search for: in_sg

...+ > dma_unmap_page(vring_dma_dev(vq), > extra[i].addr, > extra[i].len, > @@ -751,6 +755,7 @@ static inline int virtqueue_add_split(struct virtqueue *_vq, > unsigned int in_sgs, > void *data, > void *ctx, > + bool dma_map, > gfp_t gfp) > { > struct vring_virtqueue *vq = to_vvq(_vq); > @@ -767,9...

...virtqueue_add_split(struct virtqueue *_vq, > - struct scatterlist *sgs[], > - unsigned int total_sg, > - unsigned int out_sgs, > - unsigned int in_sgs, > - void *data, > - void *ctx, > - gfp_t gfp) > +static int virtqueue_map_sgs(struct vring_virtqueue *vq, > + struct scatterlist *sgs[], > +...

...dma_unmap_page(vring_dma_dev(vq), > > extra[i].addr, > > extra[i].len, > > @@ -751,6 +755,7 @@ static inline int virtqueue_add_split(struct virtqueue *_vq, > > unsigned int in_sgs, > > void *data, > > void *ctx, > > + bool dma_map, > > gfp_t gfp) > > { > > struct vring_virtqueue *vq...

...rection direction) @@ -520,6 +528,80 @@ static inline unsigned int virtqueue_add_desc_split(struct virtqueue *vq, return next; } +static void virtqueue_unmap_sgs(struct vring_virtqueue *vq, + struct scatterlist *sgs[], + unsigned int total_sg, + unsigned int out_sgs, + unsigned int in_sgs) +{ + struct scatterlist *sg; + unsigned int n; + + if (!vq->use_dma_api) + return; + + for (n = 0; n < out_sgs; n++) { + for (sg = sgs[n]; sg; sg = sg_next(sg)) { + if (!sg->dma_address) + return; + + dma_unmap_page(vring_dma_dev(vq), sg->dma_address, + sg->lengt...

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%

...ing.c @@ -520,6 +520,73 @@ static inline unsigned int virtqueue_add_desc_split(struct virtqueue *vq, return next; } +static int virtqueue_map_sgs(struct vring_virtqueue *vq, + struct scatterlist *sgs[], + unsigned int total_sg, + unsigned int out_sgs, + unsigned int in_sgs) +{ + struct scatterlist *sg; + unsigned int n; + + for (n = 0; n < out_sgs; n++) { + for (sg = sgs[n]; sg; sg = sg_next(sg)) { + dma_addr_t addr = vring_map_one_sg(vq, sg, DMA_TO_DEVICE); + + if (vring_mapping_error(vq, addr)) + return -ENOMEM; + + sg->dma_address = addr; + } + }...

OK, so I've spent a few days benchmarking. Turns out 80% of virtio_add_buf cases are uni-directional (including the always-performance-sensitive networking code), and that gets no performance penalty (though tests with real networking would be appreciated!). I'm not reposting all the "convert driver to virtio_add_outbuf()" patches: just the scsi one which I didn't have

OK, so I've spent a few days benchmarking. Turns out 80% of virtio_add_buf cases are uni-directional (including the always-performance-sensitive networking code), and that gets no performance penalty (though tests with real networking would be appreciated!). I'm not reposting all the "convert driver to virtio_add_outbuf()" patches: just the scsi one which I didn't have

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%

...rection direction) @@ -520,6 +528,80 @@ static inline unsigned int virtqueue_add_desc_split(struct virtqueue *vq, return next; } +static void virtqueue_unmap_sgs(struct vring_virtqueue *vq, + struct scatterlist *sgs[], + unsigned int total_sg, + unsigned int out_sgs, + unsigned int in_sgs) +{ + struct scatterlist *sg; + unsigned int n; + + if (!vq->use_dma_api) + return; + + for (n = 0; n < out_sgs; n++) { + for (sg = sgs[n]; sg; sg = sg_next(sg)) { + if (!sg->dma_address) + return; + + dma_unmap_page(vring_dma_dev(vq), sg->dma_address, + sg->lengt...

...rection direction) @@ -520,6 +528,80 @@ static inline unsigned int virtqueue_add_desc_split(struct virtqueue *vq, return next; } +static void virtqueue_unmap_sgs(struct vring_virtqueue *vq, + struct scatterlist *sgs[], + unsigned int total_sg, + unsigned int out_sgs, + unsigned int in_sgs) +{ + struct scatterlist *sg; + unsigned int n; + + if (!vq->use_dma_api) + return; + + for (n = 0; n < out_sgs; n++) { + for (sg = sgs[n]; sg; sg = sg_next(sg)) { + if (!sg->dma_address) + return; + + dma_unmap_page(vring_dma_dev(vq), sg->dma_address, + sg->lengt...

...ng_virtqueue *vq, struct scatterlist *sgs[], - struct scatterlist *(*next) - (struct scatterlist *, unsigned int *), unsigned int total_sg, - unsigned int total_out, - unsigned int total_in, unsigned int out_sgs, unsigned int in_sgs, gfp_t gfp) @@ -144,7 +126,7 @@ static inline int vring_add_indirect(struct vring_virtqueue *vq, /* Transfer entries from the sg lists into the indirect page */ i = 0; for (n = 0; n < out_sgs; n++) { - for (sg = sgs[n]; sg; sg = next(sg, &total_out)) { + for (sg = sgs[n];...

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%

...t; doubt that dma vs. non-dma is relevant. I tried -net user a -net tap >> with iperf running in both directions. I also tried switching >> virtio_pci into non-DMA-API mode. No errors. >> >> Is there any chance that you could instrument that BUG_ON to print n, >> i, in_sgs, out_sgs, total_sg, total_in, and total_out? I assume this >> is some oddity with patch 1, but I'm mystified. > > Yes, its triggered by patch 1. I must be doing something wrong, since virtio_net isn't sending me through that code path at all. This may be related to ethtool r...

...t; doubt that dma vs. non-dma is relevant. I tried -net user a -net tap >> with iperf running in both directions. I also tried switching >> virtio_pci into non-DMA-API mode. No errors. >> >> Is there any chance that you could instrument that BUG_ON to print n, >> i, in_sgs, out_sgs, total_sg, total_in, and total_out? I assume this >> is some oddity with patch 1, but I'm mystified. > > Yes, its triggered by patch 1. I must be doing something wrong, since virtio_net isn't sending me through that code path at all. This may be related to ethtool r...

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%

...program key or write request can also combine with a result read (VIRTIO_RPMB_REQ_RESULT_READ) for a returned result. Now I'm deep in the guts of virt queues doing the implementation I'm trying to clarify exactly what frames should be sent to the device and if they should be out_sgs or in_sgs. I suspect there is some difference between the original prototype interface and what we have now. Some operations obviously have an implied response (VIRTIO_RPMB_REQ_GET_WRITE_COUNTER and VIRTIO_RPMB_REQ_DATA_READ). As far as I could tell the frame should be simple: sg[0] (out_sg=1) - frame w...

## 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