similar to: [PATCH vhost v1 00/12] virtio core prepares for AF_XDP

On Thu, Mar 2, 2023 at 7:59?PM Xuan Zhuo <xuanzhuo at linux.alibaba.com> wrote: > > virtqueue_add_split() only supports virtual addresses, dma is completed > in virtqueue_add_split(). > > In some scenarios (such as the AF_XDP scenario), the memory is allocated > and DMA is completed in advance, so it is necessary for us to support > passing the DMA address to virtio

On Tue, 7 Mar 2023 14:43:42 +0800, Jason Wang <jasowang at redhat.com> wrote: > On Thu, Mar 2, 2023 at 7:59?PM Xuan Zhuo <xuanzhuo at linux.alibaba.com> wrote: > > > > virtqueue_add_split() only supports virtual addresses, dma is completed > > in virtqueue_add_split(). > > > > In some scenarios (such as the AF_XDP scenario), the memory is allocated >

This commit splits virtqueue_add_split() to two functions. The purpose of such splitting is to separate DMA operations. The first function includes all codes that may fail before the DMA operation. The subsequent part is used as the second function. In this way, we can perform DMA operations in the middle of the two functions. If the first function fails, we do not need to perform DMA

DMA-related logic is separated from the virtqueue_add_vring_split() to prepare for subsequent support for premapped. DMA address will be saved as sg->dma_address, then virtqueue_add_vring_split() will use it directly. If it is a premapped scene, the transmitted sgs should have saved DMA address in dma_address, and in virtio core, we need to pass virtqueue_map_sgs(). Signed-off-by: Xuan Zhuo

XDP socket(AF_XDP) is an excellent bypass kernel network framework. The zero copy feature of xsk (XDP socket) needs to be supported by the driver. The performance of zero copy is very good. ENV: Qemu with vhost. vhost cpu | Guest APP CPU |Guest Softirq CPU | PPS -----------------------------|---------------|------------------|------------ xmit by sockperf: 90% | 100%

XDP socket(AF_XDP) is an excellent bypass kernel network framework. The zero copy feature of xsk (XDP socket) needs to be supported by the driver. The performance of zero copy is very good. ENV: Qemu with vhost. vhost cpu | Guest APP CPU |Guest Softirq CPU | PPS -----------------------------|---------------|------------------|------------ xmit by sockperf: 90% | 100%

XDP socket(AF_XDP) is an excellent bypass kernel network framework. The zero copy feature of xsk (XDP socket) needs to be supported by the driver. The performance of zero copy is very good. ENV: Qemu with vhost. vhost cpu | Guest APP CPU |Guest Softirq CPU | PPS -----------------------------|---------------|------------------|------------ xmit by sockperf: 90% | 100%

XDP socket(AF_XDP) is an excellent bypass kernel network framework. The zero copy feature of xsk (XDP socket) needs to be supported by the driver. The performance of zero copy is very good. ENV: Qemu with vhost. vhost cpu | Guest APP CPU |Guest Softirq CPU | PPS -----------------------------|---------------|------------------|------------ xmit by sockperf: 90% | 100%

XDP socket(AF_XDP) is an excellent bypass kernel network framework. The zero copy feature of xsk (XDP socket) needs to be supported by the driver. The performance of zero copy is very good. ENV: Qemu with vhost. vhost cpu | Guest APP CPU |Guest Softirq CPU | PPS -----------------------------|---------------|------------------|------------ xmit by sockperf: 90% | 100%

## About DMA APIs Now, virtio may can not work with DMA APIs when virtio features do not have VIRTIO_F_ACCESS_PLATFORM. 1. I tried to let DMA APIs return phy address by virtio-device. But DMA APIs just work with the "real" devices. 2. I tried to let xsk support callballs to get phy address from virtio-net driver as the dma address. But the maintainers of xsk may want to use

On Tue, Feb 14, 2023 at 3:27 PM Xuan Zhuo <xuanzhuo at linux.alibaba.com> wrote: > > DMA-related logic is separated from the virtqueue_add_split to prepare > for subsequent support for premapped. The patch seems to do more than what is described here. To simplify reviewers, I'd suggest to split this patch into three: 1) virtqueue_add_split_prepare() (could we have a better

## About DMA APIs Now, virtio may can not work with DMA APIs when virtio features do not have VIRTIO_F_ACCESS_PLATFORM. 1. I tried to let DMA APIs return phy address by virtio-device. But DMA APIs just work with the "real" devices. 2. I tried to let xsk support callballs to get phy address from virtio-net driver as the dma address. But the maintainers of xsk may want to use

DMA-related logic is separated from the virtqueue_add_vring_packed() to prepare for subsequent support for premapped. DMA address will be saved as sg->dma_address, then virtqueue_add_vring_packed() will use it directly. If it is a premapped scene, the transmitted sgs should have saved DMA address in dma_address, and in virtio core, we need to pass virtqueue_map_sgs(). Signed-off-by: Xuan

## About DMA APIs Now, virtio may can not work with DMA APIs when virtio features do not have VIRTIO_F_ACCESS_PLATFORM. 1. I tried to let DMA APIs return phy address by virtio-device. But DMA APIs just work with the "real" devices. 2. I tried to let xsk support callballs to get phy address from virtio-net driver as the dma address. But the maintainers of xsk may want to use

virtio core only supports virtual addresses, dma is completed in virtio core. In some scenarios (such as the AF_XDP), the memory is allocated and DMA mapping is completed in advance, so it is necessary for us to support passing the DMA address to virtio core. Drives can use sg->dma_address to pass the mapped dma address to virtio core. If one sg->dma_address is used then all sgs must use

On Tue, Mar 21, 2023 at 05:34:59PM +0800, Xuan Zhuo wrote: > virtio core only supports virtual addresses, dma is completed in virtio > core. > > In some scenarios (such as the AF_XDP), the memory is allocated > and DMA mapping is completed in advance, so it is necessary for us to > support passing the DMA address to virtio core. > > Drives can use sg->dma_address to

DMA-related logic is separated from the virtqueue_add_split() to one new function. DMA address will be saved as sg->dma_address if use_dma_api is true, then virtqueue_add_split() will use it directly. Unmap operation will be simpler. The purpose of this is to facilitate subsequent support to receive dma address mapped by drivers. Signed-off-by: Xuan Zhuo <xuanzhuo at linux.alibaba.com>

Hello everyone, This RFC implements a subset of packed ring which is described at https://github.com/oasis-tcs/virtio-docs/blob/master/virtio-v1.1-packed-wd08.pdf The code was tested with DPDK vhost (testpmd/vhost-PMD) implemented by Jens at http://dpdk.org/ml/archives/dev/2018-January/089417.html Minor changes are needed for the vhost code, e.g. to kick the guest. It's not a complete

Hi, This patch set implements packed ring support in virtio driver. A performance test between pktgen (pktgen_sample03_burst_single_flow.sh) and DPDK vhost (testpmd/rxonly/vhost-PMD) has been done, I saw ~30% performance gain in packed ring in this case. To make this patch set work with below patch set for vhost, some hacks are needed to set the _F_NEXT flag in indirect descriptors (this should

Hi, This patch set implements packed ring support in virtio driver. A performance test between pktgen (pktgen_sample03_burst_single_flow.sh) and DPDK vhost (testpmd/rxonly/vhost-PMD) has been done, I saw ~30% performance gain in packed ring in this case. To make this patch set work with below patch set for vhost, some hacks are needed to set the _F_NEXT flag in indirect descriptors (this should