similar to: [PATCH 2/2] virtio_net: fix race in RX VQ processing

On 07/08/2013 05:04 PM, Michael S. Tsirkin wrote: > virtio net called virtqueue_enable_cq on RX path after napi_complete, so > with NAPI_STATE_SCHED clear - outside the implicit napi lock. > This violates the requirement to synchronize virtqueue_enable_cq wrt > virtqueue_add_buf. In particular, used event can move backwards, > causing us to lose interrupts. > In a debug build,

Hello. On 08-07-2013 13:04, Michael S. Tsirkin wrote: > virtio net called virtqueue_enable_cq on RX path after napi_complete, so > with NAPI_STATE_SCHED clear - outside the implicit napi lock. > This violates the requirement to synchronize virtqueue_enable_cq wrt > virtqueue_add_buf. In particular, used event can move backwards, > causing us to lose interrupts. > In a debug

virtio net called virtqueue_enable_cq on RX path after napi_complete, so with NAPI_STATE_SCHED clear - outside the implicit napi lock. This violates the requirement to synchronize virtqueue_enable_cq wrt virtqueue_add_buf. In particular, used event can move backwards, causing us to lose interrupts. In a debug build, this can trigger panic within START_USE. Jason Wang reports that he can trigger

virtio net called virtqueue_enable_cq on RX path after napi_complete, so with NAPI_STATE_SCHED clear - outside the implicit napi lock. This violates the requirement to synchronize virtqueue_enable_cq wrt virtqueue_add_buf. In particular, used event can move backwards, causing us to lose interrupts. In a debug build, this can trigger panic within START_USE. Jason Wang reports that he can trigger

On Fri, 2013-07-12 at 23:13 +0200, Wolfram Gloger wrote: > Hi, > > Today I merged M. Tsirkin's patch v2 "virtio_net: fix race in RX VQ > processing": > > http://lkml.indiana.edu/hypermail/linux/kernel/1307.1/00503.html > > into 3.2.48 and lightly tested the result (vhost-net, virtio-disk > and 9pfs using virtio). This sounds like it could be suitable

On Fri, 2013-07-12 at 23:13 +0200, Wolfram Gloger wrote: > Hi, > > Today I merged M. Tsirkin's patch v2 "virtio_net: fix race in RX VQ > processing": > > http://lkml.indiana.edu/hypermail/linux/kernel/1307.1/00503.html > > into 3.2.48 and lightly tested the result (vhost-net, virtio-disk > and 9pfs using virtio). This sounds like it could be suitable

According to David, proper accounting and queueing (at all levels, not just TCP sockets) is more important than trying to skim a bunch of cycles by avoiding TX interrupts. Having an event to free the SKB is absolutely essential for the stack to operate correctly. This series tries to enable tx interrupt for virtio-net. The idea is simple: enable tx interrupt and schedule a tx napi to free old

According to David, proper accounting and queueing (at all levels, not just TCP sockets) is more important than trying to skim a bunch of cycles by avoiding TX interrupts. Having an event to free the SKB is absolutely essential for the stack to operate correctly. This series tries to enable tx interrupt for virtio-net. The idea is simple: enable tx interrupt and schedule a tx napi to free old

Since commit 39e6c8208d7b ("net: solve a NAPI race") napi has been able to be rescheduled within napi_complete_done() even in non-busypoll case, but virtnet_poll() always enabled interrupts before complete, and when napi was rescheduled within napi_complete_done() it did not disable interrupts. This caused more interrupts when event idx is disabled. According to commit cbdadbbf0c79

Since commit 39e6c8208d7b ("net: solve a NAPI race") napi has been able to be rescheduled within napi_complete_done() even in non-busypoll case, but virtnet_poll() always enabled interrupts before complete, and when napi was rescheduled within napi_complete_done() it did not disable interrupts. This caused more interrupts when event idx is disabled. According to commit cbdadbbf0c79

On 10/15/2014 06:18 PM, Michael S. Tsirkin wrote: > On Wed, Oct 15, 2014 at 03:25:29PM +0800, Jason Wang wrote: >> > Orphan skb in ndo_start_xmit() breaks socket accounting and packet >> > queuing. This in fact breaks lots of things such as pktgen and several >> > TCP optimizations. And also make BQL can't be implemented for >> > virtio-net. >> >

On 10/15/2014 06:18 PM, Michael S. Tsirkin wrote: > On Wed, Oct 15, 2014 at 03:25:29PM +0800, Jason Wang wrote: >> > Orphan skb in ndo_start_xmit() breaks socket accounting and packet >> > queuing. This in fact breaks lots of things such as pktgen and several >> > TCP optimizations. And also make BQL can't be implemented for >> > virtio-net. >> >

On newer hosts that support delayed tx interrupts, we probably don't have much to gain from orphaning packets early. Based on patch by Jason Wang. Note: this will likely degrade performance for hosts without event idx support. Various fallback options are available, including orphaning conditionally. Testing TBD. Signed-off-by: Michael S. Tsirkin <mst at redhat.com> ---

On newer hosts that support delayed tx interrupts, we probably don't have much to gain from orphaning packets early. Based on patch by Jason Wang. Note: this will likely degrade performance for hosts without event idx support. Various fallback options are available, including orphaning conditionally. Testing TBD. Signed-off-by: Michael S. Tsirkin <mst at redhat.com> ---

Orphan skb in ndo_start_xmit() breaks socket accounting and packet queuing. This in fact breaks lots of things such as pktgen and several TCP optimizations. And also make BQL can't be implemented for virtio-net. This patch tries to solve this issue by enabling tx interrupt. To avoid introducing extra spinlocks, a tx napi was scheduled to free those packets. More tx interrupt mitigation

Orphan skb in ndo_start_xmit() breaks socket accounting and packet queuing. This in fact breaks lots of things such as pktgen and several TCP optimizations. And also make BQL can't be implemented for virtio-net. This patch tries to solve this issue by enabling tx interrupt. To avoid introducing extra spinlocks, a tx napi was scheduled to free those packets. More tx interrupt mitigation

Hi all: This series introduces the support for rx busy polling support. This was useful for reduing the latency for a kvm guest. Instead of introducing new states and spinlocks, this series re-uses NAPI state to synchonrize between NAPI and busy polling. This grealy simplified the codes and reduce the overheads of spinlocks for normal NAPI fast path. Test was done between a kvm guest and an

Hi all: This series introduces the support for rx busy polling support. This was useful for reduing the latency for a kvm guest. Instead of introducing new states and spinlocks, this series re-uses NAPI state to synchonrize between NAPI and busy polling. This grealy simplified the codes and reduce the overheads of spinlocks for normal NAPI fast path. Test was done between a kvm guest and an

From: Eric Dumazet <edumazet at google.com> Straightforward patch to add GRO processing to virtio_net. napi_complete_done() usage allows more aggressive aggregation, opted-in by setting /sys/class/net/xxx/gro_flush_timeout Tested: Setting /sys/class/net/xxx/gro_flush_timeout to 1000 nsec, Rick Jones reported following results. One VM of each on a pair of OpenStack compute nodes with

From: Eric Dumazet <edumazet at google.com> Straightforward patch to add GRO processing to virtio_net. napi_complete_done() usage allows more aggressive aggregation, opted-in by setting /sys/class/net/xxx/gro_flush_timeout Tested: Setting /sys/class/net/xxx/gro_flush_timeout to 1000 nsec, Rick Jones reported following results. One VM of each on a pair of OpenStack compute nodes with