qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] [RFC v4 00/16] VIRTIO-IOMMU device


From: Auger Eric
Subject: Re: [Qemu-devel] [RFC v4 00/16] VIRTIO-IOMMU device
Date: Wed, 27 Sep 2017 17:38:47 +0200
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:45.0) Gecko/20100101 Thunderbird/45.4.0

Hi Tomasz

On 27/09/2017 13:07, Tomasz Nowicki wrote:
> Hi Eric,
> 
> Out of curiosity, I compared your SMMUv3 full emulation against
> virtio-iommu. For virtio-net device behind the virtio-iommu I get 50%
> performance of what I can get for SMMUv3 emulation.
> 
> Do you have similar observations? Since there is no need to emulate HW
> behaviour in QEMU I expected virtio-iommu to be faster. I will run some
> more benchmarks.

No I don't have formal comparative figures at the moment. Benchmarking
is also on the top of my todo list though.

One reason could be that virtio-iommu sends requests for each map
operations whereas this does not happen for vsmmuv3 (in non caching
mode). You could try to turn the vsmmuv3 caching-mode option and I guess
you should have more similar perf figures between virtio-iommu and
vsmuv3.

Thanks

Eric
> 
> Thanks,
> Tomasz
> 
> On 19.09.2017 09:46, Eric Auger wrote:
>> This series implements the virtio-iommu device.
>>
>> This v4 is an upgrade to v0.4 spec [1] and applies on QEMU v2.10.0.
>> - probe request support although no reserved region is returned at
>>    the moment
>> - unmap semantics less strict, as specified in v0.4
>> - device registration, attach/detach revisited
>> - split into smaller patches to ease review
>> - propose a way to inform the IOMMU mr about the page_size_mask
>>    of underlying HW IOMMU, if any
>> - remove warning associated with the translation of the MSI doorbell
>>
>> The device gets instantiated using the "-device virtio-iommu-device"
>> option. It currently works with ARM virt machine only, as the machine
>> must handle the dt binding between the virtio-mmio "iommu" node and
>> the PCI host bridge node.
>>
>> The associated VHOST/VFIO adaptation is available on the branch
>> below but is not officially delivered as part of this series.
>>
>> Best Regards
>>
>> Eric
>>
>> This series can be found at:
>> https://github.com/eauger/qemu/tree/v2.10.0-virtio-iommu-v4
>>
>> References:
>> [1] [RFC] virtio-iommu version 0.4
>>      git://linux-arm.org/virtio-iommu.git branch viommu/v0.4
>>
>> Testing:
>> - guest kernel with v0.4 virtio-iommu driver (4kB page)
>> - tested with guest using virtio-pci-net and vhost-net
>>    (,vhost=on/off,iommu_platform,disable-modern=off,disable-legacy=on  )
>> - tested with VFIO and guest assigned with 2 VFs
>> - tested with DPDK on guest with 2 assigned VFs
>>
>> Not tested:
>> - hot-plug/hot-unplug of EP: not implemented
>>
>> History:
>> v2-> v4:
>> - see above
>>
>> v2 -> v3:
>> - rebase on top of 2.10-rc0 and especially
>>    [PATCH qemu v9 0/2] memory/iommu: QOM'fy IOMMU MemoryRegion
>> - add mutex init
>> - fix as->mappings deletion using g_tree_ref/unref
>> - when a dev is attached whereas it is already attached to
>>    another address space, first detach it
>> - fix some error values
>> - page_sizes = TARGET_PAGE_MASK;
>> - I haven't changed the unmap() semantics yet, waiting for the
>>    next virtio-iommu spec revision.
>>
>> v1 -> v2:
>> - fix redifinition of viommu_as typedef
>>
>> Eric Auger (16):
>>    update-linux-headers: import virtio_iommu.h
>>    linux-headers: Update for virtio-iommu
>>    virtio-iommu: add skeleton
>>    virtio-iommu: Decode the command payload
>>    virtio-iommu: Add the iommu regions
>>    virtio-iommu: Register attached devices
>>    virtio-iommu: Implement attach/detach command
>>    virtio-iommu: Implement map/unmap
>>    virtio-iommu: Implement translate
>>    virtio-iommu: Implement probe request
>>    hw/arm/virt: Add 2.11 machine type
>>    hw/arm/virt: Add virtio-iommu to the virt board
>>    memory.h: Add set_page_size_mask IOMMUMemoryRegion callback
>>    hw/vfio/common: Set the IOMMUMemoryRegion supported page sizes
>>    virtio-iommu: Implement set_page_size_mask
>>    hw/vfio/common: Do not print error when viommu translates into an mmio
>>      region
>>
>>   hw/arm/virt.c                                 | 115 +++-
>>   hw/vfio/common.c                              |   7 +-
>>   hw/virtio/Makefile.objs                       |   1 +
>>   hw/virtio/trace-events                        |  24 +
>>   hw/virtio/virtio-iommu.c                      | 923
>> ++++++++++++++++++++++++++
>>   include/exec/memory.h                         |   4 +
>>   include/hw/arm/virt.h                         |   5 +
>>   include/hw/vfio/vfio-common.h                 |   1 +
>>   include/hw/virtio/virtio-iommu.h              |  61 ++
>>   include/standard-headers/linux/virtio_ids.h   |   1 +
>>   include/standard-headers/linux/virtio_iommu.h | 177 +++++
>>   linux-headers/linux/virtio_iommu.h            |   1 +
>>   scripts/update-linux-headers.sh               |   3 +
>>   13 files changed, 1312 insertions(+), 11 deletions(-)
>>   create mode 100644 hw/virtio/virtio-iommu.c
>>   create mode 100644 include/hw/virtio/virtio-iommu.h
>>   create mode 100644 include/standard-headers/linux/virtio_iommu.h
>>   create mode 100644 linux-headers/linux/virtio_iommu.h
>>
> 



reply via email to

[Prev in Thread] Current Thread [Next in Thread]