[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: Re: [PATCH v4 1/1] virtio-net: define support for receive-side scaling
On Tue, Nov 12, 2019 at 07:37:48AM +0200, Yuri Benditovich wrote: > Fixes: https://github.com/oasis-tcs/virtio-spec/issues/48 > Added support for RSS receive steering mode. > > Signed-off-by: Yuri Benditovich <yuri.benditovich@daynix.com> Looks good to me. Big change and some people could be at DPDK conf, so pls give it a week or so, then request voting if no more comments. Thanks! > --- > conformance.tex | 2 + > content.tex | 234 +++++++++++++++++++++++++++++++++++++++++++----- > 2 files changed, 213 insertions(+), 23 deletions(-) > > diff --git a/conformance.tex b/conformance.tex > index 0ac58aa..01449c5 100644 > --- a/conformance.tex > +++ b/conformance.tex > @@ -101,6 +101,7 @@ \section{Conformance Targets}\label{sec:Conformance / Conformance Targets} > \item \ref{drivernormative:Device Types / Network Device / Device Operation / Control Virtqueue / Gratuitous Packet Sending} > \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) } > \end{itemize} > > \conformance{\subsection}{Block Driver Conformance}\label{sec:Conformance / Driver Conformance / Block Driver Conformance} > @@ -257,6 +258,7 @@ \section{Conformance Targets}\label{sec:Conformance / Conformance Targets} > \item \ref{devicenormative:Device Types / Network Device / Device Operation / Control Virtqueue / Setting MAC Address Filtering} > \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} > \end{itemize} > > \conformance{\subsection}{Block Device Conformance}\label{sec:Conformance / Device Conformance / Block Device Conformance} > diff --git a/content.tex b/content.tex > index 679391e..7a7abd2 100644 > --- a/content.tex > +++ b/content.tex > @@ -2752,7 +2752,7 @@ \subsection{Virtqueues}\label{sec:Device Types / Network Device / Virtqueues} > \item[2N] controlq > \end{description} > > - N=1 if VIRTIO_NET_F_MQ is not negotiated, otherwise N is set by > + N=1 if neither VIRTIO_NET_F_MQ nor VIRTIO_NET_F_RSS are negotiated, otherwise N is set by > \field{max_virtqueue_pairs}. > > controlq only exists if VIRTIO_NET_F_CTRL_VQ set. > @@ -2811,6 +2811,9 @@ \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_RSS(60)] Device supports RSS (receive-side scaling) > + with Toeplitz hash calculation and configurable hash parameters for receive steering > + > \item[VIRTIO_NET_F_RSC_EXT(61)] Device can process duplicated ACKs > and report number of coalesced segments and duplicated ACKs > > @@ -2840,6 +2843,7 @@ \subsubsection{Feature bit requirements}\label{sec:Device Types / Network Device > \item[VIRTIO_NET_F_MQ] Requires VIRTIO_NET_F_CTRL_VQ. > \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. > \end{description} > > \subsubsection{Legacy Interface: Feature bits}\label{sec:Device Types / Network Device / Feature bits / Legacy Interface: Feature bits} > @@ -2854,7 +2858,7 @@ \subsubsection{Legacy Interface: Feature bits}\label{sec:Device Types / Network > \subsection{Device configuration layout}\label{sec:Device Types / Network Device / Device configuration layout} > \label{sec:Device Types / Block Device / Feature bits / Device configuration layout} > > -Three driver-read-only configuration fields are currently defined. The \field{mac} address field > +Device configuration fields are listed below, they are read-only for a driver. The \field{mac} address field > always exists (though is only valid if VIRTIO_NET_F_MAC is set), and > \field{status} only exists if VIRTIO_NET_F_STATUS is set. Two > read-only bits (for the driver) are currently defined for the status field: > @@ -2866,23 +2870,58 @@ \subsection{Device configuration layout}\label{sec:Device Types / Network Device > \end{lstlisting} > > The following driver-read-only field, \field{max_virtqueue_pairs} only exists if > -VIRTIO_NET_F_MQ is set. This field specifies the maximum number > +VIRTIO_NET_F_MQ or VIRTIO_NET_F_RSS is set. This field specifies the maximum number > of each of transmit and receive virtqueues (receiveq1\ldots receiveqN > -and transmitq1\ldots transmitqN respectively) that can be configured once VIRTIO_NET_F_MQ > +and transmitq1\ldots transmitqN respectively) that can be configured once at least one of these features > is negotiated. > > The following driver-read-only field, \field{mtu} only exists if > VIRTIO_NET_F_MTU is set. This field specifies the maximum MTU for the driver to > use. > > +Two following fields, \field{speed} and \field{duplex} are reserved. > \begin{lstlisting} > struct virtio_net_config { > u8 mac[6]; > le16 status; > le16 max_virtqueue_pairs; > le16 mtu; > + le32 speed; > + u8 duplex; > + u8 rss_max_key_size; > + le16 rss_max_indirection_table_length; > + le32 supported_hash_types; > }; > \end{lstlisting} > +\label{sec:Device Types / Network Device / Device configuration layout / RSS} > +Three following fields, \field{rss_max_key_size}, \field{rss_max_indirection_table_length} > +and \field{supported_hash_types} only exist if VIRTIO_NET_F_RSS is set. > + > +Field \field{rss_max_key_size} specifies maximal supported length of RSS key in bytes. > + > +Field \field{rss_max_indirection_table_length} specifies maximal number of 16-bit entries in RSS indirection table. > + > +Field \field{supported_hash_types} contains bitmask of supported RSS hash types. > + > +Hash types applicable for IPv4 packets: > +\begin{lstlisting} > +#define VIRTIO_NET_RSS_HASH_TYPE_IPv4 (1 << 0) > +#define VIRTIO_NET_RSS_HASH_TYPE_TCPv4 (1 << 1) > +#define VIRTIO_NET_RSS_HASH_TYPE_UDPv4 (1 << 2) > +\end{lstlisting} > +Hash types applicable for IPv6 packets without extension headers > +\begin{lstlisting} > +#define VIRTIO_NET_RSS_HASH_TYPE_IPv6 (1 << 3) > +#define VIRTIO_NET_RSS_HASH_TYPE_TCPv6 (1 << 4) > +#define VIRTIO_NET_RSS_HASH_TYPE_UDPv6 (1 << 5) > +\end{lstlisting} > +Hash types applicable for IPv6 packets with extension headers > +\begin{lstlisting} > +#define VIRTIO_NET_RSS_HASH_TYPE_IP_EX (1 << 6) > +#define VIRTIO_NET_RSS_HASH_TYPE_TCP_EX (1 << 7) > +#define VIRTIO_NET_RSS_HASH_TYPE_UDP_EX (1 << 8) > +\end{lstlisting} > +For exact meaning of VIRTIO_NET_RSS_HASH_TYPE_ flags see \ref{sec:Device Types / Network Device / Device Operation / Control Virtqueue / Receive-side scaling (RSS) / RSS hash types}. > > \devicenormative{\subsubsection}{Device configuration layout}{Device Types / Network Device / Device configuration layout} > > @@ -2906,6 +2945,12 @@ \subsection{Device configuration layout}\label{sec:Device Types / Network Device > so without fragmentation, after VIRTIO_NET_F_MTU has been successfully > negotiated. > > +The device MUST set \field{rss_max_key_size} to at least 40, if it offers > +VIRTIO_NET_F_RSS. > + > +The device MUST set \field{rss_max_indirection_table_length} to at least 128, if it offers > +VIRTIO_NET_F_RSS. > + > If the driver negotiates the VIRTIO_NET_F_STANDBY feature, the device MAY act > as a standby device for a primary device with the same MAC address. > > @@ -3675,33 +3720,52 @@ \subsubsection{Control Virtqueue}\label{sec:Device Types / Network Device / Devi > with class VIRTIO_NET_CTRL_ANNOUNCE and command VIRTIO_NET_CTRL_ANNOUNCE_ACK > before marking the buffer as used. > > -\paragraph{Automatic receive steering in multiqueue mode}\label{sec:Device Types / Network Device / Device Operation / Control Virtqueue / Automatic receive steering in multiqueue mode} > +\paragraph{Device operation in multiqueue mode}\label{sec:Device Types / Network Device / Device Operation / Control Virtqueue / Device operation in multiqueue mode} > > -If the driver negotiates the VIRTIO_NET_F_MQ feature bit (depends > -on VIRTIO_NET_F_CTRL_VQ), it MAY transmit outgoing packets on one > -of the multiple transmitq1\ldots transmitqN and ask the device to > -queue incoming packets into one of the multiple receiveq1\ldots receiveqN > -depending on the packet flow. > +This specification defines following modes that a device MAY implement for operation with multiple transmit/receive virtqueues: > +\begin{itemize} > +\item Automatic receive steering as defined in \ref{sec:Device Types / Network Device / Device Operation / Control Virtqueue / Automatic receive steering in multiqueue mode}. > + If a device supports such mode, it offers VIRTIO_NET_F_MQ feature bit. > +\item Receive-side scaling as defined in \ref{devicenormative:Device Types / Network Device / Device Operation / Control Virtqueue / Receive-side scaling (RSS) / RSS processing}. > + If a device supports such mode, it offers VIRTIO_NET_F_RSS feature bit. > +\end{itemize} > > -\begin{lstlisting} > -struct virtio_net_ctrl_mq { > - le16 virtqueue_pairs; > -}; > +A device MAY support one of these features or both. The driver MAY negotiate any set of these features that the device supports. > > +Multiqueue is disabled by default. > + > +The driver enables multiqueue by sending a command using \field{class} VIRTIO_NET_CTRL_MQ. The \field{command} selects the mode of multiqueue operation, as follows: > +\begin{lstlisting} > #define VIRTIO_NET_CTRL_MQ 4 > - #define VIRTIO_NET_CTRL_MQ_VQ_PAIRS_SET 0 > - #define VIRTIO_NET_CTRL_MQ_VQ_PAIRS_MIN 1 > - #define VIRTIO_NET_CTRL_MQ_VQ_PAIRS_MAX 0x8000 > + #define VIRTIO_NET_CTRL_MQ_VQ_PAIRS_SET 0 (for automatic receive steering) > + #define VIRTIO_NET_CTRL_MQ_RSS_CONFIG 1 (for configurable receive steering) > \end{lstlisting} > > -Multiqueue is disabled by default. The driver enables multiqueue by > -executing the VIRTIO_NET_CTRL_MQ_VQ_PAIRS_SET command, specifying > +If more than one multiqueue mode negotiated, the resulting device configuration is defined by the last command sent by the driver. > + > +\paragraph{Automatic receive steering in multiqueue mode}\label{sec:Device Types / Network Device / Device Operation / Control Virtqueue / Automatic receive steering in multiqueue mode} > + > +If the driver negotiates the VIRTIO_NET_F_MQ feature bit (depends on VIRTIO_NET_F_CTRL_VQ), it MAY transmit outgoing packets on one > +of the multiple transmitq1\ldots transmitqN and ask the device to > +queue incoming packets into one of the multiple receiveq1\ldots receiveqN > +depending on the packet flow. > + > +The driver enables multiqueue by > +sending the VIRTIO_NET_CTRL_MQ_VQ_PAIRS_SET command, specifying > the number of the transmit and receive queues to be used up to > \field{max_virtqueue_pairs}; subsequently, > transmitq1\ldots transmitqn and receiveq1\ldots receiveqn where > n=\field{virtqueue_pairs} MAY be used. > +\begin{lstlisting} > +struct virtio_net_ctrl_mq_pairs_set { > + le16 virtqueue_pairs; > +}; > +#define VIRTIO_NET_CTRL_MQ_VQ_PAIRS_MIN 1 > +#define VIRTIO_NET_CTRL_MQ_VQ_PAIRS_MAX 0x8000 > > -When multiqueue is enabled, the device MUST use automatic receive steering > +\end{lstlisting} > + > +When multiqueue is enabled by VIRTIO_NET_CTRL_MQ_VQ_PAIRS_SET command, the device MUST use automatic receive steering > based on packet flow. Programming of the receive steering > classificator is implicit. After the driver transmitted a packet of a > flow on transmitqX, the device SHOULD cause incoming packets for that flow to > @@ -3709,7 +3773,7 @@ \subsubsection{Control Virtqueue}\label{sec:Device Types / Network Device / Devi > no packets have been transmitted yet, the device MAY steer a packet > to a random queue out of the specified receiveq1\ldots receiveqn. > > -Multiqueue is disabled by setting \field{virtqueue_pairs} to 1 (this is > +Multiqueue is disabled by VIRTIO_NET_CTRL_MQ_VQ_PAIRS_SET with \field{virtqueue_pairs} to 1 (this is > the default) and waiting for the device to use the command buffer. > > \drivernormative{\subparagraph}{Automatic receive steering in multiqueue mode}{Device Types / Network Device / Device Operation / Control Virtqueue / Automatic receive steering in multiqueue mode} > @@ -3728,8 +3792,7 @@ \subsubsection{Control Virtqueue}\label{sec:Device Types / Network Device / Devi > > \devicenormative{\subparagraph}{Automatic receive steering in multiqueue mode}{Device Types / Network Device / Device Operation / Control Virtqueue / Automatic receive steering in multiqueue mode} > > -The device MUST queue packets only on any receiveq1 before the > -VIRTIO_NET_CTRL_MQ_VQ_PAIRS_SET command. > +The device after initialization of reset MUST queue packets only on receiveq1. > > The device MUST NOT queue packets on receive queues greater than > \field{virtqueue_pairs} once it has placed the > @@ -3741,6 +3804,131 @@ \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{Receive-side scaling (RSS)}\label{sec:Device Types / Network Device / Device Operation / Control Virtqueue / Receive-side scaling (RSS)} > +A device offers feature VIRTIO_NET_F_RSS if it supports RSS receive steering with Toeplitz hash calculation and configurable parameters. > + > +A driver queries RSS capabilities of the device by reading device configuration as defined in \ref{sec:Device Types / Network Device / Device configuration layout / RSS} > + > +\subparagraph{Setting RSS parameters}\label{sec:Device Types / Network Device / Device Operation / Control Virtqueue / Receive-side scaling (RSS) / Setting RSS parameters} > + > +Driver sends VIRTIO_NET_CTRL_MQ_RSS_CONFIG command using following format for \field{command-specific-data}: > +\begin{lstlisting} > +struct virtio_net_rss_config { > + le32 hash_types; > + le16 indirection_table_mask; > + le16 unclassified_queue; > + le16 indirection_table[indirection_table_length]; > + le16 max_tx_vq; > + u8 hash_key_length; > + u8 hash_key_data[hash_key_length]; > +}; > +\end{lstlisting} > +Field \field{hash_types} contains a bitmask of allowed hash types as > +defined in \ref{sec:Device Types / Network Device / Device Operation / Control Virtqueue / Receive-side scaling (RSS) / RSS hash types}. > + > +Field \field{indirection_table_mask} is a mask to be applied to calculated hash to produce index in \field{indirection_table array}. > +Number of entries in \field{indirection_table} is (\field{indirection_table_mask} + 1). > + > +Field \field{unclassified_queue} contains 0-based index of receive virtqueue to place unclassified packets in. Index 0 corresponds to receiveq1. > + > +Field \field{indirection_table} contains array of 0-based indices of receive virtqueus. Index 0 corresponds to receiveq1. > + > +A driver sets \field{max_tx_vq} to inform a device how many transmit virtqueues it may use (transmitq1\ldots transmitq \field{max_tx_vq}). > + > +\subparagraph{RSS hash types}\label{sec:Device Types / Network Device / Device Operation / Control Virtqueue / Receive-side scaling (RSS) / RSS hash types} > + > +The device calculates hash on IPv4 packets according to the field \field{hash_types} of virtio_net_rss_config structure as follows: > +\begin{itemize} > +\item If VIRTIO_NET_RSS_HASH_TYPE_TCPv4 is set and the packet has TCP header, the hash is calculated over following fields: > +\begin{itemize} > +\item Source IP address > +\item Destination IP address > +\item Source TCP port > +\item Destination TCP port > +\end{itemize} > +\item Else if VIRTIO_NET_RSS_HASH_TYPE_UDPv4 is set and the packet has UDP header, the hash is calculated over following fields: > +\begin{itemize} > +\item Source IP address > +\item Destination IP address > +\item Source UDP port > +\item Destination UDP port > +\end{itemize} > +\item Else if VIRTIO_NET_RSS_HASH_TYPE_IPv4 is set, the hash is calculated over following fields: > +\begin{itemize} > +\item Source IP address > +\item Destination IP address > +\end{itemize} > +\item Else the device does not calculate the hash > +\end{itemize} > + > +The device calculates hash on IPv6 packets without extension headers according to the field \field{hash_types} of virtio_net_rss_config structure as follows: > +\begin{itemize} > +\item If VIRTIO_NET_RSS_HASH_TYPE_TCPv6 is set and the packet has TCPv6 header, the hash is calculated over following fields: > +\begin{itemize} > +\item Source IPv6 address > +\item Destination IPv6 address > +\item Source TCP port > +\item Destination TCP port > +\end{itemize} > +\item Else if VIRTIO_NET_RSS_HASH_TYPE_UDPv6 is set and the packet has UDPv6 header, the hash is calculated over following fields: > +\begin{itemize} > +\item Source IPv6 address > +\item Destination IPv6 address > +\item Source UDP port > +\item Destination UDP port > +\end{itemize} > +\item Else if VIRTIO_NET_RSS_HASH_TYPE_IPv6 is set, the hash is calculated over following fields: > +\begin{itemize} > +\item Source IPv6 address > +\item Destination IPv6 address > +\end{itemize} > +\item Else the device does not calculate the hash > +\end{itemize} > + > +The device calculates hash on IPv6 packets with extension headers according to the field \field{hash_types} of virtio_net_rss_config structure as follows: > +\begin{itemize} > +\item If VIRTIO_NET_RSS_HASH_TYPE_TCP_EX is set and the packet has TCPv6 header, the hash is calculated over following fields: > +\begin{itemize} > +\item Home address from the home address option in the IPv6 destination options header. If the extension header is not present, use the Source IPv6 address. > +\item IPv6 address that is contained in the Routing-Header-Type-2 from the associated extension header. If the extension header is not present, use the Destination IPv6 address. > +\item Source TCP port > +\item Destination TCP port > +\end{itemize} > +\item Else if VIRTIO_NET_RSS_HASH_TYPE_UDP_EX is set and the packet has UDPv6 header, the hash is calculated over following fields: > +\begin{itemize} > +\item Home address from the home address option in the IPv6 destination options header. If the extension header is not present, use the Source IPv6 address. > +\item IPv6 address that is contained in the Routing-Header-Type-2 from the associated extension header. If the extension header is not present, use the Destination IPv6 address. > +\item Source UDP port > +\item Destination UDP port > +\end{itemize} > +\item Else if VIRTIO_NET_RSS_HASH_TYPE_IP_EX is set, the hash is calculated over following fields: > +\begin{itemize} > +\item Home address from the home address option in the IPv6 destination options header. If the extension header is not present, use the Source IPv6 address. > +\item IPv6 address that is contained in the Routing-Header-Type-2 from the associated extension header. If the extension header is not present, use the Destination IPv6 address. > +\end{itemize} > +\item Else skip IPv6 extension headers and calculate the hash as defined above for IPv6 packet without extension headers > +\end{itemize} > + > +\drivernormative{\subparagraph}{Setting RSS parameters}{Device Types / Network Device / Device Operation / Control Virtqueue / Receive-side scaling (RSS) } > + > +A driver MUST NOT send VIRTIO_NET_CTRL_MQ_RSS_CONFIG command if the feature VIRTIO_NET_F_RSS has not been negotiated. > + > +A driver MUST fill \field{indirection_table} array only with indices of enabled queues. Index 0 corresponds to receiveq1. > + > +Number of entries in \field{indirection_table} (\field{indirection_table_mask} + 1) MUST be a power of two. > + > +A driver MUST use \field{indirection_table_mask} values that are less than \field{rss_max_indirection_table_length} reported by a device. > + > +A driver MUST NOT set any VIRTIO_NET_RSS_HASH_TYPE_ flags that are not supported by a device. > + > +\devicenormative{\subparagraph}{RSS processing}{Device Types / Network Device / Device Operation / Control Virtqueue / Receive-side scaling (RSS) / RSS processing} > +The device MUST determine destination queue for network packet as follows: > +\begin{itemize} > +\item Calculate hash of the packet as defined in \ref{sec:Device Types / Network Device / Device Operation / Control Virtqueue / Receive-side scaling (RSS) / RSS hash types} > +\item If the device did not calculate the hash for specific packet, the device directs the packet to the receiveq specified by \field{unclassified_queue} of virtio_net_rss_config structure (value of 0 corresponds to receiveq1). > +\item Apply \field{indirection_table_mask} to the calculated hash and use the result as the index in the indirection table to get 0-based number of destination receiveq (value of 0 corresponds to receiveq1). > +\end{itemize} > + > \paragraph{Offloads State Configuration}\label{sec:Device Types / Network Device / Device Operation / Control Virtqueue / Offloads State Configuration} > > If the VIRTIO_NET_F_CTRL_GUEST_OFFLOADS feature is negotiated, the driver can > -- > 2.17.2
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]