search for: desc_event_flags

...v, wrap; > + int off; > + > + /* Flush out used descriptors updates. This is paired > + * with the barrier that the Guest executes when enabling > + * interrupts. > + */ > + smp_mb(); > + > + if (vhost_get_avail(vq, event_flags, > + &vq->driver_event->desc_event_flags) < 0) { > + vq_err(vq, "Failed to get driver desc_event_flags"); > + return true; > + } > + > + if (!(event_flags & cpu_to_vhost16(vq, RING_EVENT_FLAGS_DESC))) > + return event_flags == > + cpu_to_vhost16(vq, RING_EVENT_FLAGS_ENABLE); Maybe it would...

...tic int vhost_update_used_flags(struct vhost_virtqueue *vq) return 0; } +static int vhost_update_device_flags(struct vhost_virtqueue *vq, + __virtio16 device_flags) +{ + void __user *flags; + + if (vhost_put_user(vq, cpu_to_vhost16(vq, device_flags), + &vq->device_event->desc_event_flags, + VHOST_ADDR_DESC) < 0) + return -EFAULT; + if (unlikely(vq->log_used)) { + /* Make sure the flag is seen before log. */ + smp_wmb(); + /* Log used flag write. */ + flags = &vq->device_event->desc_event_flags; + log_write(vq->log_base, vq->log_addr + + (flags -...

...false; > > > > +#endif > > > > + > > > > + off_wrap = virtio16_to_cpu(_vq->vdev, vq->vring_packed.device->off_wrap); > > > > + > > > > + if (vq->event) { > > > It looks to me we should examine RING_EVENT_FLAGS_DESC in desc_event_flags > > > instead of vq->event here. Spec does not forces to use evenf_off and > > > event_wrap if event index is enabled. > > > > > > > + // FIXME: fix this! > > > > + needs_kick = ((off_wrap >> 15) == vq->wrap_counter) && >...

...false; > > > > +#endif > > > > + > > > > + off_wrap = virtio16_to_cpu(_vq->vdev, vq->vring_packed.device->off_wrap); > > > > + > > > > + if (vq->event) { > > > It looks to me we should examine RING_EVENT_FLAGS_DESC in desc_event_flags > > > instead of vq->event here. Spec does not forces to use evenf_off and > > > event_wrap if event index is enabled. > > > > > > > + // FIXME: fix this! > > > > + needs_kick = ((off_wrap >> 15) == vq->wrap_counter) && >...

...include/uapi/linux/virtio_config.h | 12 +- >>>>> include/uapi/linux/virtio_ring.h | 61 ++ >>>>> 4 files changed, 980 insertions(+), 195 deletions(-) >>> [...] [...] >>>> It looks to me we should examine RING_EVENT_FLAGS_DESC in desc_event_flags >>>> instead of vq->event here. Spec does not forces to use evenf_off and >>>> event_wrap if event index is enabled. >>>> >>>>> + // FIXME: fix this! >>>>> + needs_kick = ((off_wrap >> 15) == vq->wrap_counter) &&amp...

...struct vring { > struct vring_used *used; > }; > >+struct vring_packed_desc_event { >+ /* Descriptor Event Offset */ >+ __virtio16 desc_event_off : 15, >+ /* Descriptor Event Wrap Counter */ >+ desc_event_wrap : 1; >+ /* Descriptor Event Flags */ >+ __virtio16 desc_event_flags : 2; >+}; Where would the virtqueue number go in driver notifications? regards, Jens

...rom: Tiwei Bie > Sent: 23 February 2018 11:18 ... > +struct vring_packed_desc_event { > + /* Descriptor Event Offset */ > + __virtio16 desc_event_off : 15, > + /* Descriptor Event Wrap Counter */ > + desc_event_wrap : 1; > + /* Descriptor Event Flags */ > + __virtio16 desc_event_flags : 2; > +}; This looks like you are assuming that a bit-field has a defined layout and can be used to map a 'hardware' structure. The don't, don't use them like that. David

...> > + } > > + vq->last_add_time_valid = false; > > +#endif > > + > > + off_wrap = virtio16_to_cpu(_vq->vdev, vq->vring_packed.device->off_wrap); > > + > > + if (vq->event) { > > It looks to me we should examine RING_EVENT_FLAGS_DESC in desc_event_flags > instead of vq->event here. Spec does not forces to use evenf_off and > event_wrap if event index is enabled. > > > + // FIXME: fix this! > > + needs_kick = ((off_wrap >> 15) == vq->wrap_counter) && > > + vring_need_event(off_wrap & ~(1&...

...> > + } > > + vq->last_add_time_valid = false; > > +#endif > > + > > + off_wrap = virtio16_to_cpu(_vq->vdev, vq->vring_packed.device->off_wrap); > > + > > + if (vq->event) { > > It looks to me we should examine RING_EVENT_FLAGS_DESC in desc_event_flags > instead of vq->event here. Spec does not forces to use evenf_off and > event_wrap if event index is enabled. > > > + // FIXME: fix this! > > + needs_kick = ((off_wrap >> 15) == vq->wrap_counter) && > > + vring_need_event(off_wrap & ~(1&...

...t;>> + vq->last_add_time_valid = false; >>> +#endif >>> + >>> + off_wrap = virtio16_to_cpu(_vq->vdev, vq->vring_packed.device->off_wrap); >>> + >>> + if (vq->event) { >> It looks to me we should examine RING_EVENT_FLAGS_DESC in desc_event_flags >> instead of vq->event here. Spec does not forces to use evenf_off and >> event_wrap if event index is enabled. >> >>> + // FIXME: fix this! >>> + needs_kick = ((off_wrap >> 15) == vq->wrap_counter) && >>> + vring_need_event...

...with split ring. > > I have to ask then - what is the benefit of this? You can use this with dpdk within guest. The DPDK version did see a performance improvement so hopefully with future versions this will too. > > > > Changes from V2: > > - do not use & in checking desc_event_flags > > - off should be most significant bit > > - remove the workaround of mergeable buffer for dpdk prototype > > - id should be in the last descriptor in the chain > > - keep _F_WRITE for write descriptor when adding used > > - device flags updating should use ADDR_USED...

...with split ring. > > I have to ask then - what is the benefit of this? You can use this with dpdk within guest. The DPDK version did see a performance improvement so hopefully with future versions this will too. > > > > Changes from V2: > > - do not use & in checking desc_event_flags > > - off should be most significant bit > > - remove the workaround of mergeable buffer for dpdk prototype > > - id should be in the last descriptor in the chain > > - keep _F_WRITE for write descriptor when adding used > > - device flags updating should use ADDR_USED...

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

...ks > > Wei > >> Changes from V3: >> - Fix math on event idx checking >> - Sync last avail wrap counter through GET/SET_VRING_BASE >> - remove desc_event prefix in the driver/device structure >> >> Changes from V2: >> - do not use & in checking desc_event_flags >> - off should be most significant bit >> - remove the workaround of mergeable buffer for dpdk prototype >> - id should be in the last descriptor in the chain >> - keep _F_WRITE for write descriptor when adding used >> - device flags updating should use ADDR_USED type...

...;> Changes from RFC V3: >> >> - Fix math on event idx checking >> - Sync last avail wrap counter through GET/SET_VRING_BASE >> - remove desc_event prefix in the driver/device structure >> >> Changes from RFC V2: >> >> - do not use & in checking desc_event_flags >> - off should be most significant bit >> - remove the workaround of mergeable buffer for dpdk prototype >> - id should be in the last descriptor in the chain >> - keep _F_WRITE for write descriptor when adding used >> - device flags updating should use ADDR_USED type...

...;> Changes from RFC V3: >> >> - Fix math on event idx checking >> - Sync last avail wrap counter through GET/SET_VRING_BASE >> - remove desc_event prefix in the driver/device structure >> >> Changes from RFC V2: >> >> - do not use & in checking desc_event_flags >> - off should be most significant bit >> - remove the workaround of mergeable buffer for dpdk prototype >> - id should be in the last descriptor in the chain >> - keep _F_WRITE for write descriptor when adding used >> - device flags updating should use ADDR_USED type...

...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 - remove the workaround of mergeable buffer for dpdk prototype - id should be in the last descriptor in the chain - keep _F_WRITE for write descriptor when adding used - device flags updating should use ADDR_USED type - return error on unexpected unavail descrip...

...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 - remove the workaround of mergeable buffer for dpdk prototype - id should be in the last descriptor in the chain - keep _F_WRITE for write descriptor when adding used - device flags updating should use ADDR_USED type - return error on unexpected unavail descrip...

...l/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. I have to ask then - what is the benefit of this? > > Changes from V2: > - do not use & in checking desc_event_flags > - off should be most significant bit > - remove the workaround of mergeable buffer for dpdk prototype > - id should be in the last descriptor in the chain > - keep _F_WRITE for write descriptor when adding used > - device flags updating should use ADDR_USED type > - return error...