[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: Re: [virtio-dev] Re: [PATCH v2] virtio_net: support split header
On Tue, May 31, 2022 at 2:23 AM Michael S. Tsirkin <mst@redhat.com> wrote: > > On Mon, May 30, 2022 at 01:56:06PM +0800, Jason Wang wrote: > > On Thu, May 26, 2022 at 2:31 PM Xuan Zhuo <xuanzhuo@linux.alibaba.com> wrote: > > > > > > On Mon, 9 May 2022 16:41:11 +0800, Jason Wang <jasowang@redhat.com> wrote: > > > > > > > > å 2022/5/7 15:15, Xuan Zhuo åé: > > > > > The purpose of this feature is to split the header and the payload of > > > > > the packet. > > > > > > > > > > | receive buffer | > > > > > | 0th descriptor | 1th descriptor | > > > > > | virtnet hdr | mac | ip hdr | tcp hdr|<-- hold -->| payload | > > > > > > > > > > We can use a buffer plus a separate page when allocating the receive > > > > > buffer. In this way, we can ensure that all payloads can be > > > > > independently in a page, which is very beneficial for the zerocopy > > > > > implemented by the upper layer. > > > > > > > > > > Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com> > > > > > --- > > > > > conformance.tex | 2 ++ > > > > > content.tex | 72 +++++++++++++++++++++++++++++++++++++++++++++++++ > > > > > 2 files changed, 74 insertions(+) > > > > > > > > > > diff --git a/conformance.tex b/conformance.tex > > > > > index 663e7c3..6f561fb 100644 > > > > > --- a/conformance.tex > > > > > +++ b/conformance.tex > > > > > @@ -149,6 +149,7 @@ \section{Conformance Targets}\label{sec:Conformance / Conformance Targets} > > > > > \item \ref{drivernormative:Device Types / Network Device / Device Operation / Control Virtqueue / Automatic receive steering in multiqueue mode} > > > > > \item \ref{drivernormative:Device Types / Network Device / Device Operation / Control Virtqueue / Offloads State Configuration / Setting Offloads State} > > > > > \item \ref{drivernormative:Device Types / Network Device / Device Operation / Control Virtqueue / Receive-side scaling (RSS) } > > > > > +\item \ref{drivernormative:Device Types / Network Device / Device Operation / Control Virtqueue / Split Header} > > > > > \end{itemize} > > > > > > > > > > \conformance{\subsection}{Block Driver Conformance}\label{sec:Conformance / Driver Conformance / Block Driver Conformance} > > > > > @@ -411,6 +412,7 @@ \section{Conformance Targets}\label{sec:Conformance / Conformance Targets} > > > > > \item \ref{devicenormative:Device Types / Network Device / Device Operation / Control Virtqueue / Gratuitous Packet Sending} > > > > > \item \ref{devicenormative:Device Types / Network Device / Device Operation / Control Virtqueue / Automatic receive steering in multiqueue mode} > > > > > \item \ref{devicenormative:Device Types / Network Device / Device Operation / Control Virtqueue / Receive-side scaling (RSS) / RSS processing} > > > > > +\item \ref{devicenormative:Device Types / Network Device / Device Operation / Control Virtqueue / Split Header} > > > > > \end{itemize} > > > > > > > > > > \conformance{\subsection}{Block Device Conformance}\label{sec:Conformance / Device Conformance / Block Device Conformance} > > > > > diff --git a/content.tex b/content.tex > > > > > index 060bdab..3340402 100644 > > > > > --- a/content.tex > > > > > +++ b/content.tex > > > > > @@ -3092,6 +3092,9 @@ \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_SPLIT_HEADER (55)] Device supports to split the header and > > > > > + the payload. > > > > > + > > > > > \item[VIRTIO_NET_F_HOST_USO (56)] Device can receive USO packets. Unlike UFO > > > > > (fragmenting the packet) the USO splits large UDP packet > > > > > to several segments when each of these smaller packets has UDP header. > > > > > @@ -3139,6 +3142,7 @@ \subsubsection{Feature bit requirements}\label{sec:Device Types / Network Device > > > > > \item[VIRTIO_NET_F_CTRL_MAC_ADDR] Requires VIRTIO_NET_F_CTRL_VQ. > > > > > \item[VIRTIO_NET_F_RSC_EXT] Requires VIRTIO_NET_F_HOST_TSO4 or VIRTIO_NET_F_HOST_TSO6. > > > > > \item[VIRTIO_NET_F_RSS] Requires VIRTIO_NET_F_CTRL_VQ. > > > > > +\item[VIRTIO_NET_F_SPLIT_HEADER] Requires VIRTIO_NET_F_CTRL_VQ. > > > > > \end{description} > > > > > > > > > > \subsubsection{Legacy Interface: Feature bits}\label{sec:Device Types / Network Device / Feature bits / Legacy Interface: Feature bits} > > > > > @@ -3370,6 +3374,7 @@ \subsection{Device Operation}\label{sec:Device Types / Network Device / Device O > > > > > #define VIRTIO_NET_HDR_F_NEEDS_CSUM 1 > > > > > #define VIRTIO_NET_HDR_F_DATA_VALID 2 > > > > > #define VIRTIO_NET_HDR_F_RSC_INFO 4 > > > > > +#define VIRTIO_NET_HDR_F_SPLIT_HEADER 8 > > > > > u8 flags; > > > > > #define VIRTIO_NET_HDR_GSO_NONE 0 > > > > > #define VIRTIO_NET_HDR_GSO_TCPV4 1 > > > > > @@ -4471,6 +4476,73 @@ \subsubsection{Control Virtqueue}\label{sec:Device Types / Network Device / Devi > > > > > according to the native endian of the guest rather than > > > > > (necessarily when not using the legacy interface) little-endian. > > > > > > > > > > +\paragraph{Split Header}\label{sec:Device Types / Network Device / Device Operation / Control Virtqueue / Split Header} > > > > > + > > > > > +If the VIRTIO_NET_F_SPLIT_HEADER feature is negotiated, > > > > > +the device supports to split the header and the payload. > > > > > +The header and payload will be separated into different buffers. > > > > > > > > > > > > I think you meant "descriptors" instead of "buffers". > > > > > > Will fix. > > No, actually, I think this is exactly buffers. > device is always in terms of buffers, descriptors is > a means to implement buffers. > > And the feature should > depend on mergeable buffers. Without mergeable buffers you do not > need these tricks, just add two descriptors and be done > with it. How do we know the length of a header in advance to allocate the exact descriptor properly in this case? Thanks > > Right? > > > > > > > > > > > > > > > > > + > > > > > +\subparagraph{Split Header}\label{sec:Device Types / Network Device / Device Operation / Control Virtqueue / Split Header / Setting Split Header} > > > > > + > > > > > +To configure the split header, the following layout structure and definitions > > > > > +are used: > > > > > + > > > > > +\begin{lstlisting} > > > > > +struct virtio_net_split_header_config { > > > > > +#define VIRTIO_NET_SPLIT_HEADER_TYPE_TCPv4 1 > > > > > > > > > > > > I think it's better to capitalize: TCPV4 or TCP4. > > > > > > OK. > > > > > > > > > > > > > > > > +#define VIRTIO_NET_SPLIT_HEADER_TYPE_TCPv6 2 > > > > > +#define VIRTIO_NET_SPLIT_HEADER_TYPE_UDPv4 4 > > > > > +#define VIRTIO_NET_SPLIT_HEADER_TYPE_UDPv6 8 > > > > > + le64 type; > > > > > +}; > > > > > + > > > > > +#define VIRTIO_NET_CTRL_SPLIT_HEADER 6 > > > > > + #define VIRTIO_NET_CTRL_SPLIT_HEADER_SET 0 > > > > > +\end{lstlisting} > > > > > + > > > > > +The class VIRTIO_NET_CTRL_SPLIT_HEADER has one command: > > > > > +VIRTIO_NET_CTRL_SPLIT_HEADER_SET applies the new split header configuration. > > > > > + > > > > > +\field{type} passed as command data is a bitmask, bits set define > > > > > +packet types to split header, bits cleared - split header to be disabled. > > > > > > > > > > > > (Not a native speaker, try my best to help) > > > > > > > > This sentence might need some tweaks: > > > > > > > > The driver can enable or disable the split by setting or clearing > > > > corresponding bits in \field{type}. > > > > > > Thanks. > > > > > > > > > > > > > > > > + > > > > > +The header contains the struct virtio_net_hdr and the header of the package. > > > > > +Such as \field{VIRTIO_NET_SPLIT_HEADER_TYPE_TCPv4} specified header contains > > > > > +virtio_net_hdr, MAC header, IPv4 header (including IPv4 options), TCP header > > > > > +(include TCP options). The back part is the payload. > > > > > > > > > > > > I think it's better to either explain each type one by one instead of > > > > only describing the TCPv4. Or we can simply say the head contains all > > > > headers before level 4 payload? > > > > > > OK. Thanks. > > > > So this always splits out at protocol level? > What if device wants to align e.g. IP protocol header? > Ethernet? > > > Also, what about aligning at an offset? > E.g. NET_IP_ALIGN? > > > > > > > > > > > > > > > > > > + > > > > > +\devicenormative{\subparagraph}{Setting Split Header}{Device Types / Network Device / Device Operation / Control Virtqueue / Split Header} > > > > > + > > > > > +Split header MUST be disabled after device initialization. > > > > > > > > > > > > I think you actually mean "upon device reset"? > > > > > > Yes. > > > > > > > > > > > > > > > > > > > + > > > > > +A device MUST NOT perform split header in the following cases: > > > > > +\begin{itemize} > > > > > + \item device does not recognize protocol of the packet. > > > > > + \item \field{type} does not include the protocol of the packet. > > > > > + \item the packet is a IP fragmentation. > > > > > > > > > > > > What's the reason for this limit? > > > > > > Rethinking this problem, some implementation difficulties that I originally > > > thought can be solved. So I think this limitation can be removed. > > > > > > > > > > > > > > > > + \item the receive buffer consists of only one descriptor. > > > > > > > > > > > > descriptor actually. > > > > > > Will fix. > > > > > > > > > > > > > > > > + \item the header exceeds the size of the 0th descriptor. > > > > > > > > > > > > It looks to me the spec count from the first. > > > > > > Will fix. > > > > > > > > > > > > > > > > + \item If VIRTIO_NET_F_MRG_RXBUF is not negotiated and the size of the > > > > > + payload is greater than the total size of the 1th\ldots Nth descriptor. > > > > > > > > > > > > Maybe something like the following is better: > > > > > > > > The size of the payload exceeds the length of the descriptor or buffer > > > > chain starting from the 2nd descriptor. > > > > > > OK. > > > > > > > > > > > > > > > > +\end{itemize} > > > > > + > > > > > +If the split header completed, > > > > > > > > > > > > If the header is split by the device? > > > > > > Yes. > > > > > > > > > > > > > > > > then the \field{flags} of virtnet hdr MUST > > > > > > > > > > > > We can drop "then" here and it's better to use structure virtio_net_hdr > > > > > > Will fix. > > > > > > > > > > > > > > > > +contains VIRTIO_NET_HDR_F_SPLIT_HEADER. The header MUST is on the buffer of the > > > > > +0th descriptor, > > > > > > > > > > > > first and second should be used here. > > > > > > OK > > > > > > > > > > > > > > > > and the payload MUST starts from the buffer of the 1th descriptor. > > > > > +The device MUST set \field{hdr_len} of virtnet hdr. > > > > > > > > > > > > To what value? > > > > > > I'll make it clear in the next version. > > > > > > > > > > > > > > > > + > > > > > +If VIRTIO_NET_F_MRG_RXBUF is negotiated and the device is to use multiple > > > > > +receive buffers, each subsequent receive buffer MUST skip the 0th descriptor. > > > > > > > > > > > > I don't see why we need this. > > > > > > If used to implement zerocopy, the second desc points to a page-aligned buffer. > > > The first desc points to a small buffer for saving the header. The first buffer > > > and subsequent buffers are generally discontinuous. So giving up the first > > > buffer and directly guaranteeing the data to the page-aligned buffer can achieve > > > better performance. And the driver can reuse the first buffer. > > > > I see, it might be better if we can explain this here. > > > > Thanks > > > > > > > > > > > > > > > > > > + > > > > > +\drivernormative{\subparagraph}{Setting Split Header}{Device Types / Network Device / Device Operation / Control Virtqueue / Split Header} > > > > > + > > > > > +If VIRTIO_NET_HDR_F_SPLIT_HEADER bit in \field{flags} is set, the driver MUST > > > > > +believe \field{hdr_len}, the length of the header in the 0th descriptor is equal > > > > > +to the length of struct virtio_net_hdr plus \field{hdr_len}. > > > > > > > > > > > > See above we can reuse the device normative an drop the second part > > > > starting from " the length of the header ..." > > > > > > OK > > > > > > > > > > > > > > > > + > > > > > +If the split header function is enable, the buffers submitted by the driver > > > > > +SHOULD at least be composed of two descriptors. > > > > > > > > > > > > Do we need to mention it's only used for RX? (I think we don't need this > > > > for TX). > > > > > > Yes. > > > > > > > > > > > > > > > > The buffer specified by the 0th > > > > > +descriptor SHOULD be able to accommodate the header. > > > > > > > > > > > > We probably need to clarify that the header here means both the virtio > > > > net header and the protocol header. > > > > > > Will fix. > > > > > > Thanks. > > > > > > > > > > > Thanks > > > > > > > > > > > > > > > > > > \subsubsection{Legacy Interface: Framing Requirements}\label{sec:Device > > > > > Types / Network Device / Legacy Interface: Framing Requirements} > > > > > > > > > > > > --------------------------------------------------------------------- > > > > To unsubscribe, e-mail: virtio-dev-unsubscribe@lists.oasis-open.org > > > > For additional commands, e-mail: virtio-dev-help@lists.oasis-open.org > > > > > > > >
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]