similar to: [RFC V4 PATCH 0/8] Packed ring layout for vhost

Hi all: This RFC implement packed ring layout. The code were tested with Tiwei's RFC V2 a thttps://lkml.org/lkml/2018/4/1/48. Some fixups and tweaks were needed on top of Tiwei's code to make it run. TCP stream and pktgen does not show obvious difference compared with split ring. Changes from V2: - do not use & in checking desc_event_flags - off should be most significant bit -

Hi all: This RFC implement packed ring layout. The code were tested with Tiwei's RFC V2 a thttps://lkml.org/lkml/2018/4/1/48. Some fixups and tweaks were needed on top of Tiwei's code to make it run. TCP stream and pktgen does not show obvious difference compared with split ring. Changes from V2: - do not use & in checking desc_event_flags - off should be most significant bit -

Hi all: This RFC implement packed ring layout. The code were tested with Tiwei's RFC V5 at https://lkml.org/lkml/2018/5/22/138. Some fixups and tweaks were needed on top of Tiwei's code to make it run for event index. Pktgen reports about 20% improvement on TX PPS when doing pktgen from guest to host. No ovbious improvement on RX PPS. We can do lots of optimizations on top but for simple

Hi all: This RFC implement packed ring layout. The code were tested with pmd implement by Jens at http://dpdk.org/ml/archives/dev/2018-January/089417.html. Minor change was needed for pmd codes to kick virtqueue since it assumes a busy polling backend. Test were done between localhost and guest. Testpmd (rxonly) in guest reports 2.4Mpps. Testpmd (txonly) repots about 2.1Mpps. Notes: The event

Hi all: This RFC implement packed ring layout. The code were tested with pmd implement by Jens at http://dpdk.org/ml/archives/dev/2018-January/089417.html. Minor change was needed for pmd codes to kick virtqueue since it assumes a busy polling backend. Test were done between localhost and guest. Testpmd (rxonly) in guest reports 2.4Mpps. Testpmd (txonly) repots about 2.1Mpps. Notes: The event

Hi all: This series implements packed virtqueues. The code were tested with Tiwei's guest driver series at https://patchwork.ozlabs.org/cover/942297/ Pktgen test for both RX and TX does not show obvious difference with split virtqueues. The main bottleneck is the guest Linux driver, since it can not stress vhost for a 100% CPU utilization. A full TCP benchmark is ongoing. Will test

Hi all: This series implements packed virtqueues. The code were tested with Tiwei's guest driver series at https://patchwork.ozlabs.org/cover/942297/ Pktgen test for both RX and TX does not show obvious difference with split virtqueues. The main bottleneck is the guest Linux driver, since it can not stress vhost for a 100% CPU utilization. A full TCP benchmark is ongoing. Will test

Hi all: This series implements packed virtqueues. The code were tested with Tiwei's RFC V6 at https://lkml.org/lkml/2018/6/5/120. Pktgen test for both RX and TX does not show obvious difference with split virtqueues. The main bottleneck is the guest Linux driver, since it can not stress vhost for a 100% CPU utilization. A full TCP benchmark is ongoing. Will test virtio-net pmd as well when

Hi all: This series implements packed virtqueues. The code were tested with Tiwei's RFC V6 at https://lkml.org/lkml/2018/6/5/120. Pktgen test for both RX and TX does not show obvious difference with split virtqueues. The main bottleneck is the guest Linux driver, since it can not stress vhost for a 100% CPU utilization. A full TCP benchmark is ongoing. Will test virtio-net pmd as well when

Hi all: This series implements packed virtqueues which were described at [1]. In this version we try to address the performance regression saw by V2. The root cause is packed virtqueue need more times of userspace memory accesssing which turns out to be very expensive. Thanks to the help of 7f466032dc9e ("vhost: access vq metadata through kernel virtual address"), such overhead cold be

Hi all: This series implements packed virtqueues which were described at [1]. In this version we try to address the performance regression saw by V2. The root cause is packed virtqueue need more times of userspace memory accesssing which turns out to be very expensive. Thanks to the help of 7f466032dc9e ("vhost: access vq metadata through kernel virtual address"), such overhead cold be

We let the specifics of the ring format seep through to vhost API callers - mostly because there was only one format so it was hard to imagine what an independent API would look like. Now that there's an alternative in form of the packed ring, it's easier to see the issues, and fixing them is perhaps the cleanest way to add support for more formats. This patchset does this by indtroducing

This adds infrastructure required for supporting multiple ring formats. The idea is as follows: we convert descriptors to an independent format first, and process that converting to iov later. Used ring is similar: we fetch into an independent struct first, convert that to IOV later. The point is that we have a tight loop that fetches descriptors, which is good for cache utilization. This will

Below should be useful for some experiments Jason is doing. I thought I'd send it out for early review/feedback. Compiled-only at this point. event idx feature allows us to defer interrupts until a specific # of descriptors were used. Sometimes it might be useful to get an interrupt after a specific descriptor, regardless. This adds a descriptor flag for this, and an API to create an urgent

Below should be useful for some experiments Jason is doing. I thought I'd send it out for early review/feedback. Compiled-only at this point. event idx feature allows us to defer interrupts until a specific # of descriptors were used. Sometimes it might be useful to get an interrupt after a specific descriptor, regardless. This adds a descriptor flag for this, and an API to create an urgent

This adds infrastructure required for supporting multiple ring formats. The idea is as follows: we convert descriptors to an independent format first, and process that converting to iov later. Used ring is similar: we fetch into an independent struct first, convert that to IOV later. The point is that we have a tight loop that fetches descriptors, which is good for cache utilization. This will

This intentionally leaves "fixup" changes separate - hopefully that is enough to fix vhost-net crashes reported here, but it helps me keep track of what changed. I will naturally squash them later when we are done. This adds infrastructure required for supporting multiple ring formats. The idea is as follows: we convert descriptors to an independent format first, and process that

This intentionally leaves "fixup" changes separate - hopefully that is enough to fix vhost-net crashes reported here, but it helps me keep track of what changed. I will naturally squash them later when we are done. This adds infrastructure required for supporting multiple ring formats. The idea is as follows: we convert descriptors to an independent format first, and process that

This patch adds mergeable receive buffer support to vhost_net. Signed-off-by: David L Stevens <dlstevens at us.ibm.com> diff -ruNp net-next-v0/drivers/vhost/net.c net-next-v7/drivers/vhost/net.c --- net-next-v0/drivers/vhost/net.c 2010-04-24 21:36:54.000000000 -0700 +++ net-next-v7/drivers/vhost/net.c 2010-04-28 12:26:18.000000000 -0700 @@ -74,6 +74,23 @@ static int move_iovec_hdr(struct

This patch adds mergeable receive buffer support to vhost_net. Signed-off-by: David L Stevens <dlstevens at us.ibm.com> diff -ruNp net-next-v0/drivers/vhost/net.c net-next-v7/drivers/vhost/net.c --- net-next-v0/drivers/vhost/net.c 2010-04-24 21:36:54.000000000 -0700 +++ net-next-v7/drivers/vhost/net.c 2010-04-28 12:26:18.000000000 -0700 @@ -74,6 +74,23 @@ static int move_iovec_hdr(struct