| Message ID | 20200329150953.23812-1-yuri.benditovich@daynix.com |
|---|---|
| Headers | show |
| Series | reference implementation of RSS and hash report | expand |
On Sun, Mar 29, 2020 at 06:09:46PM +0300, Yuri Benditovich wrote: > Support for VIRTIO_NET_F_RSS and VIRTIO_NET_F_HASH_REPORT > features in QEMU for reference purpose. > Implements Toeplitz hash calculation for incoming > packets according to configuration provided by driver. > Uses calculated hash for decision on receive virtqueue > and/or reports the hash in the virtio header Reviewed-by: Michael S. Tsirkin <mst@redhat.com> Probably post 5.0 material. > Changes from v6: > Fixed a bug in patch 5 "reference implementation of hash report" > that caused the ASAN test to fail > was: n->rss_data.populate_hash = true; > fixed: n->rss_data.populate_hash = !!hash_report; > > Yuri Benditovich (7): > virtio-net: introduce RSS and hash report features > virtio-net: implement RSS configuration command > virtio-net: implement RX RSS processing > tap: allow extended virtio header with hash info > virtio-net: reference implementation of hash report > vmstate.h: provide VMSTATE_VARRAY_UINT16_ALLOC macro > virtio-net: add migration support for RSS and hash report > > hw/net/trace-events | 3 + > hw/net/virtio-net.c | 448 +++++++++++++++++++++++++++++++-- > include/hw/virtio/virtio-net.h | 16 ++ > include/migration/vmstate.h | 10 + > net/tap.c | 11 +- > 5 files changed, 460 insertions(+), 28 deletions(-) > > -- > 2.17.1
Michael/Jason, As Linux headers was updated in qemu and now include RSC/RSS/Hash definitions, please let me know what you prefer: 1. You apply this series as is, then I submit clean-up series that will remove all the redundant defines from virtio-net.c 2. I post v8 of this series with cleanup of all the redundant defines and also RSC ones 3. Something other Thanks, Yuri Benditovich On Tue, Mar 31, 2020 at 5:26 PM Michael S. Tsirkin <mst@redhat.com> wrote: > On Sun, Mar 29, 2020 at 06:09:46PM +0300, Yuri Benditovich wrote: > > Support for VIRTIO_NET_F_RSS and VIRTIO_NET_F_HASH_REPORT > > features in QEMU for reference purpose. > > Implements Toeplitz hash calculation for incoming > > packets according to configuration provided by driver. > > Uses calculated hash for decision on receive virtqueue > > and/or reports the hash in the virtio header > > > Reviewed-by: Michael S. Tsirkin <mst@redhat.com> > > Probably post 5.0 material. > > > Changes from v6: > > Fixed a bug in patch 5 "reference implementation of hash report" > > that caused the ASAN test to fail > > was: n->rss_data.populate_hash = true; > > fixed: n->rss_data.populate_hash = !!hash_report; > > > > Yuri Benditovich (7): > > virtio-net: introduce RSS and hash report features > > virtio-net: implement RSS configuration command > > virtio-net: implement RX RSS processing > > tap: allow extended virtio header with hash info > > virtio-net: reference implementation of hash report > > vmstate.h: provide VMSTATE_VARRAY_UINT16_ALLOC macro > > virtio-net: add migration support for RSS and hash report > > > > hw/net/trace-events | 3 + > > hw/net/virtio-net.c | 448 +++++++++++++++++++++++++++++++-- > > include/hw/virtio/virtio-net.h | 16 ++ > > include/migration/vmstate.h | 10 + > > net/tap.c | 11 +- > > 5 files changed, 460 insertions(+), 28 deletions(-) > > > > -- > > 2.17.1 > >
We are in freeze so nothing's applied right now. v8 which has all the bits will be a good step so we are ready for after freeze. On Fri, May 01, 2020 at 07:01:58AM +0300, Yuri Benditovich wrote: > Michael/Jason, > > As Linux headers was updated in qemu and now include RSC/RSS/Hash definitions, > please let me know what you prefer: > 1. You apply this series as is, then I submit clean-up series that will remove > all the redundant defines from virtio-net.c > 2. I post v8 of this series with cleanup of all the redundant defines and also > RSC ones > 3. Something other > > Thanks, > Yuri Benditovich > > On Tue, Mar 31, 2020 at 5:26 PM Michael S. Tsirkin <mst@redhat.com> wrote: > > On Sun, Mar 29, 2020 at 06:09:46PM +0300, Yuri Benditovich wrote: > > Support for VIRTIO_NET_F_RSS and VIRTIO_NET_F_HASH_REPORT > > features in QEMU for reference purpose. > > Implements Toeplitz hash calculation for incoming > > packets according to configuration provided by driver. > > Uses calculated hash for decision on receive virtqueue > > and/or reports the hash in the virtio header > > > Reviewed-by: Michael S. Tsirkin <mst@redhat.com> > > Probably post 5.0 material. > > > Changes from v6: > > Fixed a bug in patch 5 "reference implementation of hash report" > > that caused the ASAN test to fail > > was: n->rss_data.populate_hash = true; > > fixed: n->rss_data.populate_hash = !!hash_report; > > > > Yuri Benditovich (7): > > virtio-net: introduce RSS and hash report features > > virtio-net: implement RSS configuration command > > virtio-net: implement RX RSS processing > > tap: allow extended virtio header with hash info > > virtio-net: reference implementation of hash report > > vmstate.h: provide VMSTATE_VARRAY_UINT16_ALLOC macro > > virtio-net: add migration support for RSS and hash report > > > > hw/net/trace-events | 3 + > > hw/net/virtio-net.c | 448 +++++++++++++++++++++++++++++++-- > > include/hw/virtio/virtio-net.h | 16 ++ > > include/migration/vmstate.h | 10 + > > net/tap.c | 11 +- > > 5 files changed, 460 insertions(+), 28 deletions(-) > > > > -- > > 2.17.1 > >
OK so now 5.0's out, I think it's easier if you just send v8, but it's up to Jason. On Fri, May 01, 2020 at 10:44:06AM -0400, Michael S. Tsirkin wrote: > We are in freeze so nothing's applied right now. > v8 which has all the bits will be a good step so we > are ready for after freeze. > > On Fri, May 01, 2020 at 07:01:58AM +0300, Yuri Benditovich wrote: > > Michael/Jason, > > > > As Linux headers was updated in qemu and now include RSC/RSS/Hash definitions, > > please let me know what you prefer: > > 1. You apply this series as is, then I submit clean-up series that will remove > > all the redundant defines from virtio-net.c > > 2. I post v8 of this series with cleanup of all the redundant defines and also > > RSC ones > > 3. Something other > > > > Thanks, > > Yuri Benditovich > > > > On Tue, Mar 31, 2020 at 5:26 PM Michael S. Tsirkin <mst@redhat.com> wrote: > > > > On Sun, Mar 29, 2020 at 06:09:46PM +0300, Yuri Benditovich wrote: > > > Support for VIRTIO_NET_F_RSS and VIRTIO_NET_F_HASH_REPORT > > > features in QEMU for reference purpose. > > > Implements Toeplitz hash calculation for incoming > > > packets according to configuration provided by driver. > > > Uses calculated hash for decision on receive virtqueue > > > and/or reports the hash in the virtio header > > > > > > Reviewed-by: Michael S. Tsirkin <mst@redhat.com> > > > > Probably post 5.0 material. > > > > > Changes from v6: > > > Fixed a bug in patch 5 "reference implementation of hash report" > > > that caused the ASAN test to fail > > > was: n->rss_data.populate_hash = true; > > > fixed: n->rss_data.populate_hash = !!hash_report; > > > > > > Yuri Benditovich (7): > > > virtio-net: introduce RSS and hash report features > > > virtio-net: implement RSS configuration command > > > virtio-net: implement RX RSS processing > > > tap: allow extended virtio header with hash info > > > virtio-net: reference implementation of hash report > > > vmstate.h: provide VMSTATE_VARRAY_UINT16_ALLOC macro > > > virtio-net: add migration support for RSS and hash report > > > > > > hw/net/trace-events | 3 + > > > hw/net/virtio-net.c | 448 +++++++++++++++++++++++++++++++-- > > > include/hw/virtio/virtio-net.h | 16 ++ > > > include/migration/vmstate.h | 10 + > > > net/tap.c | 11 +- > > > 5 files changed, 460 insertions(+), 28 deletions(-) > > > > > > -- > > > 2.17.1 > > > >
On Sun, Mar 29, 2020 at 06:09:46PM +0300, Yuri Benditovich wrote: > Support for VIRTIO_NET_F_RSS and VIRTIO_NET_F_HASH_REPORT > features in QEMU for reference purpose. > Implements Toeplitz hash calculation for incoming > packets according to configuration provided by driver. > Uses calculated hash for decision on receive virtqueue > and/or reports the hash in the virtio header Series: Reviewed-by: Michael S. Tsirkin <mst@redhat.com> to be queued through Jason's tree. > Changes from v6: > Fixed a bug in patch 5 "reference implementation of hash report" > that caused the ASAN test to fail > was: n->rss_data.populate_hash = true; > fixed: n->rss_data.populate_hash = !!hash_report; > > Yuri Benditovich (7): > virtio-net: introduce RSS and hash report features > virtio-net: implement RSS configuration command > virtio-net: implement RX RSS processing > tap: allow extended virtio header with hash info > virtio-net: reference implementation of hash report > vmstate.h: provide VMSTATE_VARRAY_UINT16_ALLOC macro > virtio-net: add migration support for RSS and hash report > > hw/net/trace-events | 3 + > hw/net/virtio-net.c | 448 +++++++++++++++++++++++++++++++-- > include/hw/virtio/virtio-net.h | 16 ++ > include/migration/vmstate.h | 10 + > net/tap.c | 11 +- > 5 files changed, 460 insertions(+), 28 deletions(-) > > -- > 2.17.1
On 2020/5/1 下午12:01, Yuri Benditovich wrote: > Michael/Jason, > > As Linux headers was updated in qemu and now include RSC/RSS/Hash > definitions, please let me know what you prefer: > 1. You apply this series as is, then I submit clean-up series that > will remove all the redundant defines from virtio-net.c > 2. I post v8 of this series with cleanup of all the redundant defines > and also RSC ones > 3. Something other Hi Yuri: Though I've queued this series but consider we have new headers, I think it's better to post v8. Thanks
I'll send v8 soon Thanks, Yuri On Wed, May 6, 2020 at 8:37 AM Jason Wang <jasowang@redhat.com> wrote: > > On 2020/5/1 下午12:01, Yuri Benditovich wrote: > > Michael/Jason, > > > > As Linux headers was updated in qemu and now include RSC/RSS/Hash > > definitions, please let me know what you prefer: > > 1. You apply this series as is, then I submit clean-up series that > > will remove all the redundant defines from virtio-net.c > > 2. I post v8 of this series with cleanup of all the redundant defines > > and also RSC ones > > 3. Something other > > > Hi Yuri: > > Though I've queued this series but consider we have new headers, I think > it's better to post v8. > > Thanks > >