 6 lines

#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO 46

// ...

The class VIRTIO_NET_CTRL_GUEST_OFFLOADS has one command:
VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET applies the new offloads

configuration.

Quoted 12 lines

le64 value passed as command data is a bitmask, bits set define offloads to be enabled, bits cleared - offloads to be disabled.

There is a corresponding device feature for each offload. Upon feature negotiation corresponding offload gets enabled to preserve backward compatibility """

The "corresponding device feature" has the same numerical value of the selected offloads, except for UDP tunnels related one (which are mapped to bits corresponding to reserved features).

Quoted 8 lines
Quoted 4 lines

It's unclear to me which should be the better way to address this inconsistency.

/P

Parav, what's your take here? Given your change broke VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET, do you want to revert it?

I see two options.
Opt_1:
Open source Linux kernel driver and DPDK PMD has not used RSS_CONTEXT yet.
If Heng from Alibaba acks that they do not have any internal implementation either, it may be safe to shift _all_ feature > 63 to lower position. We can get Yuri's feedback, if at all windows driver has used RSS context.

And once for all we mark it that feature bits are limited to 0-63.
There is enough infrastructure in place in virtio spec to not try to squeeze things in feature bits.
And these 4 bits are good example of it already, which could have been negotiated/communicated at later phase of driver at runtime. Only bit required was a bit to expand vnet header size at early stage.

Advantage: brings the good practice to adapt to the modern and efficient driver->device interface.
Risk: May break RSS_CONTEXT (risk looks low)

Opt_2. VIRTIO_NET_CTRL_GUEST_OFFLOADS command text to be updated to indicate that,

Below defines corresponds to respective feature bits 65 to 68. There is still one to one mapping, its just position is different inside the class. This is clarification text to be added and sw can adjust for it.

#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO 46
#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_CSUM 47
#define VIRTIO_NET_F_GUEST_USO4 54
#define VIRTIO_NET_F_GUEST_USO6 55

Advantages of 2nd option are:
a. featuring bits remain open upto 127.
b. Does not break RSS_CONTEXT.

Both options are practical to me.
I prefer #1, if Heng acks it, but also ok for #2.

#20 lore raw
In reply to Parav Pandit
Quoted 82 lines · Parav Pandit
Quoted 49 lines

From: Michael S. Tsirkin <[email protected]>
Sent: Wednesday, April 30, 2025 2:14 AM

Quoted 3 lines · Paolo Abeni
Quoted 2 lines · Michael S. Tsirkin
Quoted 1 line · Daniel Verkamp

On Tue, Apr 22, 2025 at 10:46 PM Michael S. Tsirkin <[email protected]>

Quoted 4 lines · wrote:
Quoted 4 lines
Quoted 4 lines
Quoted 1 line

I'm afraid we'll have to bite the bullet.

One other issue with bits > 63 is that the vhost-user protocol VHOST_USER_GET_FEATURES and VHOST_USER_SET_FEATURES

messages use

Quoted 7 lines
Quoted 7 lines
Quoted 4 lines

u64 to represent the features, so vhost-user-net devices can't query or enable these features. vhost-user is outside the scope of the virtio spec, though, and I think it's reasonable to extend the protocol to enable high feature bits rather than avoiding them forever.

Yes you would use VHOST_USER_SET_PROTOCOL_FEATURES to make VHOST_USER_GET_FEATURES return two u64s, or even a new message

returning an array.

Quoted 9 lines

I think that additionally the VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET command will need some clarification, as in the current text looks a bit
inconsistent:

"""
// in Offloads State Configuration / Setting Offloads State:

Quoted 6 lines

#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO 46

// ...

The class VIRTIO_NET_CTRL_GUEST_OFFLOADS has one command:
VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET applies the new offloads

configuration.

Quoted 12 lines

le64 value passed as command data is a bitmask, bits set define offloads to be enabled, bits cleared - offloads to be disabled.

There is a corresponding device feature for each offload. Upon feature negotiation corresponding offload gets enabled to preserve backward compatibility """

The "corresponding device feature" has the same numerical value of the selected offloads, except for UDP tunnels related one (which are mapped to bits corresponding to reserved features).

Quoted 8 lines
Quoted 4 lines

It's unclear to me which should be the better way to address this inconsistency.

/P

Parav, what's your take here? Given your change broke VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET, do you want to revert it?

I see two options.
Opt_1:
Open source Linux kernel driver and DPDK PMD has not used RSS_CONTEXT yet.
If Heng from Alibaba acks that they do not have any internal implementation either, it may be safe to shift _all_ feature > 63 to lower position. We can get Yuri's feedback, if at all windows driver has used RSS context.

And once for all we mark it that feature bits are limited to 0-63.
There is enough infrastructure in place in virtio spec to not try to squeeze things in feature bits.
And these 4 bits are good example of it already, which could have been negotiated/communicated at later phase of driver at runtime. Only bit required was a bit to expand vnet header size at early stage.

Advantage: brings the good practice to adapt to the modern and efficient driver->device interface.
Risk: May break RSS_CONTEXT (risk looks low)

Opt_2. VIRTIO_NET_CTRL_GUEST_OFFLOADS command text to be updated to indicate that,

Below defines corresponds to respective feature bits 65 to 68. There is still one to one mapping, its just position is different inside the class. This is clarification text to be added and sw can adjust for it.

#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO 46
#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_CSUM 47
#define VIRTIO_NET_F_GUEST_USO4 54
#define VIRTIO_NET_F_GUEST_USO6 55

Quoted 6 lines

Advantages of 2nd option are:
a. featuring bits remain open upto 127.
b. Does not break RSS_CONTEXT.

Both options are practical to me.
I prefer #1, if Heng acks it, but also ok for #2.

I spent quite of bit of time trying to evaluate the scope of features bit expansion (implied by the option 2 above).

While strictly speaking I haven't hit yet a complete blocker, implementation-wise it's going to be huge and error prone, as great deal of both the kernel and the user-space/qemu infrastructure hard-codes the 64 bit limit.

Even exposing the feature extensions only the the virtio-net device (AFAICS it will "minimize" the code churn) a lot of code and devices implementations are going to be impacted.

I expect a far away in time timeline for implementations based on option 2.

/P

#21 lore raw
Quoted 112 lines

From: Paolo Abeni <[email protected]>
Sent: Wednesday, April 30, 2025 3:42 PM

Quoted 56 lines · Parav Pandit
Quoted 56 lines

From: Michael S. Tsirkin <[email protected]>
Sent: Wednesday, April 30, 2025 2:14 AM

Quoted 4 lines · Paolo Abeni
Quoted 3 lines · Michael S. Tsirkin
Quoted 2 lines · Daniel Verkamp

On Tue, Apr 22, 2025 at 10:46 PM Michael S. Tsirkin <[email protected]>

Quoted 4 lines · wrote:
Quoted 4 lines
Quoted 4 lines
Quoted 1 line

I'm afraid we'll have to bite the bullet.

One other issue with bits > 63 is that the vhost-user protocol VHOST_USER_GET_FEATURES and VHOST_USER_SET_FEATURES

messages use

Quoted 7 lines
Quoted 7 lines
Quoted 4 lines

u64 to represent the features, so vhost-user-net devices can't query or enable these features. vhost-user is outside the scope of the virtio spec, though, and I think it's reasonable to extend the protocol to enable high feature bits rather than avoiding them forever.

Yes you would use VHOST_USER_SET_PROTOCOL_FEATURES to make VHOST_USER_GET_FEATURES return two u64s, or even a new message

returning an array.

Quoted 9 lines

I think that additionally the VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET command will need some clarification, as in the current text looks a bit
inconsistent:

"""
// in Offloads State Configuration / Setting Offloads State:

Quoted 6 lines

#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO 46

// ...

The class VIRTIO_NET_CTRL_GUEST_OFFLOADS has one command:
VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET applies the new offloads

configuration.

Quoted 16 lines

le64 value passed as command data is a bitmask, bits set define offloads to be enabled, bits cleared - offloads to be disabled.

There is a corresponding device feature for each offload. Upon feature negotiation corresponding offload gets enabled to preserve backward compatibility """

The "corresponding device feature" has the same numerical value of the selected offloads, except for UDP tunnels related one (which are mapped to bits corresponding to reserved features).

It's unclear to me which should be the better way to address this inconsistency.

/P

Quoted 6 lines
Quoted 2 lines

Parav, what's your take here? Given your change broke VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET, do you want to revert it?

I see two options.
Opt_1:
Open source Linux kernel driver and DPDK PMD has not used RSS_CONTEXT

yet.

Quoted 1 line

If Heng from Alibaba acks that they do not have any internal

implementation either, it may be safe to shift _all_ feature > 63 to lower position.

Quoted 4 lines

We can get Yuri's feedback, if at all windows driver has used RSS context.

And once for all we mark it that feature bits are limited to 0-63. There is enough infrastructure in place in virtio spec to not try to squeeze

things in feature bits.

Quoted 1 line

And these 4 bits are good example of it already, which could have been

negotiated/communicated at later phase of driver at runtime.

Quoted 3 lines

Only bit required was a bit to expand vnet header size at early stage.

Advantage: brings the good practice to adapt to the modern and efficient

driver->device interface.

Quoted 3 lines

Risk: May break RSS_CONTEXT (risk looks low)

Opt_2. VIRTIO_NET_CTRL_GUEST_OFFLOADS command text to be updated

to

Quoted 3 lines

indicate that,

Below defines corresponds to respective feature bits 65 to 68. There is still

one to one mapping, its just position is different inside the class.

Quoted 6 lines

This is clarification text to be added and sw can adjust for it.

#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO 46 #define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_CSUM 47 #define VIRTIO_NET_F_GUEST_USO4 54 #define VIRTIO_NET_F_GUEST_USO6 55

Quoted 6 lines

Advantages of 2nd option are:
a. featuring bits remain open upto 127.
b. Does not break RSS_CONTEXT.

Both options are practical to me.
I prefer #1, if Heng acks it, but also ok for #2.

I spent quite of bit of time trying to evaluate the scope of features bit expansion (implied by the option 2 above).

While strictly speaking I haven't hit yet a complete blocker, implementation- wise it's going to be huge and error prone, as great deal of both the kernel and the user-space/qemu infrastructure hard-codes the
64 bit limit.

Quoted 7 lines

Even exposing the feature extensions only the the virtio-net device (AFAICS it will "minimize" the code churn) a lot of code and devices implementations are going to be impacted.

I expect a far away in time timeline for implementations based on option 2.

/P

Given that its painful enough, we can update the virtio spec to limit to 64-bit features as option_1.

A virtio spec diff like below,
- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 50 to 63 Feature bits for the specific device type

Otherwise, we are delaying the problem to next feature, which is not good.

Heng,
Yuri already acked, so we are good from Windows side. With that Windows, Linux kernel, DPDK PMD are safe.

Can you also please confirm that there are no users within Alibaba relying on RSS_CONTEXT, so that shifting the feature bit will not break any existing functionality? I think it will not break, because we can update the non-open-source software to detect RSS_CONTEXT on a new bit.

For example, driver can do,
Enable_rss_context if bit X OR Bit 64 is set.

#22 lore raw
Quoted 131 lines · Parav Pandit
Quoted 112 lines

From: Paolo Abeni <[email protected]>
Sent: Wednesday, April 30, 2025 3:42 PM

Quoted 56 lines · Parav Pandit
Quoted 56 lines

From: Michael S. Tsirkin <[email protected]>
Sent: Wednesday, April 30, 2025 2:14 AM

Quoted 4 lines · Paolo Abeni
Quoted 3 lines · Michael S. Tsirkin
Quoted 2 lines · Daniel Verkamp

On Tue, Apr 22, 2025 at 10:46 PM Michael S. Tsirkin <[email protected]>

Quoted 4 lines · wrote:
Quoted 4 lines
Quoted 4 lines
Quoted 1 line

I'm afraid we'll have to bite the bullet.

One other issue with bits > 63 is that the vhost-user protocol VHOST_USER_GET_FEATURES and VHOST_USER_SET_FEATURES

messages use

Quoted 7 lines
Quoted 7 lines
Quoted 4 lines

u64 to represent the features, so vhost-user-net devices can't query or enable these features. vhost-user is outside the scope of the virtio spec, though, and I think it's reasonable to extend the protocol to enable high feature bits rather than avoiding them forever.

Yes you would use VHOST_USER_SET_PROTOCOL_FEATURES to make VHOST_USER_GET_FEATURES return two u64s, or even a new message

returning an array.

Quoted 9 lines

I think that additionally the VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET command will need some clarification, as in the current text looks a bit
inconsistent:

"""
// in Offloads State Configuration / Setting Offloads State:

Quoted 6 lines

#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO 46

// ...

The class VIRTIO_NET_CTRL_GUEST_OFFLOADS has one command:
VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET applies the new offloads

configuration.

Quoted 16 lines

le64 value passed as command data is a bitmask, bits set define offloads to be enabled, bits cleared - offloads to be disabled.

There is a corresponding device feature for each offload. Upon feature negotiation corresponding offload gets enabled to preserve backward compatibility """

The "corresponding device feature" has the same numerical value of the selected offloads, except for UDP tunnels related one (which are mapped to bits corresponding to reserved features).

It's unclear to me which should be the better way to address this inconsistency.

/P

Quoted 6 lines
Quoted 2 lines

Parav, what's your take here? Given your change broke VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET, do you want to revert it?

I see two options.
Opt_1:
Open source Linux kernel driver and DPDK PMD has not used RSS_CONTEXT

yet.

Quoted 1 line

If Heng from Alibaba acks that they do not have any internal

implementation either, it may be safe to shift _all_ feature > 63 to lower position.

Quoted 4 lines

We can get Yuri's feedback, if at all windows driver has used RSS context.

And once for all we mark it that feature bits are limited to 0-63. There is enough infrastructure in place in virtio spec to not try to squeeze

things in feature bits.

Quoted 1 line

And these 4 bits are good example of it already, which could have been

negotiated/communicated at later phase of driver at runtime.

Quoted 3 lines

Only bit required was a bit to expand vnet header size at early stage.

Advantage: brings the good practice to adapt to the modern and efficient

driver->device interface.

Quoted 3 lines

Risk: May break RSS_CONTEXT (risk looks low)

Opt_2. VIRTIO_NET_CTRL_GUEST_OFFLOADS command text to be updated

to

Quoted 3 lines

indicate that,

Below defines corresponds to respective feature bits 65 to 68. There is still

one to one mapping, its just position is different inside the class.

Quoted 6 lines

This is clarification text to be added and sw can adjust for it.

#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO 46 #define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_CSUM 47 #define VIRTIO_NET_F_GUEST_USO4 54 #define VIRTIO_NET_F_GUEST_USO6 55

Quoted 6 lines

Advantages of 2nd option are:
a. featuring bits remain open upto 127.
b. Does not break RSS_CONTEXT.

Both options are practical to me.
I prefer #1, if Heng acks it, but also ok for #2.

I spent quite of bit of time trying to evaluate the scope of features bit expansion (implied by the option 2 above).

While strictly speaking I haven't hit yet a complete blocker, implementation- wise it's going to be huge and error prone, as great deal of both the kernel and the user-space/qemu infrastructure hard-codes the
64 bit limit.

Quoted 7 lines

Even exposing the feature extensions only the the virtio-net device (AFAICS it will "minimize" the code churn) a lot of code and devices implementations are going to be impacted.

I expect a far away in time timeline for implementations based on option 2.

/P

Given that its painful enough, we can update the virtio spec to limit to 64-bit features as option_1.

A virtio spec diff like below,
- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 50 to 63 Feature bits for the specific device type

Otherwise, we are delaying the problem to next feature, which is not good.

Heng,
Yuri already acked, so we are good from Windows side. With that Windows, Linux kernel, DPDK PMD are safe.

Quoted 5 lines

Can you also please confirm that there are no users within Alibaba relying on RSS_CONTEXT, so that shifting the feature bit will not break any existing functionality? I think it will not break, because we can update the non-open-source software to detect RSS_CONTEXT on a new bit.

For example, driver can do,
Enable_rss_context if bit X OR Bit 64 is set.

sorry if I am unclear, I think it's too late to move RSS really. It has been out there for two years. The reason we can be reasonably sure no one implemented the offloads so far is because the set offloads command is broken and no one complained. We need to fix them in the spec, current one is broken. So the tunnel offloads can move if it's convenient.

Signature · 1 line
MST
#23 lore raw
Quoted 6 lines · Michael S. Tsirkin

sorry if I am unclear, I think it's too late to move RSS really. It has been out there for two years. The reason we can be reasonably sure no one implemented the offloads so far is because the set offloads command is broken and no one complained. We need to fix them in the spec, current one is broken. So the tunnel offloads can move if it's convenient.

I understand the argument WRT RSS. I think and hope RSS and GSO over tunnel offload can follow different paths: I would opt to move around the tunnel offload bit in the reserved range (and possibly make a little more room there for other future features).

Cheers,

Paolo

#24 lore raw
In reply to Michael S. Tsirkin
Quoted 164 lines

From: Michael S. Tsirkin <[email protected]>
Sent: Thursday, May 1, 2025 7:13 PM
To: Parav Pandit <[email protected]>
Cc: Paolo Abeni <[email protected]>; [email protected]; virtio- [email protected]; [email protected]; [email protected]; Jason Wang <[email protected]>; Shahaf Shuler <[email protected]>; Willem de Bruijn <[email protected]>; Daniel Verkamp <[email protected]>
Subject: Re: [PATCH v1] virtio-net: Fix to avoid using reserved feature bits

Quoted 17 lines · Parav Pandit
Quoted 17 lines

From: Paolo Abeni <[email protected]>
Sent: Wednesday, April 30, 2025 3:42 PM

Quoted 11 lines · Parav Pandit
Quoted 11 lines

From: Michael S. Tsirkin <[email protected]>
Sent: Wednesday, April 30, 2025 2:14 AM

Quoted 4 lines · Paolo Abeni
Quoted 3 lines · Michael S. Tsirkin
Quoted 2 lines · Daniel Verkamp

On Tue, Apr 22, 2025 at 10:46 PM Michael S. Tsirkin <[email protected]>

Quoted 2 lines · wrote:
Quoted 2 lines
Quoted 2 lines
Quoted 1 line

I'm afraid we'll have to bite the bullet.

Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

One other issue with bits > 63 is that the vhost-user protocol VHOST_USER_GET_FEATURES and VHOST_USER_SET_FEATURES

messages use

Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

u64 to represent the features, so vhost-user-net devices can't query or enable these features. vhost-user is outside the scope of the virtio spec, though, and I think it's reasonable to extend the protocol to enable high feature bits rather than

avoiding them forever.

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

Yes you would use VHOST_USER_SET_PROTOCOL_FEATURES to make VHOST_USER_GET_FEATURES return two u64s, or even a new

message

Quoted 42 lines
Quoted 42 lines
Quoted 42 lines
Quoted 29 lines

returning an array.

Quoted 9 lines

I think that additionally the VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET command will need some clarification, as in the current text looks a bit
inconsistent:

"""
// in Offloads State Configuration / Setting Offloads State:

Quoted 6 lines

#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO 46

// ...

The class VIRTIO_NET_CTRL_GUEST_OFFLOADS has one command:
VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET applies the new offloads

configuration.

Quoted 12 lines

le64 value passed as command data is a bitmask, bits set define offloads to be enabled, bits cleared - offloads to be disabled.

There is a corresponding device feature for each offload. Upon feature negotiation corresponding offload gets enabled to preserve backward compatibility """

The "corresponding device feature" has the same numerical value of the selected offloads, except for UDP tunnels related one (which are mapped to bits corresponding to reserved features).

Quoted 8 lines
Quoted 4 lines

It's unclear to me which should be the better way to address this inconsistency.

/P

Parav, what's your take here? Given your change broke VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET, do you want to revert it?

I see two options.
Opt_1:
Open source Linux kernel driver and DPDK PMD has not used RSS_CONTEXT

Quoted 5 lines
Quoted 5 lines

yet.

Quoted 1 line

If Heng from Alibaba acks that they do not have any internal

implementation either, it may be safe to shift _all_ feature > 63 to lower position.

Quoted 1 line

We can get Yuri's feedback, if at all windows driver has used RSS

context.

Quoted 8 lines
Quoted 8 lines
Quoted 4 lines

And once for all we mark it that feature bits are limited to 0-63. There is enough infrastructure in place in virtio spec to not try to squeeze

things in feature bits.

Quoted 2 lines

And these 4 bits are good example of it already, which could have been

negotiated/communicated at later phase of driver at runtime.

Quoted 8 lines
Quoted 8 lines
Quoted 4 lines

Only bit required was a bit to expand vnet header size at early stage.

Advantage: brings the good practice to adapt to the modern and efficient

driver->device interface.

Quoted 3 lines

Risk: May break RSS_CONTEXT (risk looks low)

Opt_2. VIRTIO_NET_CTRL_GUEST_OFFLOADS command text to be

updated

Quoted 28 lines
Quoted 28 lines

to

Quoted 4 lines

indicate that,

Below defines corresponds to respective feature bits 65 to 68. There is still

one to one mapping, its just position is different inside the class.

Quoted 6 lines

This is clarification text to be added and sw can adjust for it.

#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO 46 #define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_CSUM 47 #define VIRTIO_NET_F_GUEST_USO4 54 #define VIRTIO_NET_F_GUEST_USO6 55

Quoted 6 lines

Advantages of 2nd option are:
a. featuring bits remain open upto 127.
b. Does not break RSS_CONTEXT.

Both options are practical to me.
I prefer #1, if Heng acks it, but also ok for #2.

I spent quite of bit of time trying to evaluate the scope of features bit expansion (implied by the option 2 above).

While strictly speaking I haven't hit yet a complete blocker, implementation- wise it's going to be huge and error prone, as great deal of both the kernel and the user-space/qemu infrastructure hard-codes the
64 bit limit.

Quoted 5 lines
Quoted 5 lines

Even exposing the feature extensions only the the virtio-net device (AFAICS it will "minimize" the code churn) a lot of code and devices implementations are going to be impacted.

I expect a far away in time timeline for implementations based on option

2.

Quoted 3 lines
Quoted 2 lines

/P

Given that its painful enough, we can update the virtio spec to limit to 64-bit

features as option_1.

Quoted 5 lines

A virtio spec diff like below,
- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 50 to 63 Feature bits for the specific device type

Quoted 7 lines

Otherwise, we are delaying the problem to next feature, which is not good.

Heng,
Yuri already acked, so we are good from Windows side. With that Windows, Linux kernel, DPDK PMD are safe.

Can you also please confirm that there are no users within Alibaba relying

on RSS_CONTEXT, so that shifting the feature bit will not break any existing functionality?

Quoted 1 line

I think it will not break, because we can update the non-open-source

software to detect RSS_CONTEXT on a new bit.

Quoted 3 lines

For example, driver can do,
Enable_rss_context if bit X OR Bit 64 is set.

Quoted 5 lines

sorry if I am unclear, I think it's too late to move RSS really. It has been out there for two years. The reason we can be reasonably sure no one implemented the offloads so far is because the set offloads command is broken and no one complained. We need to fix them in the spec, current one is broken. So the tunnel offloads can move if it's convenient.

There are few proposals on table.

1. From Paolo,
- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 46 to 127 Feature bits for the specific device type

This does not have good reason of why it should still be 127.

2. From me:
- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 45 to 64 Feature bits for the specific device type

This is an extension of Paolo, to justify that implementing feature bits is extremely hard even for experts as pointed by Paolo. It is worth to not extend it further.
RSS can be negotiated via new bit 44 in future bit as OR of 44 and 64 so that more wider users (Linux, freebsd, qnx, Windows, dpdk pmd) can pick 44.

3. From me:
Keep the feature bits encoding as is up to 127 bits, because may be there is (unknown and weird) value in having 127 feature bits. (unknown because the reasoning of #1 and #3 mismatch).

In that case,
VIRTIO_NET_CTRL_GUEST_OFFLOADS command text to be updated to indicate feature fits A to D map to VIRTIO_NET_F_CTRL_GUEST_OFFLOADS.offloads bits A' to D'.

I am fine with option #2 and #3.
Doing #1 for sure is wrong.
Wrong because it delays the problem of #1 from this to another feature [A] who's voting already completed.

[A] https://lore.kernel.org/virtio-comment/DM4PR18MB4269F73B786E83EF68A70F37DFB82@DM4PR18MB4269.namprd18.prod.outlook.com/T/#t

#25 lore raw
Quoted 193 lines · Parav Pandit
Quoted 164 lines

From: Michael S. Tsirkin <[email protected]>
Sent: Thursday, May 1, 2025 7:13 PM
To: Parav Pandit <[email protected]>
Cc: Paolo Abeni <[email protected]>; [email protected]; virtio- [email protected]; [email protected]; [email protected]; Jason Wang <[email protected]>; Shahaf Shuler <[email protected]>; Willem de Bruijn <[email protected]>; Daniel Verkamp <[email protected]>
Subject: Re: [PATCH v1] virtio-net: Fix to avoid using reserved feature bits

Quoted 17 lines · Parav Pandit
Quoted 17 lines

From: Paolo Abeni <[email protected]>
Sent: Wednesday, April 30, 2025 3:42 PM

Quoted 11 lines · Parav Pandit
Quoted 11 lines

From: Michael S. Tsirkin <[email protected]>
Sent: Wednesday, April 30, 2025 2:14 AM

Quoted 4 lines · Paolo Abeni
Quoted 3 lines · Michael S. Tsirkin
Quoted 2 lines · Daniel Verkamp

On Tue, Apr 22, 2025 at 10:46 PM Michael S. Tsirkin <[email protected]>

Quoted 2 lines · wrote:
Quoted 2 lines
Quoted 2 lines
Quoted 1 line

I'm afraid we'll have to bite the bullet.

Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

One other issue with bits > 63 is that the vhost-user protocol VHOST_USER_GET_FEATURES and VHOST_USER_SET_FEATURES

messages use

Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

u64 to represent the features, so vhost-user-net devices can't query or enable these features. vhost-user is outside the scope of the virtio spec, though, and I think it's reasonable to extend the protocol to enable high feature bits rather than

avoiding them forever.

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

Yes you would use VHOST_USER_SET_PROTOCOL_FEATURES to make VHOST_USER_GET_FEATURES return two u64s, or even a new

message

Quoted 42 lines
Quoted 42 lines
Quoted 42 lines
Quoted 29 lines

returning an array.

Quoted 9 lines

I think that additionally the VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET command will need some clarification, as in the current text looks a bit
inconsistent:

"""
// in Offloads State Configuration / Setting Offloads State:

Quoted 6 lines

#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO 46

// ...

The class VIRTIO_NET_CTRL_GUEST_OFFLOADS has one command:
VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET applies the new offloads

configuration.

Quoted 12 lines

le64 value passed as command data is a bitmask, bits set define offloads to be enabled, bits cleared - offloads to be disabled.

There is a corresponding device feature for each offload. Upon feature negotiation corresponding offload gets enabled to preserve backward compatibility """

The "corresponding device feature" has the same numerical value of the selected offloads, except for UDP tunnels related one (which are mapped to bits corresponding to reserved features).

Quoted 8 lines
Quoted 4 lines

It's unclear to me which should be the better way to address this inconsistency.

/P

Parav, what's your take here? Given your change broke VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET, do you want to revert it?

I see two options.
Opt_1:
Open source Linux kernel driver and DPDK PMD has not used RSS_CONTEXT

Quoted 5 lines
Quoted 5 lines

yet.

Quoted 1 line

If Heng from Alibaba acks that they do not have any internal

implementation either, it may be safe to shift _all_ feature > 63 to lower position.

Quoted 1 line

We can get Yuri's feedback, if at all windows driver has used RSS

context.

Quoted 8 lines
Quoted 8 lines
Quoted 4 lines

And once for all we mark it that feature bits are limited to 0-63. There is enough infrastructure in place in virtio spec to not try to squeeze

things in feature bits.

Quoted 2 lines

And these 4 bits are good example of it already, which could have been

negotiated/communicated at later phase of driver at runtime.

Quoted 8 lines
Quoted 8 lines
Quoted 4 lines

Only bit required was a bit to expand vnet header size at early stage.

Advantage: brings the good practice to adapt to the modern and efficient

driver->device interface.

Quoted 3 lines

Risk: May break RSS_CONTEXT (risk looks low)

Opt_2. VIRTIO_NET_CTRL_GUEST_OFFLOADS command text to be

updated

Quoted 28 lines
Quoted 28 lines

to

Quoted 4 lines

indicate that,

Below defines corresponds to respective feature bits 65 to 68. There is still

one to one mapping, its just position is different inside the class.

Quoted 6 lines

This is clarification text to be added and sw can adjust for it.

#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO 46 #define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_CSUM 47 #define VIRTIO_NET_F_GUEST_USO4 54 #define VIRTIO_NET_F_GUEST_USO6 55

Quoted 6 lines

Advantages of 2nd option are:
a. featuring bits remain open upto 127.
b. Does not break RSS_CONTEXT.

Both options are practical to me.
I prefer #1, if Heng acks it, but also ok for #2.

I spent quite of bit of time trying to evaluate the scope of features bit expansion (implied by the option 2 above).

While strictly speaking I haven't hit yet a complete blocker, implementation- wise it's going to be huge and error prone, as great deal of both the kernel and the user-space/qemu infrastructure hard-codes the
64 bit limit.

Quoted 5 lines
Quoted 5 lines

Even exposing the feature extensions only the the virtio-net device (AFAICS it will "minimize" the code churn) a lot of code and devices implementations are going to be impacted.

I expect a far away in time timeline for implementations based on option

2.

Quoted 3 lines
Quoted 2 lines

/P

Given that its painful enough, we can update the virtio spec to limit to 64-bit

features as option_1.

Quoted 5 lines

A virtio spec diff like below,
- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 50 to 63 Feature bits for the specific device type

Quoted 7 lines

Otherwise, we are delaying the problem to next feature, which is not good.

Heng,
Yuri already acked, so we are good from Windows side. With that Windows, Linux kernel, DPDK PMD are safe.

Can you also please confirm that there are no users within Alibaba relying

on RSS_CONTEXT, so that shifting the feature bit will not break any existing functionality?

Quoted 1 line

I think it will not break, because we can update the non-open-source

software to detect RSS_CONTEXT on a new bit.

Quoted 3 lines

For example, driver can do,
Enable_rss_context if bit X OR Bit 64 is set.

Quoted 5 lines

sorry if I am unclear, I think it's too late to move RSS really. It has been out there for two years. The reason we can be reasonably sure no one implemented the offloads so far is because the set offloads command is broken and no one complained. We need to fix them in the spec, current one is broken. So the tunnel offloads can move if it's convenient.

There are few proposals on table.

1. From Paolo,
- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 46 to 127 Feature bits for the specific device type

This does not have good reason of why it should still be 127.

2. From me:
- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 45 to 64 Feature bits for the specific device type

This is an extension of Paolo, to justify that implementing feature bits is extremely hard even for experts as pointed by Paolo. It is worth to not extend it further.
RSS can be negotiated via new bit 44 in future bit as OR of 44 and 64 so that more wider users (Linux, freebsd, qnx, Windows, dpdk pmd) can pick 44.

3. From me:
Keep the feature bits encoding as is up to 127 bits, because may be there is (unknown and weird) value in having 127 feature bits. (unknown because the reasoning of #1 and #3 mismatch).
In that case,
VIRTIO_NET_CTRL_GUEST_OFFLOADS command text to be updated to indicate feature fits A to D map to VIRTIO_NET_F_CTRL_GUEST_OFFLOADS.offloads bits A' to D'.

Quoted 5 lines

I am fine with option #2 and #3.
Doing #1 for sure is wrong.
Wrong because it delays the problem of #1 from this to another feature [A] who's voting already completed.

[A] https://lore.kernel.org/virtio-comment/DM4PR18MB4269F73B786E83EF68A70F37DFB82@DM4PR18MB4269.namprd18.prod.outlook.com/T/#t

#3 seems more conservate.

Signature · 1 line
MST
#26 lore raw

Hi Paolo,

Quoted 226 lines

From: Michael S. Tsirkin <[email protected]>
Sent: Tuesday, May 6, 2025 1:26 PM

Quoted 6 lines · Parav Pandit
Quoted 6 lines

From: Michael S. Tsirkin <[email protected]>
Sent: Thursday, May 1, 2025 7:13 PM
To: Parav Pandit <[email protected]>
Cc: Paolo Abeni <[email protected]>; [email protected]; virtio- [email protected]; [email protected]; [email protected]; Jason Wang <[email protected]>; Shahaf

Shuler

Quoted 18 lines
Quoted 18 lines

<[email protected]>; Willem de Bruijn <[email protected]>; Daniel Verkamp <[email protected]>
Subject: Re: [PATCH v1] virtio-net: Fix to avoid using reserved feature bits

Quoted 12 lines · Parav Pandit
Quoted 12 lines

From: Paolo Abeni <[email protected]>
Sent: Wednesday, April 30, 2025 3:42 PM

Quoted 6 lines · Parav Pandit
Quoted 6 lines

From: Michael S. Tsirkin <[email protected]>
Sent: Wednesday, April 30, 2025 2:14 AM

Quoted 2 lines · Paolo Abeni
Quoted 1 line · Michael S. Tsirkin

On Wed, Apr 23, 2025 at 09:29:11AM -0700, Daniel Verkamp

Quoted 9 lines · wrote:
Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

On Tue, Apr 22, 2025 at 10:46 PM Michael S. Tsirkin <[email protected]>

Quoted 5 lines · wrote:
Quoted 5 lines
Quoted 5 lines
Quoted 1 line

I'm afraid we'll have to bite the bullet.

One other issue with bits > 63 is that the vhost-user protocol VHOST_USER_GET_FEATURES and VHOST_USER_SET_FEATURES

messages use

Quoted 8 lines
Quoted 8 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines

u64 to represent the features, so vhost-user-net devices can't query or enable these features. vhost-user is outside the scope of the virtio spec, though, and I think it's reasonable to extend the protocol to enable high feature bits rather than

avoiding them forever.

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

Yes you would use VHOST_USER_SET_PROTOCOL_FEATURES to

make

Quoted 10 lines
Quoted 10 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

VHOST_USER_GET_FEATURES return two u64s, or even a new

message

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines

returning an array.

Quoted 7 lines

I think that additionally the VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET
command will need some clarification, as in the current text looks a bit
inconsistent:

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines

"""
// in Offloads State Configuration / Setting Offloads State:

#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO 46

// ...

The class VIRTIO_NET_CTRL_GUEST_OFFLOADS has one

command:

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET applies the new

offloads

Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

configuration.

Quoted 3 lines

le64 value passed as command data is a bitmask, bits set define offloads to be enabled, bits cleared - offloads to be

disabled.

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines

There is a corresponding device feature for each offload. Upon feature negotiation corresponding offload gets enabled to preserve backward compatibility """

The "corresponding device feature" has the same numerical value of the selected offloads, except for UDP tunnels related one (which are mapped to bits corresponding to reserved

features).

Quoted 41 lines
Quoted 41 lines
Quoted 14 lines
Quoted 14 lines
Quoted 14 lines
Quoted 9 lines
Quoted 5 lines

It's unclear to me which should be the better way to address this inconsistency.

/P

Parav, what's your take here? Given your change broke VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET, do you want to revert it?

I see two options.
Opt_1:
Open source Linux kernel driver and DPDK PMD has not used RSS_CONTEXT

Quoted 6 lines
Quoted 6 lines

yet.

Quoted 1 line

If Heng from Alibaba acks that they do not have any internal

implementation either, it may be safe to shift _all_ feature > 63 to lower position.

Quoted 2 lines

We can get Yuri's feedback, if at all windows driver has used RSS

context.

Quoted 8 lines
Quoted 8 lines
Quoted 4 lines

And once for all we mark it that feature bits are limited to 0-63. There is enough infrastructure in place in virtio spec to not try to squeeze

things in feature bits.

Quoted 2 lines

And these 4 bits are good example of it already, which could have been

negotiated/communicated at later phase of driver at runtime.

Quoted 8 lines
Quoted 8 lines
Quoted 4 lines

Only bit required was a bit to expand vnet header size at early stage.

Advantage: brings the good practice to adapt to the modern and efficient

driver->device interface.

Quoted 3 lines

Risk: May break RSS_CONTEXT (risk looks low)

Opt_2. VIRTIO_NET_CTRL_GUEST_OFFLOADS command text to be

updated

Quoted 3 lines
Quoted 3 lines

to

Quoted 2 lines

indicate that,

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 2 lines

Below defines corresponds to respective feature bits 65 to 68. There is still

one to one mapping, its just position is different inside the class.

Quoted 5 lines

This is clarification text to be added and sw can adjust for it.

#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO 46 #define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_CSUM 47 #define VIRTIO_NET_F_GUEST_USO4 54 #define

VIRTIO_NET_F_GUEST_USO6 55

Quoted 28 lines
Quoted 28 lines
Quoted 17 lines
Quoted 17 lines
Quoted 7 lines

Advantages of 2nd option are:
a. featuring bits remain open upto 127.
b. Does not break RSS_CONTEXT.

Both options are practical to me.
I prefer #1, if Heng acks it, but also ok for #2.

I spent quite of bit of time trying to evaluate the scope of features bit expansion (implied by the option 2 above).

While strictly speaking I haven't hit yet a complete blocker, implementation- wise it's going to be huge and error prone, as great deal of both the kernel and the user-space/qemu infrastructure hard-codes the
64 bit limit.

Quoted 6 lines
Quoted 6 lines

Even exposing the feature extensions only the the virtio-net device (AFAICS it will "minimize" the code churn) a lot of code and devices implementations are going to be impacted.

I expect a far away in time timeline for implementations based on option

2.

Quoted 4 lines
Quoted 2 lines

/P

Given that its painful enough, we can update the virtio spec to limit to 64-bit

Quoted 7 lines
Quoted 7 lines

features as option_1.

Quoted 6 lines

A virtio spec diff like below,
- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 50 to 63 Feature bits for the specific device type

Otherwise, we are delaying the problem to next feature, which is not

good.

Quoted 17 lines
Quoted 17 lines
Quoted 7 lines

Heng,
Yuri already acked, so we are good from Windows side. With that Windows, Linux kernel, DPDK PMD are safe.

Can you also please confirm that there are no users within Alibaba relying

on RSS_CONTEXT, so that shifting the feature bit will not break any existing functionality?

Quoted 2 lines

I think it will not break, because we can update the non-open-source

software to detect RSS_CONTEXT on a new bit.

Quoted 3 lines

For example, driver can do,
Enable_rss_context if bit X OR Bit 64 is set.

Quoted 5 lines
Quoted 5 lines

sorry if I am unclear, I think it's too late to move RSS really. It has been out there for two years. The reason we can be reasonably sure no one implemented the offloads so far is because the set offloads command is broken and no one complained. We need to fix them in the spec, current one is broken. So the tunnel offloads can move

if it's convenient.

Quoted 7 lines

There are few proposals on table.

1. From Paolo,
- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 46 to 127 Feature bits for the specific device type

Quoted 7 lines

This does not have good reason of why it should still be 127.

2. From me:
- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 45 to 64 Feature bits for the specific device type

This is an extension of Paolo, to justify that implementing feature bits is

extremely hard even for experts as pointed by Paolo.

Quoted 2 lines

It is worth to not extend it further.
RSS can be negotiated via new bit 44 in future bit as OR of 44 and 64 so that

more wider users (Linux, freebsd, qnx, Windows, dpdk pmd) can pick 44.

Quoted 3 lines

3. From me:
Keep the feature bits encoding as is up to 127 bits, because may be there is

(unknown and weird) value in having 127 feature bits.

Quoted 3 lines

(unknown because the reasoning of #1 and #3 mismatch). In that case,
VIRTIO_NET_CTRL_GUEST_OFFLOADS command text to be updated to

indicate

Quoted 1 line

feature fits A to D map to VIRTIO_NET_F_CTRL_GUEST_OFFLOADS.offloads

bits A' to D'.

Quoted 4 lines

I am fine with option #2 and #3.
Doing #1 for sure is wrong.
Wrong because it delays the problem of #1 from this to another feature [A]

who's voting already completed.

Quoted 1 line
Quoted 7 lines
Quoted 2 lines

[A]
https://lore.kernel.org/virtio-

comment/DM4PR18MB4269F73B786E83EF68A70F

Quoted 1 line

[email protected]/T/#t

#3 seems more conservate.

Are you good with #3?

#27 lore raw

Hi all,

Quoted 37 lines · Parav Pandit
Quoted 37 lines

From: Michael S. Tsirkin <[email protected]>
Sent: Tuesday, May 6, 2025 1:26 PM

Quoted 6 lines

There are few proposals on table.

1. From Paolo,
- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 46 to 127 Feature bits for the specific device type

Quoted 7 lines

This does not have good reason of why it should still be 127.

2. From me:
- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 45 to 64 Feature bits for the specific device type

This is an extension of Paolo, to justify that implementing feature bits is

extremely hard even for experts as pointed by Paolo.

Quoted 2 lines

It is worth to not extend it further.
RSS can be negotiated via new bit 44 in future bit as OR of 44 and 64 so that

more wider users (Linux, freebsd, qnx, Windows, dpdk pmd) can pick 44.

Quoted 3 lines

3. From me:
Keep the feature bits encoding as is up to 127 bits, because may be there is

(unknown and weird) value in having 127 feature bits.

Quoted 3 lines

(unknown because the reasoning of #1 and #3 mismatch). In that case,
VIRTIO_NET_CTRL_GUEST_OFFLOADS command text to be updated to

indicate

Quoted 1 line

feature fits A to D map to VIRTIO_NET_F_CTRL_GUEST_OFFLOADS.offloads

bits A' to D'.

Quoted 4 lines

I am fine with option #2 and #3.
Doing #1 for sure is wrong.
Wrong because it delays the problem of #1 from this to another feature [A]

who's voting already completed.

Quoted 3 lines

[A]
https://lore.kernel.org/virtio-

Quoted 7 lines
Quoted 5 lines

comment/DM4PR18MB4269F73B786E83EF68A70F

Quoted 1 line

[email protected]/T/#t

#3 seems more conservate.

Are you good with #3?

I'm sorry for the latency. Let me double check to avoid possible misunderstanding; #3 means:

- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 45 to 127 Feature bits for the specific device type

using bits 46-39 for UDP tunnel offloads and likely bit 45 for VIRTIO_NET_F_OUT_NET_HEADER.

The VIRTIO_NET_F_CTRL_GUEST_OFFLOADS mapping should be specified after eventually a new offload feature will be defined using a bit >= 64.

Am I correct?

Thanks,

Paolo

#28 lore raw
Quoted 56 lines

From: Paolo Abeni <[email protected]>
Sent: Tuesday, May 6, 2025 8:09 PM

Hi all,

Quoted 25 lines · Parav Pandit
Quoted 25 lines

From: Michael S. Tsirkin <[email protected]>
Sent: Tuesday, May 6, 2025 1:26 PM

Quoted 6 lines

There are few proposals on table.

1. From Paolo,
- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 46 to 127 Feature bits for the specific device type

Quoted 8 lines

This does not have good reason of why it should still be 127.

2. From me:
- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 45 to 64 Feature bits for the specific device type

This is an extension of Paolo, to justify that implementing feature bits is

extremely hard even for experts as pointed by Paolo.

Quoted 3 lines

It is worth to not extend it further.
RSS can be negotiated via new bit 44 in future bit as OR of 44 and 64 so that

more wider users (Linux, freebsd, qnx, Windows, dpdk pmd) can pick 44.

Quoted 4 lines

3. From me:
Keep the feature bits encoding as is up to 127 bits, because may be there is

Quoted 6 lines
Quoted 6 lines

(unknown and weird) value in having 127 feature bits.

Quoted 3 lines

(unknown because the reasoning of #1 and #3 mismatch). In that case,
VIRTIO_NET_CTRL_GUEST_OFFLOADS command text to be updated to

indicate

Quoted 1 line

feature fits A to D map to

VIRTIO_NET_F_CTRL_GUEST_OFFLOADS.offloads

Quoted 16 lines
Quoted 15 lines

bits A' to D'.

Quoted 5 lines

I am fine with option #2 and #3.
Doing #1 for sure is wrong.
Wrong because it delays the problem of #1 from this to another feature [A]

who's voting already completed.

Quoted 3 lines

[A]
https://lore.kernel.org/virtio-

comment/DM4PR18MB4269F73B786E83EF68A70F

Quoted 1 line

[email protected]/T/#t

#3 seems more conservate.

Quoted 8 lines
Quoted 1 line

Are you good with #3?

I'm sorry for the latency. Let me double check to avoid possible misunderstanding; #3 means:

- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 45 to 127 Feature bits for the specific device type

No change in above feature bits.

Quoted 3 lines

using bits 46-39 for UDP tunnel offloads and likely bit 45 for VIRTIO_NET_F_OUT_NET_HEADER.

This also to use bit 69 as proposed.

Quoted 3 lines

The VIRTIO_NET_F_CTRL_GUEST_OFFLOADS mapping should be specified after eventually a new offload feature will be defined using a bit >= 64.

No. UDP tunnel feature bits 65 to 68 maps to command bits 46,47,48,49. This is the only description change in VIRTIO_NET_F_CTRL_GUEST_OFFLOADS command. Would it work?

In fact, this command extension alone would have been sufficient if the vnet_hdr had a fixed size where new fields could be added.

Quoted 5 lines

Am I correct?

Thanks,

Paolo

#29 lore raw
Quoted 64 lines · Parav Pandit
Quoted 51 lines

From: Paolo Abeni <[email protected]>
Sent: Tuesday, May 6, 2025 8:09 PM

Quoted 25 lines · Parav Pandit
Quoted 25 lines

From: Michael S. Tsirkin <[email protected]>
Sent: Tuesday, May 6, 2025 1:26 PM

Quoted 6 lines

There are few proposals on table.

1. From Paolo,
- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 46 to 127 Feature bits for the specific device type

Quoted 8 lines

This does not have good reason of why it should still be 127.

2. From me:
- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 45 to 64 Feature bits for the specific device type

This is an extension of Paolo, to justify that implementing feature bits is

extremely hard even for experts as pointed by Paolo.

Quoted 3 lines

It is worth to not extend it further.
RSS can be negotiated via new bit 44 in future bit as OR of 44 and 64 so that

more wider users (Linux, freebsd, qnx, Windows, dpdk pmd) can pick 44.

Quoted 4 lines

3. From me:
Keep the feature bits encoding as is up to 127 bits, because may be there is

Quoted 6 lines
Quoted 6 lines

(unknown and weird) value in having 127 feature bits.

Quoted 3 lines

(unknown because the reasoning of #1 and #3 mismatch). In that case,
VIRTIO_NET_CTRL_GUEST_OFFLOADS command text to be updated to

indicate

Quoted 1 line

feature fits A to D map to

VIRTIO_NET_F_CTRL_GUEST_OFFLOADS.offloads

Quoted 16 lines
Quoted 15 lines

bits A' to D'.

Quoted 5 lines

I am fine with option #2 and #3.
Doing #1 for sure is wrong.
Wrong because it delays the problem of #1 from this to another feature [A]

who's voting already completed.

Quoted 3 lines

[A]
https://lore.kernel.org/virtio-

comment/DM4PR18MB4269F73B786E83EF68A70F

Quoted 1 line

[email protected]/T/#t

#3 seems more conservate.

Quoted 8 lines
Quoted 1 line

Are you good with #3?

I'm sorry for the latency. Let me double check to avoid possible misunderstanding; #3 means:

- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 45 to 127 Feature bits for the specific device type

No change in above feature bits.

Quoted 3 lines

using bits 46-39 for UDP tunnel offloads and likely bit 45 for VIRTIO_NET_F_OUT_NET_HEADER.

Quoted 8 lines

This also to use bit 69 as proposed.

Quoted 3 lines

The VIRTIO_NET_F_CTRL_GUEST_OFFLOADS mapping should be specified after eventually a new offload feature will be defined using a bit >= 64.

No. UDP tunnel feature bits 65 to 68 maps to command bits 46,47,48,49. This is the only description change in VIRTIO_NET_F_CTRL_GUEST_OFFLOADS command. Would it work?

AFAICT, yes, it should work.

But it will not avoid the immediate need to expand the virtio features negotiation above 64 bits, with the already mentioned complexity.

I would preferably avoid that, if possible: I restarted this thread with such a goal.

Note: the implementation complexity spark from u64 hard-coded "everywhere" in both the kernel, qemu and the related APIs.

Quoted 1 line

In fact, this command extension alone would have been sufficient if the vnet_hdr had a fixed size where new fields could be added.

I could not parse this last statement. The virtio_net_hdr size depends on the negotiated features, and AFAICT VIRTIO_NET_F_CTRL_GUEST_OFFLOADS can't replace features negotiation. Could you please rephrase?

Thanks,

Paolo

#30 lore raw
Quoted 80 lines

From: Paolo Abeni <[email protected]>
Sent: Tuesday, May 6, 2025 9:10 PM

Quoted 64 lines · Parav Pandit
Quoted 51 lines

From: Paolo Abeni <[email protected]>
Sent: Tuesday, May 6, 2025 8:09 PM

Quoted 25 lines · Parav Pandit
Quoted 25 lines

From: Michael S. Tsirkin <[email protected]>
Sent: Tuesday, May 6, 2025 1:26 PM

Quoted 6 lines

There are few proposals on table.

1. From Paolo,
- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 46 to 127 Feature bits for the specific device type

Quoted 8 lines

This does not have good reason of why it should still be 127.

2. From me:
- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 45 to 64 Feature bits for the specific device type

This is an extension of Paolo, to justify that implementing feature bits is

extremely hard even for experts as pointed by Paolo.

Quoted 3 lines

It is worth to not extend it further.
RSS can be negotiated via new bit 44 in future bit as OR of 44 and 64 so that

more wider users (Linux, freebsd, qnx, Windows, dpdk pmd) can pick 44.

Quoted 4 lines

3. From me:
Keep the feature bits encoding as is up to 127 bits, because may be there is

Quoted 6 lines
Quoted 6 lines

(unknown and weird) value in having 127 feature bits.

Quoted 3 lines

(unknown because the reasoning of #1 and #3 mismatch). In that case,
VIRTIO_NET_CTRL_GUEST_OFFLOADS command text to be updated to

indicate

Quoted 1 line

feature fits A to D map to

VIRTIO_NET_F_CTRL_GUEST_OFFLOADS.offloads

Quoted 16 lines
Quoted 15 lines

bits A' to D'.

Quoted 5 lines

I am fine with option #2 and #3.
Doing #1 for sure is wrong.
Wrong because it delays the problem of #1 from this to another feature [A]

who's voting already completed.

Quoted 3 lines

[A]
https://lore.kernel.org/virtio-

comment/DM4PR18MB4269F73B786E83EF68A70F

Quoted 1 line

[email protected]/T/#t

#3 seems more conservate.

Quoted 8 lines
Quoted 1 line

Are you good with #3?

I'm sorry for the latency. Let me double check to avoid possible misunderstanding; #3 means:

- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 45 to 127 Feature bits for the specific device type

No change in above feature bits.

Quoted 3 lines

using bits 46-39 for UDP tunnel offloads and likely bit 45 for VIRTIO_NET_F_OUT_NET_HEADER.

Quoted 7 lines

This also to use bit 69 as proposed.

Quoted 3 lines

The VIRTIO_NET_F_CTRL_GUEST_OFFLOADS mapping should be specified after eventually a new offload feature will be defined using a bit >= 64.

No. UDP tunnel feature bits 65 to 68 maps to command bits 46,47,48,49. This is the only description change in

VIRTIO_NET_F_CTRL_GUEST_OFFLOADS command.

Quoted 1 line

Would it work?

Quoted 8 lines

AFAICT, yes, it should work.

But it will not avoid the immediate need to expand the virtio features negotiation above 64 bits, with the already mentioned complexity.

I would preferably avoid that, if possible: I restarted this thread with such a goal.

In that case we should adopt #2.

Quoted 3 lines

Note: the implementation complexity spark from u64 hard-coded "everywhere" in both the kernel, qemu and the related APIs.

Hence it applies to already 2nd feature which also needs new feature bit.

Quoted 7 lines
Quoted 1 line

In fact, this command extension alone would have been sufficient if the

vnet_hdr had a fixed size where new fields could be added.

I could not parse this last statement. The virtio_net_hdr size depends on the negotiated features, and AFAICT VIRTIO_NET_F_CTRL_GUEST_OFFLOADS can't replace features negotiation. Could you please rephrase?

I meant to say, if we have fixed size vnet hdr, say 64B, than there is no need of feature negotiation per offload. VIRTIO_NET_F_CTRL_GUEST_OFFLOADS or any other command can dynamically enable/disable a feature on invocation of ethtool/netlink/devlink or similar OS UAPIs. It does not apply presently to this feature as the infrastructure of 64B vnet hdr is not in present and it cannot a fix anyway for this issue. It was just a forward thinking on how to build the optimize data path.

Quoted 3 lines

Thanks,

Paolo

#31 lore raw
Quoted 29 lines · Parav Pandit

From: Paolo Abeni <[email protected] Sent: Tuesday, May 6, 2025 9:10 PM

Quoted 28 lines · Parav Pandit
Quoted 15 lines

From: Paolo Abeni <[email protected] Sent: Tuesday, May 6, 2025 8:09 PM

Quoted 9 lines · Parav Pandit
Quoted 1 line

Are you good with #3?

I'm sorry for the latency. Let me double check to avoid possible misunderstanding; #3 means:

- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 45 to 127 Feature bits for the specific device type

No change in above feature bits.

Quoted 3 lines

using bits 46-39 for UDP tunnel offloads and likely bit 45 for VIRTIO_NET_F_OUT_NET_HEADER.

Quoted 7 lines

This also to use bit 69 as proposed.

Quoted 3 lines

The VIRTIO_NET_F_CTRL_GUEST_OFFLOADS mapping should be specified after eventually a new offload feature will be defined using a bit >= 64.

No. UDP tunnel feature bits 65 to 68 maps to command bits 46,47,48,49. This is the only description change in

VIRTIO_NET_F_CTRL_GUEST_OFFLOADS command.

Quoted 1 line

Would it work?

AFAICT, yes, it should work.

Quoted 7 lines
Quoted 6 lines

But it will not avoid the immediate need to expand the virtio features negotiation above 64 bits, with the already mentioned complexity.

I would preferably avoid that, if possible: I restarted this thread with such a goal.

In that case we should adopt #2.

Do we have quorum? Should I send a patch?

Quoted 11 lines
Quoted 7 lines
Quoted 1 line

In fact, this command extension alone would have been sufficient if the

vnet_hdr had a fixed size where new fields could be added.

I could not parse this last statement. The virtio_net_hdr size depends on the negotiated features, and AFAICT VIRTIO_NET_F_CTRL_GUEST_OFFLOADS can't replace features negotiation. Could you please rephrase?

I meant to say, if we have fixed size vnet hdr, say 64B, than there is no need of feature negotiation per offload. VIRTIO_NET_F_CTRL_GUEST_OFFLOADS or any other command can dynamically enable/disable a feature on invocation of ethtool/netlink/devlink or similar OS UAPIs. It does not apply presently to this feature as the infrastructure of 64B vnet hdr is not in present and it cannot a fix anyway for this issue. It was just a forward thinking on how to build the optimize data path.

Understood. I'm unsure if that schema could have issues with migration vs VMs with different offloads, but let's put that aside from the main topic here ;)

Thanks,

Paolo

#32 lore raw
Quoted 33 lines · Paolo Abeni
Quoted 32 lines · Parav Pandit

From: Paolo Abeni <[email protected] Sent: Tuesday, May 6, 2025 9:10 PM

Quoted 31 lines · Parav Pandit
Quoted 15 lines

From: Paolo Abeni <[email protected] Sent: Tuesday, May 6, 2025 8:09 PM

Quoted 9 lines · Parav Pandit
Quoted 1 line

Are you good with #3?

I'm sorry for the latency. Let me double check to avoid possible misunderstanding; #3 means:

- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 45 to 127 Feature bits for the specific device type

No change in above feature bits.

Quoted 3 lines

using bits 46-39 for UDP tunnel offloads and likely bit 45 for VIRTIO_NET_F_OUT_NET_HEADER.

Quoted 7 lines

This also to use bit 69 as proposed.

Quoted 3 lines

The VIRTIO_NET_F_CTRL_GUEST_OFFLOADS mapping should be specified after eventually a new offload feature will be defined using a bit >= 64.

No. UDP tunnel feature bits 65 to 68 maps to command bits 46,47,48,49. This is the only description change in

VIRTIO_NET_F_CTRL_GUEST_OFFLOADS command.

Quoted 1 line

Would it work?

AFAICT, yes, it should work.

But it will not avoid the immediate need to expand the virtio features negotiation above 64 bits, with the already mentioned complexity.

Quoted 6 lines
Quoted 4 lines
Quoted 3 lines

I would preferably avoid that, if possible: I restarted this thread with such a goal.

In that case we should adopt #2.

Do we have quorum? Should I send a patch?

I'm curious how hard is 128 bit support, gimme a couple of days to try and maybe post a patch, just for comparison.

#33 lore raw
In reply to Paolo Abeni
Quoted 33 lines · Paolo Abeni
Quoted 32 lines · Parav Pandit

From: Paolo Abeni <[email protected] Sent: Tuesday, May 6, 2025 9:10 PM

Quoted 31 lines · Parav Pandit
Quoted 15 lines

From: Paolo Abeni <[email protected] Sent: Tuesday, May 6, 2025 8:09 PM

Quoted 9 lines · Parav Pandit
Quoted 1 line

Are you good with #3?

I'm sorry for the latency. Let me double check to avoid possible misunderstanding; #3 means:

- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 45 to 127 Feature bits for the specific device type

No change in above feature bits.

Quoted 3 lines

using bits 46-39 for UDP tunnel offloads and likely bit 45 for VIRTIO_NET_F_OUT_NET_HEADER.

Quoted 7 lines

This also to use bit 69 as proposed.

Quoted 3 lines

The VIRTIO_NET_F_CTRL_GUEST_OFFLOADS mapping should be specified after eventually a new offload feature will be defined using a bit >= 64.

No. UDP tunnel feature bits 65 to 68 maps to command bits 46,47,48,49. This is the only description change in

VIRTIO_NET_F_CTRL_GUEST_OFFLOADS command.

Quoted 1 line

Would it work?

AFAICT, yes, it should work.

But it will not avoid the immediate need to expand the virtio features negotiation above 64 bits, with the already mentioned complexity.

Quoted 6 lines
Quoted 4 lines
Quoted 3 lines

I would preferably avoid that, if possible: I restarted this thread with such a goal.

In that case we should adopt #2.

Do we have quorum? Should I send a patch?

As per off-list discussion with Michel, there is no agreement on reserved bits re-use.

That means the only available option is #3 above.

@Parav, would you send a patch to fix the offloads <> features mapping, or do you prefer I'll do it?

Thanks,

Paolo

#34 lore raw
Quoted 50 lines

From: Paolo Abeni <[email protected]>
Sent: Monday, May 19, 2025 2:28 PM

Quoted 15 lines · Paolo Abeni
Quoted 14 lines · Parav Pandit

From: Paolo Abeni <[email protected] Sent: Tuesday, May 6, 2025 9:10 PM

Quoted 12 lines · Parav Pandit
Quoted 11 lines

From: Paolo Abeni <[email protected] Sent: Tuesday, May 6, 2025 8:09 PM

Quoted 9 lines · Parav Pandit
Quoted 1 line

Are you good with #3?

I'm sorry for the latency. Let me double check to avoid possible misunderstanding; #3 means:

- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 45 to 127 Feature bits for the specific device type

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines

No change in above feature bits.

Quoted 3 lines

using bits 46-39 for UDP tunnel offloads and likely bit 45 for VIRTIO_NET_F_OUT_NET_HEADER.

This also to use bit 69 as proposed.

Quoted 1 line

The VIRTIO_NET_F_CTRL_GUEST_OFFLOADS mapping should be

specified

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 2 lines

after eventually a new offload feature will be defined using a bit >= 64.

No. UDP tunnel feature bits 65 to 68 maps to command bits

46,47,48,49.

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 1 line

This is the only description change in

VIRTIO_NET_F_CTRL_GUEST_OFFLOADS command.

Quoted 1 line

Would it work?

AFAICT, yes, it should work.

But it will not avoid the immediate need to expand the virtio features negotiation above 64 bits, with the already mentioned

complexity.

Quoted 7 lines
Quoted 5 lines
Quoted 4 lines

I would preferably avoid that, if possible: I restarted this thread with such a goal.

In that case we should adopt #2.

Do we have quorum? Should I send a patch?

Quoted 8 lines

As per off-list discussion with Michel, there is no agreement on reserved bits re-use.

That means the only available option is #3 above.

@Parav, would you send a patch to fix the offloads <> features mapping, or do you prefer I'll do it?

If its urgent, please do it.
If it can wait till 22 May this week, I will prepare one.

Quoted 3 lines

Thanks,

Paolo

#35 lore raw
Quoted 51 lines · Parav Pandit
Quoted 51 lines

From: Paolo Abeni <[email protected]>
Sent: Monday, May 19, 2025 2:28 PM

Quoted 15 lines · Paolo Abeni
Quoted 14 lines · Parav Pandit

From: Paolo Abeni <[email protected] Sent: Tuesday, May 6, 2025 9:10 PM

Quoted 12 lines · Parav Pandit
Quoted 11 lines

From: Paolo Abeni <[email protected] Sent: Tuesday, May 6, 2025 8:09 PM

Quoted 9 lines · Parav Pandit
Quoted 1 line

Are you good with #3?

I'm sorry for the latency. Let me double check to avoid possible misunderstanding; #3 means:

- 0 to 23, and 50 to 127 Feature bits for the specific device type
+ 0 to 23, and 45 to 127 Feature bits for the specific device type

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines

No change in above feature bits.

Quoted 3 lines

using bits 46-39 for UDP tunnel offloads and likely bit 45 for VIRTIO_NET_F_OUT_NET_HEADER.

This also to use bit 69 as proposed.

Quoted 1 line

The VIRTIO_NET_F_CTRL_GUEST_OFFLOADS mapping should be

specified

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 2 lines

after eventually a new offload feature will be defined using a bit >= 64.

No. UDP tunnel feature bits 65 to 68 maps to command bits

46,47,48,49.

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 1 line

This is the only description change in

VIRTIO_NET_F_CTRL_GUEST_OFFLOADS command.

Quoted 1 line

Would it work?

AFAICT, yes, it should work.

But it will not avoid the immediate need to expand the virtio features negotiation above 64 bits, with the already mentioned

complexity.

Quoted 7 lines
Quoted 5 lines
Quoted 4 lines

I would preferably avoid that, if possible: I restarted this thread with such a goal.

In that case we should adopt #2.

Do we have quorum? Should I send a patch?

As per off-list discussion with Michel, there is no agreement on reserved bits re-use.

Quoted 7 lines
Quoted 5 lines

That means the only available option is #3 above.

@Parav, would you send a patch to fix the offloads <> features mapping, or do you prefer I'll do it?

If its urgent, please do it.
If it can wait till 22 May this week, I will prepare one.

I'll do, because I want to move this thing forward. Note that I think it's better to change the offload bit name definition for UDP related offload, to avoid multiple definitions using specifying the string and different value. i.e.:

---

(quoted diff) +0 -0

b/device-types/net/description.tex
index 1b6b54d..5d869bc 100644

device-types/net/description.tex +0 -0
@@ -2186,8 +2186,8 @@ \subsubsection{Control Virtqueue}\label{sec:Device

Types / Network Device / Devi

 #define VIRTIO_NET_F_GUEST_TSO6       8
 #define VIRTIO_NET_F_GUEST_ECN        9
 #define VIRTIO_NET_F_GUEST_UFO        10

-#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO 46 -#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_CSUM 47 +#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_MAPPED 46 +#define VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO_CSUM_MAPPED 47
#define VIRTIO_NET_F_GUEST_USO4 54
#define VIRTIO_NET_F_GUEST_USO6 55

@@ -2205,6 +2205,9 @@ \subsubsection{Control Virtqueue}\label{sec:Device Types / Network Device / Devi
negotiation corresponding offload gets enabled to preserve backward compatibility.

+Note that device features bit in the [65 to 68] range correspond to +offloads bits in the [46 to 49] range.
+
\drivernormative{\subparagraph}{Setting Offloads State}{Device Types /
Network Device / Device Operation / Control Virtqueue / Offloads State Configuration / Setting Offloads State}

A driver MUST NOT enable an offload for which the appropriate feature