| Message ID | 20240416061943.407082-1-liangchen.linux@gmail.com (mailing list archive) |
|---|---|
| State | Superseded |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [net-next,v8] virtio_net: Support RX hash XDP hint | expand |
On Tue, Apr 16, 2024 at 2:20 PM Liang Chen <liangchen.linux@gmail.com> wrote: > > The RSS hash report is a feature that's part of the virtio specification. > Currently, virtio backends like qemu, vdpa (mlx5), and potentially vhost > (still a work in progress as per [1]) support this feature. While the > capability to obtain the RSS hash has been enabled in the normal path, > it's currently missing in the XDP path. Therefore, we are introducing > XDP hints through kfuncs to allow XDP programs to access the RSS hash. > > 1. > https://lore.kernel.org/all/20231015141644.260646-1-akihiko.odaki@daynix.com/#r > > Signed-off-by: Liang Chen <liangchen.linux@gmail.com> > --- > Changes from v7: > - use table lookup for rss hash type > Changes from v6: > - fix a coding style issue > Changes from v5: > - Preservation of the hash value has been dropped, following the conclusion > from discussions in V3 reviews. The virtio_net driver doesn't > accessing/using the virtio_net_hdr after the XDP program execution, so > nothing tragic should happen. As to the xdp program, if it smashes the > entry in virtio header, it is likely buggy anyways. Additionally, looking > up the Intel IGC driver, it also does not bother with this particular > aspect. > --- > drivers/net/virtio_net.c | 42 +++++++++++++++++++++++++++++++++ > include/uapi/linux/virtio_net.h | 1 + > 2 files changed, 43 insertions(+) > > diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c > index c22d1118a133..1d750009f615 100644 > --- a/drivers/net/virtio_net.c > +++ b/drivers/net/virtio_net.c > @@ -4621,6 +4621,47 @@ static void virtnet_set_big_packets(struct virtnet_info *vi, const int mtu) > } > } > > +static enum xdp_rss_hash_type > +virtnet_xdp_rss_type[VIRTIO_NET_HASH_REPORT_MAX_TABLE] = { > + [VIRTIO_NET_HASH_REPORT_NONE] = XDP_RSS_TYPE_NONE, > + [VIRTIO_NET_HASH_REPORT_IPv4] = XDP_RSS_TYPE_L3_IPV4, > + [VIRTIO_NET_HASH_REPORT_TCPv4] = XDP_RSS_TYPE_L4_IPV4_TCP, > + [VIRTIO_NET_HASH_REPORT_UDPv4] = XDP_RSS_TYPE_L4_IPV4_UDP, > + [VIRTIO_NET_HASH_REPORT_IPv6] = XDP_RSS_TYPE_L3_IPV6, > + [VIRTIO_NET_HASH_REPORT_TCPv6] = XDP_RSS_TYPE_L4_IPV6_TCP, > + [VIRTIO_NET_HASH_REPORT_UDPv6] = XDP_RSS_TYPE_L4_IPV6_UDP, > + [VIRTIO_NET_HASH_REPORT_IPv6_EX] = XDP_RSS_TYPE_L3_IPV6_EX, > + [VIRTIO_NET_HASH_REPORT_TCPv6_EX] = XDP_RSS_TYPE_L4_IPV6_TCP_EX, > + [VIRTIO_NET_HASH_REPORT_UDPv6_EX] = XDP_RSS_TYPE_L4_IPV6_UDP_EX > +}; > + > +static int virtnet_xdp_rx_hash(const struct xdp_md *_ctx, u32 *hash, > + enum xdp_rss_hash_type *rss_type) > +{ > + const struct xdp_buff *xdp = (void *)_ctx; > + struct virtio_net_hdr_v1_hash *hdr_hash; > + struct virtnet_info *vi; > + u16 hash_report; > + > + if (!(xdp->rxq->dev->features & NETIF_F_RXHASH)) > + return -ENODATA; > + > + vi = netdev_priv(xdp->rxq->dev); > + hdr_hash = (struct virtio_net_hdr_v1_hash *)(xdp->data - vi->hdr_len); > + hash_report = __le16_to_cpu(hdr_hash->hash_report); > + > + if (hash_report >= VIRTIO_NET_HASH_REPORT_MAX_TABLE) > + hash_report = VIRTIO_NET_HASH_REPORT_NONE; > + > + *rss_type = virtnet_xdp_rss_type[hash_report]; > + *hash = __le32_to_cpu(hdr_hash->hash_value); > + return 0; > +} > + > +static const struct xdp_metadata_ops virtnet_xdp_metadata_ops = { > + .xmo_rx_hash = virtnet_xdp_rx_hash, > +}; > + > static int virtnet_probe(struct virtio_device *vdev) > { > int i, err = -ENOMEM; > @@ -4747,6 +4788,7 @@ static int virtnet_probe(struct virtio_device *vdev) > VIRTIO_NET_RSS_HASH_TYPE_UDP_EX); > > dev->hw_features |= NETIF_F_RXHASH; > + dev->xdp_metadata_ops = &virtnet_xdp_metadata_ops; > } > > if (vi->has_rss_hash_report) > diff --git a/include/uapi/linux/virtio_net.h b/include/uapi/linux/virtio_net.h > index cc65ef0f3c3e..3ee695450096 100644 > --- a/include/uapi/linux/virtio_net.h > +++ b/include/uapi/linux/virtio_net.h > @@ -176,6 +176,7 @@ struct virtio_net_hdr_v1_hash { > #define VIRTIO_NET_HASH_REPORT_IPv6_EX 7 > #define VIRTIO_NET_HASH_REPORT_TCPv6_EX 8 > #define VIRTIO_NET_HASH_REPORT_UDPv6_EX 9 > +#define VIRTIO_NET_HASH_REPORT_MAX_TABLE 10 This should not be part of uAPI. It may confuse the userspace. Others look good. Thanks > __le16 hash_report; > __le16 padding; > }; > -- > 2.40.1 >
在 2024/4/16 下午2:19, Liang Chen 写道: > The RSS hash report is a feature that's part of the virtio specification. > Currently, virtio backends like qemu, vdpa (mlx5), and potentially vhost > (still a work in progress as per [1]) support this feature. While the > capability to obtain the RSS hash has been enabled in the normal path, > it's currently missing in the XDP path. Therefore, we are introducing > XDP hints through kfuncs to allow XDP programs to access the RSS hash. > > 1. > https://lore.kernel.org/all/20231015141644.260646-1-akihiko.odaki@daynix.com/#r > > Signed-off-by: Liang Chen <liangchen.linux@gmail.com> > --- > Changes from v7: > - use table lookup for rss hash type > Changes from v6: > - fix a coding style issue > Changes from v5: > - Preservation of the hash value has been dropped, following the conclusion > from discussions in V3 reviews. The virtio_net driver doesn't > accessing/using the virtio_net_hdr after the XDP program execution, so > nothing tragic should happen. As to the xdp program, if it smashes the > entry in virtio header, it is likely buggy anyways. Additionally, looking > up the Intel IGC driver, it also does not bother with this particular > aspect. > --- > drivers/net/virtio_net.c | 42 +++++++++++++++++++++++++++++++++ > include/uapi/linux/virtio_net.h | 1 + > 2 files changed, 43 insertions(+) > > diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c > index c22d1118a133..1d750009f615 100644 > --- a/drivers/net/virtio_net.c > +++ b/drivers/net/virtio_net.c > @@ -4621,6 +4621,47 @@ static void virtnet_set_big_packets(struct virtnet_info *vi, const int mtu) > } > } > > +static enum xdp_rss_hash_type > +virtnet_xdp_rss_type[VIRTIO_NET_HASH_REPORT_MAX_TABLE] = { > + [VIRTIO_NET_HASH_REPORT_NONE] = XDP_RSS_TYPE_NONE, > + [VIRTIO_NET_HASH_REPORT_IPv4] = XDP_RSS_TYPE_L3_IPV4, > + [VIRTIO_NET_HASH_REPORT_TCPv4] = XDP_RSS_TYPE_L4_IPV4_TCP, > + [VIRTIO_NET_HASH_REPORT_UDPv4] = XDP_RSS_TYPE_L4_IPV4_UDP, > + [VIRTIO_NET_HASH_REPORT_IPv6] = XDP_RSS_TYPE_L3_IPV6, > + [VIRTIO_NET_HASH_REPORT_TCPv6] = XDP_RSS_TYPE_L4_IPV6_TCP, > + [VIRTIO_NET_HASH_REPORT_UDPv6] = XDP_RSS_TYPE_L4_IPV6_UDP, > + [VIRTIO_NET_HASH_REPORT_IPv6_EX] = XDP_RSS_TYPE_L3_IPV6_EX, > + [VIRTIO_NET_HASH_REPORT_TCPv6_EX] = XDP_RSS_TYPE_L4_IPV6_TCP_EX, > + [VIRTIO_NET_HASH_REPORT_UDPv6_EX] = XDP_RSS_TYPE_L4_IPV6_UDP_EX > +}; > + > +static int virtnet_xdp_rx_hash(const struct xdp_md *_ctx, u32 *hash, > + enum xdp_rss_hash_type *rss_type) > +{ > + const struct xdp_buff *xdp = (void *)_ctx; > + struct virtio_net_hdr_v1_hash *hdr_hash; > + struct virtnet_info *vi; > + u16 hash_report; > + > + if (!(xdp->rxq->dev->features & NETIF_F_RXHASH)) > + return -ENODATA; > + > + vi = netdev_priv(xdp->rxq->dev); > + hdr_hash = (struct virtio_net_hdr_v1_hash *)(xdp->data - vi->hdr_len); > + hash_report = __le16_to_cpu(hdr_hash->hash_report); > + > + if (hash_report >= VIRTIO_NET_HASH_REPORT_MAX_TABLE) > + hash_report = VIRTIO_NET_HASH_REPORT_NONE; When this happens, it may mean an error or user modification of the header occurred. Should the following *hash* value be cleared to 0? Thanks, Heng > + > + *rss_type = virtnet_xdp_rss_type[hash_report]; > + *hash = __le32_to_cpu(hdr_hash->hash_value); > + return 0; > +} > + > +static const struct xdp_metadata_ops virtnet_xdp_metadata_ops = { > + .xmo_rx_hash = virtnet_xdp_rx_hash, > +}; > + > static int virtnet_probe(struct virtio_device *vdev) > { > int i, err = -ENOMEM; > @@ -4747,6 +4788,7 @@ static int virtnet_probe(struct virtio_device *vdev) > VIRTIO_NET_RSS_HASH_TYPE_UDP_EX); > > dev->hw_features |= NETIF_F_RXHASH; > + dev->xdp_metadata_ops = &virtnet_xdp_metadata_ops; > } > > if (vi->has_rss_hash_report) > diff --git a/include/uapi/linux/virtio_net.h b/include/uapi/linux/virtio_net.h > index cc65ef0f3c3e..3ee695450096 100644 > --- a/include/uapi/linux/virtio_net.h > +++ b/include/uapi/linux/virtio_net.h > @@ -176,6 +176,7 @@ struct virtio_net_hdr_v1_hash { > #define VIRTIO_NET_HASH_REPORT_IPv6_EX 7 > #define VIRTIO_NET_HASH_REPORT_TCPv6_EX 8 > #define VIRTIO_NET_HASH_REPORT_UDPv6_EX 9 > +#define VIRTIO_NET_HASH_REPORT_MAX_TABLE 10 > __le16 hash_report; > __le16 padding; > };
On Tue, Apr 16, 2024 at 3:20 PM Jason Wang <jasowang@redhat.com> wrote: > > On Tue, Apr 16, 2024 at 2:20 PM Liang Chen <liangchen.linux@gmail.com> wrote: > > > > The RSS hash report is a feature that's part of the virtio specification. > > Currently, virtio backends like qemu, vdpa (mlx5), and potentially vhost > > (still a work in progress as per [1]) support this feature. While the > > capability to obtain the RSS hash has been enabled in the normal path, > > it's currently missing in the XDP path. Therefore, we are introducing > > XDP hints through kfuncs to allow XDP programs to access the RSS hash. > > > > 1. > > https://lore.kernel.org/all/20231015141644.260646-1-akihiko.odaki@daynix.com/#r > > > > Signed-off-by: Liang Chen <liangchen.linux@gmail.com> > > --- > > Changes from v7: > > - use table lookup for rss hash type > > Changes from v6: > > - fix a coding style issue > > Changes from v5: > > - Preservation of the hash value has been dropped, following the conclusion > > from discussions in V3 reviews. The virtio_net driver doesn't > > accessing/using the virtio_net_hdr after the XDP program execution, so > > nothing tragic should happen. As to the xdp program, if it smashes the > > entry in virtio header, it is likely buggy anyways. Additionally, looking > > up the Intel IGC driver, it also does not bother with this particular > > aspect. > > --- > > drivers/net/virtio_net.c | 42 +++++++++++++++++++++++++++++++++ > > include/uapi/linux/virtio_net.h | 1 + > > 2 files changed, 43 insertions(+) > > > > diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c > > index c22d1118a133..1d750009f615 100644 > > --- a/drivers/net/virtio_net.c > > +++ b/drivers/net/virtio_net.c > > @@ -4621,6 +4621,47 @@ static void virtnet_set_big_packets(struct virtnet_info *vi, const int mtu) > > } > > } > > > > +static enum xdp_rss_hash_type > > +virtnet_xdp_rss_type[VIRTIO_NET_HASH_REPORT_MAX_TABLE] = { > > + [VIRTIO_NET_HASH_REPORT_NONE] = XDP_RSS_TYPE_NONE, > > + [VIRTIO_NET_HASH_REPORT_IPv4] = XDP_RSS_TYPE_L3_IPV4, > > + [VIRTIO_NET_HASH_REPORT_TCPv4] = XDP_RSS_TYPE_L4_IPV4_TCP, > > + [VIRTIO_NET_HASH_REPORT_UDPv4] = XDP_RSS_TYPE_L4_IPV4_UDP, > > + [VIRTIO_NET_HASH_REPORT_IPv6] = XDP_RSS_TYPE_L3_IPV6, > > + [VIRTIO_NET_HASH_REPORT_TCPv6] = XDP_RSS_TYPE_L4_IPV6_TCP, > > + [VIRTIO_NET_HASH_REPORT_UDPv6] = XDP_RSS_TYPE_L4_IPV6_UDP, > > + [VIRTIO_NET_HASH_REPORT_IPv6_EX] = XDP_RSS_TYPE_L3_IPV6_EX, > > + [VIRTIO_NET_HASH_REPORT_TCPv6_EX] = XDP_RSS_TYPE_L4_IPV6_TCP_EX, > > + [VIRTIO_NET_HASH_REPORT_UDPv6_EX] = XDP_RSS_TYPE_L4_IPV6_UDP_EX > > +}; > > + > > +static int virtnet_xdp_rx_hash(const struct xdp_md *_ctx, u32 *hash, > > + enum xdp_rss_hash_type *rss_type) > > +{ > > + const struct xdp_buff *xdp = (void *)_ctx; > > + struct virtio_net_hdr_v1_hash *hdr_hash; > > + struct virtnet_info *vi; > > + u16 hash_report; > > + > > + if (!(xdp->rxq->dev->features & NETIF_F_RXHASH)) > > + return -ENODATA; > > + > > + vi = netdev_priv(xdp->rxq->dev); > > + hdr_hash = (struct virtio_net_hdr_v1_hash *)(xdp->data - vi->hdr_len); > > + hash_report = __le16_to_cpu(hdr_hash->hash_report); > > + > > + if (hash_report >= VIRTIO_NET_HASH_REPORT_MAX_TABLE) > > + hash_report = VIRTIO_NET_HASH_REPORT_NONE; > > + > > + *rss_type = virtnet_xdp_rss_type[hash_report]; > > + *hash = __le32_to_cpu(hdr_hash->hash_value); > > + return 0; > > +} > > + > > +static const struct xdp_metadata_ops virtnet_xdp_metadata_ops = { > > + .xmo_rx_hash = virtnet_xdp_rx_hash, > > +}; > > + > > static int virtnet_probe(struct virtio_device *vdev) > > { > > int i, err = -ENOMEM; > > @@ -4747,6 +4788,7 @@ static int virtnet_probe(struct virtio_device *vdev) > > VIRTIO_NET_RSS_HASH_TYPE_UDP_EX); > > > > dev->hw_features |= NETIF_F_RXHASH; > > + dev->xdp_metadata_ops = &virtnet_xdp_metadata_ops; > > } > > > > if (vi->has_rss_hash_report) > > diff --git a/include/uapi/linux/virtio_net.h b/include/uapi/linux/virtio_net.h > > index cc65ef0f3c3e..3ee695450096 100644 > > --- a/include/uapi/linux/virtio_net.h > > +++ b/include/uapi/linux/virtio_net.h > > @@ -176,6 +176,7 @@ struct virtio_net_hdr_v1_hash { > > #define VIRTIO_NET_HASH_REPORT_IPv6_EX 7 > > #define VIRTIO_NET_HASH_REPORT_TCPv6_EX 8 > > #define VIRTIO_NET_HASH_REPORT_UDPv6_EX 9 > > +#define VIRTIO_NET_HASH_REPORT_MAX_TABLE 10 > > This should not be part of uAPI. It may confuse the userspace. > Sure. I will just move it to virtio_net.c right above the table definition. Thanks! > Others look good. > > Thanks > > > __le16 hash_report; > > __le16 padding; > > }; > > -- > > 2.40.1 > > >
On Tue, Apr 16, 2024 at 3:35 PM Heng Qi <hengqi@linux.alibaba.com> wrote: > > > > 在 2024/4/16 下午2:19, Liang Chen 写道: > > The RSS hash report is a feature that's part of the virtio specification. > > Currently, virtio backends like qemu, vdpa (mlx5), and potentially vhost > > (still a work in progress as per [1]) support this feature. While the > > capability to obtain the RSS hash has been enabled in the normal path, > > it's currently missing in the XDP path. Therefore, we are introducing > > XDP hints through kfuncs to allow XDP programs to access the RSS hash. > > > > 1. > > https://lore.kernel.org/all/20231015141644.260646-1-akihiko.odaki@daynix.com/#r > > > > Signed-off-by: Liang Chen <liangchen.linux@gmail.com> > > --- > > Changes from v7: > > - use table lookup for rss hash type > > Changes from v6: > > - fix a coding style issue > > Changes from v5: > > - Preservation of the hash value has been dropped, following the conclusion > > from discussions in V3 reviews. The virtio_net driver doesn't > > accessing/using the virtio_net_hdr after the XDP program execution, so > > nothing tragic should happen. As to the xdp program, if it smashes the > > entry in virtio header, it is likely buggy anyways. Additionally, looking > > up the Intel IGC driver, it also does not bother with this particular > > aspect. > > --- > > drivers/net/virtio_net.c | 42 +++++++++++++++++++++++++++++++++ > > include/uapi/linux/virtio_net.h | 1 + > > 2 files changed, 43 insertions(+) > > > > diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c > > index c22d1118a133..1d750009f615 100644 > > --- a/drivers/net/virtio_net.c > > +++ b/drivers/net/virtio_net.c > > @@ -4621,6 +4621,47 @@ static void virtnet_set_big_packets(struct virtnet_info *vi, const int mtu) > > } > > } > > > > +static enum xdp_rss_hash_type > > +virtnet_xdp_rss_type[VIRTIO_NET_HASH_REPORT_MAX_TABLE] = { > > + [VIRTIO_NET_HASH_REPORT_NONE] = XDP_RSS_TYPE_NONE, > > + [VIRTIO_NET_HASH_REPORT_IPv4] = XDP_RSS_TYPE_L3_IPV4, > > + [VIRTIO_NET_HASH_REPORT_TCPv4] = XDP_RSS_TYPE_L4_IPV4_TCP, > > + [VIRTIO_NET_HASH_REPORT_UDPv4] = XDP_RSS_TYPE_L4_IPV4_UDP, > > + [VIRTIO_NET_HASH_REPORT_IPv6] = XDP_RSS_TYPE_L3_IPV6, > > + [VIRTIO_NET_HASH_REPORT_TCPv6] = XDP_RSS_TYPE_L4_IPV6_TCP, > > + [VIRTIO_NET_HASH_REPORT_UDPv6] = XDP_RSS_TYPE_L4_IPV6_UDP, > > + [VIRTIO_NET_HASH_REPORT_IPv6_EX] = XDP_RSS_TYPE_L3_IPV6_EX, > > + [VIRTIO_NET_HASH_REPORT_TCPv6_EX] = XDP_RSS_TYPE_L4_IPV6_TCP_EX, > > + [VIRTIO_NET_HASH_REPORT_UDPv6_EX] = XDP_RSS_TYPE_L4_IPV6_UDP_EX > > +}; > > + > > +static int virtnet_xdp_rx_hash(const struct xdp_md *_ctx, u32 *hash, > > + enum xdp_rss_hash_type *rss_type) > > +{ > > + const struct xdp_buff *xdp = (void *)_ctx; > > + struct virtio_net_hdr_v1_hash *hdr_hash; > > + struct virtnet_info *vi; > > + u16 hash_report; > > + > > + if (!(xdp->rxq->dev->features & NETIF_F_RXHASH)) > > + return -ENODATA; > > + > > + vi = netdev_priv(xdp->rxq->dev); > > + hdr_hash = (struct virtio_net_hdr_v1_hash *)(xdp->data - vi->hdr_len); > > + hash_report = __le16_to_cpu(hdr_hash->hash_report); > > + > > + if (hash_report >= VIRTIO_NET_HASH_REPORT_MAX_TABLE) > > + hash_report = VIRTIO_NET_HASH_REPORT_NONE; > > When this happens, it may mean an error or user modification of the > header occurred. > Should the following *hash* value be cleared to 0? > Referring to the igc and mlx5 drivers, the hash value is not cleared to 0 in such cases either. This is because if the rss type is XDP_RSS_TYPE_NONE, the hash value is considered to be not relevant. Thanks! > Thanks, > Heng > > > + > > + *rss_type = virtnet_xdp_rss_type[hash_report]; > > + *hash = __le32_to_cpu(hdr_hash->hash_value); > > + return 0; > > +} > > + > > +static const struct xdp_metadata_ops virtnet_xdp_metadata_ops = { > > + .xmo_rx_hash = virtnet_xdp_rx_hash, > > +}; > > + > > static int virtnet_probe(struct virtio_device *vdev) > > { > > int i, err = -ENOMEM; > > @@ -4747,6 +4788,7 @@ static int virtnet_probe(struct virtio_device *vdev) > > VIRTIO_NET_RSS_HASH_TYPE_UDP_EX); > > > > dev->hw_features |= NETIF_F_RXHASH; > > + dev->xdp_metadata_ops = &virtnet_xdp_metadata_ops; > > } > > > > if (vi->has_rss_hash_report) > > diff --git a/include/uapi/linux/virtio_net.h b/include/uapi/linux/virtio_net.h > > index cc65ef0f3c3e..3ee695450096 100644 > > --- a/include/uapi/linux/virtio_net.h > > +++ b/include/uapi/linux/virtio_net.h > > @@ -176,6 +176,7 @@ struct virtio_net_hdr_v1_hash { > > #define VIRTIO_NET_HASH_REPORT_IPv6_EX 7 > > #define VIRTIO_NET_HASH_REPORT_TCPv6_EX 8 > > #define VIRTIO_NET_HASH_REPORT_UDPv6_EX 9 > > +#define VIRTIO_NET_HASH_REPORT_MAX_TABLE 10 > > __le16 hash_report; > > __le16 padding; > > }; >
diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c index c22d1118a133..1d750009f615 100644 --- a/drivers/net/virtio_net.c +++ b/drivers/net/virtio_net.c @@ -4621,6 +4621,47 @@ static void virtnet_set_big_packets(struct virtnet_info *vi, const int mtu) } } +static enum xdp_rss_hash_type +virtnet_xdp_rss_type[VIRTIO_NET_HASH_REPORT_MAX_TABLE] = { + [VIRTIO_NET_HASH_REPORT_NONE] = XDP_RSS_TYPE_NONE, + [VIRTIO_NET_HASH_REPORT_IPv4] = XDP_RSS_TYPE_L3_IPV4, + [VIRTIO_NET_HASH_REPORT_TCPv4] = XDP_RSS_TYPE_L4_IPV4_TCP, + [VIRTIO_NET_HASH_REPORT_UDPv4] = XDP_RSS_TYPE_L4_IPV4_UDP, + [VIRTIO_NET_HASH_REPORT_IPv6] = XDP_RSS_TYPE_L3_IPV6, + [VIRTIO_NET_HASH_REPORT_TCPv6] = XDP_RSS_TYPE_L4_IPV6_TCP, + [VIRTIO_NET_HASH_REPORT_UDPv6] = XDP_RSS_TYPE_L4_IPV6_UDP, + [VIRTIO_NET_HASH_REPORT_IPv6_EX] = XDP_RSS_TYPE_L3_IPV6_EX, + [VIRTIO_NET_HASH_REPORT_TCPv6_EX] = XDP_RSS_TYPE_L4_IPV6_TCP_EX, + [VIRTIO_NET_HASH_REPORT_UDPv6_EX] = XDP_RSS_TYPE_L4_IPV6_UDP_EX +}; + +static int virtnet_xdp_rx_hash(const struct xdp_md *_ctx, u32 *hash, + enum xdp_rss_hash_type *rss_type) +{ + const struct xdp_buff *xdp = (void *)_ctx; + struct virtio_net_hdr_v1_hash *hdr_hash; + struct virtnet_info *vi; + u16 hash_report; + + if (!(xdp->rxq->dev->features & NETIF_F_RXHASH)) + return -ENODATA; + + vi = netdev_priv(xdp->rxq->dev); + hdr_hash = (struct virtio_net_hdr_v1_hash *)(xdp->data - vi->hdr_len); + hash_report = __le16_to_cpu(hdr_hash->hash_report); + + if (hash_report >= VIRTIO_NET_HASH_REPORT_MAX_TABLE) + hash_report = VIRTIO_NET_HASH_REPORT_NONE; + + *rss_type = virtnet_xdp_rss_type[hash_report]; + *hash = __le32_to_cpu(hdr_hash->hash_value); + return 0; +} + +static const struct xdp_metadata_ops virtnet_xdp_metadata_ops = { + .xmo_rx_hash = virtnet_xdp_rx_hash, +}; + static int virtnet_probe(struct virtio_device *vdev) { int i, err = -ENOMEM; @@ -4747,6 +4788,7 @@ static int virtnet_probe(struct virtio_device *vdev) VIRTIO_NET_RSS_HASH_TYPE_UDP_EX); dev->hw_features |= NETIF_F_RXHASH; + dev->xdp_metadata_ops = &virtnet_xdp_metadata_ops; } if (vi->has_rss_hash_report) diff --git a/include/uapi/linux/virtio_net.h b/include/uapi/linux/virtio_net.h index cc65ef0f3c3e..3ee695450096 100644 --- a/include/uapi/linux/virtio_net.h +++ b/include/uapi/linux/virtio_net.h @@ -176,6 +176,7 @@ struct virtio_net_hdr_v1_hash { #define VIRTIO_NET_HASH_REPORT_IPv6_EX 7 #define VIRTIO_NET_HASH_REPORT_TCPv6_EX 8 #define VIRTIO_NET_HASH_REPORT_UDPv6_EX 9 +#define VIRTIO_NET_HASH_REPORT_MAX_TABLE 10 __le16 hash_report; __le16 padding; };
The RSS hash report is a feature that's part of the virtio specification. Currently, virtio backends like qemu, vdpa (mlx5), and potentially vhost (still a work in progress as per [1]) support this feature. While the capability to obtain the RSS hash has been enabled in the normal path, it's currently missing in the XDP path. Therefore, we are introducing XDP hints through kfuncs to allow XDP programs to access the RSS hash. 1. https://lore.kernel.org/all/20231015141644.260646-1-akihiko.odaki@daynix.com/#r Signed-off-by: Liang Chen <liangchen.linux@gmail.com> --- Changes from v7: - use table lookup for rss hash type Changes from v6: - fix a coding style issue Changes from v5: - Preservation of the hash value has been dropped, following the conclusion from discussions in V3 reviews. The virtio_net driver doesn't accessing/using the virtio_net_hdr after the XDP program execution, so nothing tragic should happen. As to the xdp program, if it smashes the entry in virtio header, it is likely buggy anyways. Additionally, looking up the Intel IGC driver, it also does not bother with this particular aspect. --- drivers/net/virtio_net.c | 42 +++++++++++++++++++++++++++++++++ include/uapi/linux/virtio_net.h | 1 + 2 files changed, 43 insertions(+)