Add a front end VDPA driver that registers in the VDPA bus and provides
networking to a guest. The VDPA driver creates the necessary resources
on the VF it is driving such that data path will be offloaded.
Notifications are being communicated through the driver.
Currently, only VFs are supported. In subsequent patches we will have
devlink support to control which VF is used for VDPA and which function
is used for regular networking.
Reviewed-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Eli Cohen <eli@mellanox.com>
Link: https://lore.kernel.org/r/20200804162048.22587-13-eli@mellanox.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Add code to support registering address space region for the device. The
virtio driver can run as either:
1. Guest virtio driver
2. Userspace virtio driver on the host
3. Kernel virtio driver on the host
In any case a memory key object is required to provide access to memory
for the device.
This code will be shared by network or block driver implementations.
Reviewed-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Eli Cohen <eli@mellanox.com>
Link: https://lore.kernel.org/r/20200804162048.22587-12-eli@mellanox.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
For now VQ state involves 16 bit available index value encoded in u64
variable. In the future it will be extended to contain more fields. Use
struct to contain the state, now containing only a single u16 for the
available index. In the future we can add fields to this struct.
Reviewed-by: Parav Pandit <parav@mellanox.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Eli Cohen <eli@mellanox.com>
Link: https://lore.kernel.org/r/20200804162048.22587-8-eli@mellanox.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The vDPA simulator support both set_map() and dma_map()/dma_unmap()
operations. But vhost-vdpa can only use one of them. So this patch
introduce a module parameter (batch_mapping) that let vpda_sim to
support only one of those dma operations. The batched mapping via
set_map() is enabled by default.
Signed-off-by: Jason Wang <jasowang@redhat.com>
Link: https://lore.kernel.org/r/20200804162048.22587-6-eli@mellanox.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
This patches extend the vhost IOTLB API to accept batch updating hints
form userspace. When userspace wants update the device IOTLB in a
batch, it may do:
1) Write vhost_iotlb_msg with VHOST_IOTLB_BATCH_BEGIN flag
2) Perform a batch of IOTLB updating via VHOST_IOTLB_UPDATE/INVALIDATE
3) Write vhost_iotlb_msg with VHOST_IOTLB_BATCH_END flag
Vhost-vdpa may decide to batch the IOMMU/IOTLB updating in step 3 when
vDPA device support set_map() ops. This is useful for the vDPA device
that want to know all the mappings to tweak their own DMA translation
logic.
For vDPA device that doesn't require set_map(), no behavior changes.
This capability is advertised via VHOST_BACKEND_F_IOTLB_BATCH capability.
Signed-off-by: Jason Wang <jasowang@redhat.com>
Link: https://lore.kernel.org/r/20200804162048.22587-5-eli@mellanox.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Merge shared code that's necessary for mlx5 vdpa bits to build.
The branch itself includes a small number of patches and is also
independently merged by a couple of other trees.
That shouldn't cause any conflicts by itself.
Saeed Mahameed <saeedm@mellanox.com> says:
---
mlx5-next is a very small branch based on a very early rc that includes
mlx5 shared stuff between rdma and net-next, and now virtio as well.
---
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
This patch introduce a set of functions for setup/unsetup
and update irq offloading respectively by register/unregister
and re-register the irq_bypass_producer.
With these functions, this commit can setup/unsetup
irq offloading through setting DRIVER_OK/!DRIVER_OK, and
update irq offloading through SET_VRING_CALL.
Signed-off-by: Zhu Lingshan <lingshan.zhu@intel.com>
Suggested-by: Jason Wang <jasowang@redhat.com>
Link: https://lore.kernel.org/r/20200731065533.4144-5-lingshan.zhu@intel.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
vDPA devices has dedicated backed hardware like
passthrough-ed devices. Then it is possible to setup irq
offloading to vCPU for vDPA devices. Thus this patch tries to
manipulated assigned device counters by
kvm_arch_start/end_assignment() in irqbypass manager, so that
assigned devices could be detected in update_pi_irte()
We will increase/decrease the assigned device counter in kvm/x86.
Both vDPA and VFIO would go through this code path.
Only X86 uses these counters and kvm_arch_start/end_assignment(),
so this code path only affect x86 for now.
Signed-off-by: Zhu Lingshan <lingshan.zhu@intel.com>
Suggested-by: Jason Wang <jasowang@redhat.com>
Link: https://lore.kernel.org/r/20200731065533.4144-3-lingshan.zhu@intel.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Make use of the flex_array_size() helper to calculate the size of a
flexible array member within an enclosing structure.
This helper offers defense-in-depth against potential integer
overflows, while at the same time makes it explicitly clear that
we are dealing with a flexible array member.
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Link: https://lore.kernel.org/r/20200731130956.GA30525@embeddedor
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
We used to have a per device feature whitelist to filter out the
unsupported virtio features. But this seems unnecessary since:
- the main idea behind feature whitelist is to block control vq
feature until we finalize the control virtqueue API. But the current
vhost-vDPA uAPI is sufficient to support control virtqueue. For
device that has hardware control virtqueue, the vDPA device driver
can just setup the hardware virtqueue and let userspace to use
hardware virtqueue directly. For device that doesn't have a control
virtqueue, the vDPA device driver need to use e.g vringh to emulate
a software control virtqueue.
- we don't do it in virtio-vDPA driver
So remove this limitation.
Signed-off-by: Jason Wang <jasowang@redhat.com>
Link: https://lore.kernel.org/r/20200720085043.16485-1-jasowang@redhat.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The loop may exist if vq->broken is true,
virtqueue_get_buf_ctx_packed or virtqueue_get_buf_ctx_split
will return NULL, so virtnet_poll will reschedule napi to
receive packet, it will lead cpu usage(si) to 100%.
call trace as below:
virtnet_poll
virtnet_receive
virtqueue_get_buf_ctx
virtqueue_get_buf_ctx_packed
virtqueue_get_buf_ctx_split
virtqueue_napi_complete
virtqueue_poll //return true
virtqueue_napi_schedule //it will reschedule napi
to fix this, return false if vq is broken in virtqueue_poll.
Signed-off-by: Mao Wenan <wenan.mao@linux.alibaba.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Link: https://lore.kernel.org/r/1596354249-96204-1-git-send-email-wenan.mao@linux.alibaba.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Speed and duplex config fields depend on VIRTIO_NET_F_SPEED_DUPLEX
which being 63>31 depends on VIRTIO_F_VERSION_1.
Accordingly, use LE accessors for these fields.
Reported-by: Cornelia Huck <cohuck@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
All drivers now use virtio_cread/write_le for LE config
space fields. Drop LE option from virtio_cread/write, only leaving
the option to access transitional fields.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
_Generic version allowed __uXX types but that is no longer necessary:
Transitional devices should all use __virtioXX types (and __leXX for
fields not present in the legacy devices).
Modern ones should use __leXX.
_uXX type would be a bug.
Let's prevent that.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
VDPA sim accesses config space as native endian - this is
wrong since it's a modern device and actually uses LE.
It only supports modern guests so we could punt and
just force LE, but let's use the full virtio APIs since people
tend to copy/paste code, and this is not data path anyway.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
We normally expect vdpa to use the modern interface.
However for consistency, let's use same APIs as vhost
for legacy guests.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Some legacy guests just assume features are 0 after reset.
We detect that config space is accessed before features are
set and set features to 0 automatically.
Note: some legacy guests might not even access config space, if this is
reported in the field we might need to catch a kick to handle these.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
mlxbf-tmfifo accesses config space using native types -
which works for it since the legacy virtio native types.
This will break if it ever needs to support modern virtio,
so with new tags previously introduced for virtio net config,
sparse now warns for this in drivers.
Since this is a legacy only device, fix it up using
virtio_legacy_is_little_endian for now.
No functional changes.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Cornelia Huck <cohuck@redhat.com>
Acked-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Acked-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Transitional devices should all use __virtioXX types (and __leXX for
fields not present in legacy devices).
Modern ones should use __leXX.
_uXX type would be a bug.
Let's prevent that.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Tag config space fields as having virtio endian-ness.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Since this is a modern-only device,
tag config space fields as having little endian-ness.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
