OK my bad, missed the update to the working copy of the spec. Rob Miller rob.miller@broadcom.com (919)721-3339 On Mon, Mar 2, 2020 at 8:04 AM Yuri Benditovich wrote: > On Mon, Mar 2, 2020 at 2:17 PM Rob Miller wrote: > > > > I think it would be good to have a means to negotiate which types of RSS > that a device is willing to do. Perhaps a CTRLQ command to get RSS > capabilities then only set those. For example, if a device wishes not to > RSS on UDPx but will RSS on TCPx, and a HASH request on TCP|USP, this would > result in a failure. > > > > Also, what is the definitions of the xxx_EX options? > > > > The device indicates its capabilities in its configuration space. > The exact details are in the latest master branch of the virtio-spec > (the discussion was in virtio-comment list). > These commits are to align Linux kernel header to what the spec defines. > > > Rob Miller > > rob.miller@broadcom.com > > (919)721-3339 > > > > > > On Sun, Mar 1, 2020 at 9:33 AM Yuri Benditovich < > yuri.benditovich@daynix.com> wrote: > >> > >> RSS (Receive-side scaling) defines hash calculation > >> rules and decision on receive virtqueue according to > >> the calculated hash, provided mask to apply and > >> provided indirection table containing indices of > >> receive virqueues. The driver sends the control > >> command to enable multiqueue and provide parameters > >> for receive steering. > >> > >> Signed-off-by: Yuri Benditovich > >> --- > >> include/uapi/linux/virtio_net.h | 42 +++++++++++++++++++++++++++++++-- > >> 1 file changed, 40 insertions(+), 2 deletions(-) > >> > >> diff --git a/include/uapi/linux/virtio_net.h > b/include/uapi/linux/virtio_net.h > >> index 19e76b3e3a64..188ad3eecdc8 100644 > >> --- a/include/uapi/linux/virtio_net.h > >> +++ b/include/uapi/linux/virtio_net.h > >> @@ -57,6 +57,7 @@ > >> * Steering */ > >> #define VIRTIO_NET_F_CTRL_MAC_ADDR 23 /* Set MAC address */ > >> > >> +#define VIRTIO_NET_F_RSS 60 /* Supports RSS RX steering */ > >> #define VIRTIO_NET_F_RSC_EXT 61 /* extended coalescing info */ > >> #define VIRTIO_NET_F_STANDBY 62 /* Act as standby for another > device > >> * with the same MAC. > >> @@ -70,6 +71,17 @@ > >> #define VIRTIO_NET_S_LINK_UP 1 /* Link is up */ > >> #define VIRTIO_NET_S_ANNOUNCE 2 /* Announcement is needed */ > >> > >> +/* supported/enabled hash types */ > >> +#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) > >> +#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) > >> +#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) > >> + > >> struct virtio_net_config { > >> /* The config defining mac address (if VIRTIO_NET_F_MAC) */ > >> __u8 mac[ETH_ALEN]; > >> @@ -93,6 +105,12 @@ struct virtio_net_config { > >> * Any other value stands for unknown. > >> */ > >> __u8 duplex; > >> + /* maximum size of RSS key */ > >> + __u8 rss_max_key_size; > >> + /* maximum number of indirection table entries */ > >> + __le16 rss_max_indirection_table_length; > >> + /* bitmask of supported VIRTIO_NET_RSS_HASH_ types */ > >> + __le32 supported_hash_types; > >> } __attribute__((packed)); > >> > >> /* > >> @@ -246,7 +264,9 @@ struct virtio_net_ctrl_mac { > >> > >> /* > >> * Control Receive Flow Steering > >> - * > >> + */ > >> +#define VIRTIO_NET_CTRL_MQ 4 > >> +/* > >> * The command VIRTIO_NET_CTRL_MQ_VQ_PAIRS_SET > >> * enables Receive Flow Steering, specifying the number of the > transmit and > >> * receive queues that will be used. After the command is consumed and > acked by > >> @@ -259,11 +279,29 @@ struct virtio_net_ctrl_mq { > >> __virtio16 virtqueue_pairs; > >> }; > >> > >> -#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 > >> > >> +/* > >> + * The command VIRTIO_NET_CTRL_MQ_RSS_CONFIG has the same effect as > >> + * VIRTIO_NET_CTRL_MQ_VQ_PAIRS_SET does and additionally configures > >> + * the receive steering to use a hash calculated for incoming packet > >> + * to decide on receive virtqueue to place the packet. The command > >> + * also provides parameters to calculate a hash and receive virtqueue. > >> + */ > >> +struct virtio_net_rss_config { > >> + __le32 hash_types; > >> + __le16 indirection_table_mask; > >> + __le16 unclassified_queue; > >> + __le16 indirection_table[1/* + indirection_table_mask */]; > >> + __le16 max_tx_vq; > >> + __u8 hash_key_length; > >> + __u8 hash_key_data[/* hash_key_length */]; > >> +}; > >> + > >> + #define VIRTIO_NET_CTRL_MQ_RSS_CONFIG 1 > >> + > >> /* > >> * Control network offloads > >> * > >> -- > >> 2.17.1 > >> > >> > >> --------------------------------------------------------------------- > >> To unsubscribe, e-mail: virtio-dev-unsubscribe@lists.oasis-open.org > >> For additional commands, e-mail: virtio-dev-help@lists.oasis-open.org > >> >