similar to: [RFC PATCH 00/24] Control VQ support in vDPA

v2: - rebased on Linus' tree, commit ae3419fbac84 ("vc_screen: don't clobber return value in vcs_read") - removed `struct task_struct *owner` param (unused for now, maybe ?useful to support cgroups) [Jason] - add unbind_mm callback [Jason] - call the new unbind_mm callback during the release [Jason] - avoid to call bind_mm callback after the reset, since the device ?is not

In order to reduce needlessly high setup and teardown cost of iotlb mapping during live migration, it's crucial to decouple the vhost-vdpa iotlb abstraction from the virtio device life cycle, i.e. iotlb mappings should be left intact across virtio device reset [1]. For it to work, the on-chip IOMMU parent device should implement a separate .reset_map() operation callback to restore 1:1 DMA

In order to reduce needlessly high setup and teardown cost of iotlb mapping during live migration, it's crucial to decouple the vhost-vdpa iotlb abstraction from the virtio device life cycle, i.e. iotlb mappings should be left intact across virtio device reset [1]. For it to work, the on-chip IOMMU parent device should implement a separate .reset_map() operation callback to restore 1:1 DMA

Hi All: This series introduces API for reporing IOVA range. This is a must for userspace to work correclty: - for the process that uses vhost-vDPA directly to properly allocate IOVA - for VM(qemu), when vIOMMU is not enabled, fail early if GPA is out of range - for VM(qemu), when vIOMMU is enabled, determine a valid guest address width Please review. Changes from V1: - do not mandate

Hi all: This is an update version of vDPA support in kernel. vDPA device is a device that uses a datapath which complies with the virtio specifications with vendor specific control path. vDPA devices can be both physically located on the hardware or emulated by software. vDPA hardware devices are usually implemented through PCIE with the following types: - PF (Physical Function) - A single

Hi all: This is an update version of vDPA support in kernel. vDPA device is a device that uses a datapath which complies with the virtio specifications with vendor specific control path. vDPA devices can be both physically located on the hardware or emulated by software. vDPA hardware devices are usually implemented through PCIE with the following types: - PF (Physical Function) - A single

This series adds support for the use of user virtual addresses in the vDPA simulator devices. The main reason for this change is to lift the pinning of all guest memory. Especially with virtio devices implemented in software. The next step would be to generalize the code in vdpa-sim to allow the implementation of in-kernel software devices. Similar to vhost, but using vDPA so we can reuse the

Hi all: This is an updated version of kernel support for vDPA device. Various changes were made based on the feedback since last verion. One major change is to drop the sysfs API and leave the management interface for future development, and introudce the incremental DMA bus operations. Please see changelog for more information. The work on vhost, IFCVF (intel VF driver for vDPA) and qemu is

Hi All: This series introduces API for reporing IOVA range. This is a must for userspace to work correclty: - for the process that uses vhost-vDPA directly to properly allocate IOVA - for VM(qemu), when vIOMMU is not enabled, fail early if GPA is out of range - for VM(qemu), when vIOMMU is enabled, determine a valid guest address width Please review. Thanks Jason Wang (4): vdpa:

Hi All: This series introduces API for reporing IOVA range. This is a must for userspace to work correclty: - for the process that uses vhost-vDPA directly to properly allocate IOVA - for VM(qemu), when vIOMMU is not enabled, fail early if GPA is out of range - for VM(qemu), when vIOMMU is enabled, determine a valid guest address width Please review. Thanks Jason Wang (4): vdpa:

Hi all: This series tries to support batched IOTLB updating vhost-vdpa. Currently vhost-vdpa accepts userspace mapping via IOTLB API, and it can only forward one mapping to IOMMU or device through IOMMU API or dma_map(). Though set_map() is deisgend to have the capability to pass an rbtree based mapping to vDPA device, it's still be called at least once for each VHOST_IOTLB_UPDATE or

This series adds support for the use of user virtual addresses in the vDPA simulator devices. The main reason for this change is to lift the pinning of all guest memory. Especially with virtio devices implemented in software. The next step would be to generalize the code in vdpa-sim to allow the implementation of in-kernel software devices. Similar to vhost, but using vDPA so we can reuse the

Hi all: This series tries to support batched IOTLB updating vhost-vdpa. Currently vhost-vdpa accepts userspace mapping via IOTLB API, and it can only forward one mapping to IOMMU or device through IOMMU API or dma_map(). Though set_map() is designed to have the capability to pass an rbtree based mapping to vDPA device, it's still be called at least once for each VHOST_IOTLB_UPDATE or

Let's move work management inside the vdpa_sim core. This way we can easily change how we manage the works, without having to change the devices each time. Acked-by: Eugenio P??rez Martin <eperezma at redhat.com> Acked-by: Jason Wang <jasowang at redhat.com> Signed-off-by: Stefano Garzarella <sgarzare at redhat.com> --- drivers/vdpa/vdpa_sim/vdpa_sim.h | 3 ++-

The spinlock we use to protect the state of the simulator is sometimes held for a long time (for example, when devices handle requests). This also prevents us from calling functions that might sleep (such as kthread_flush_work() in the next patch), and thus having to release and retake the lock. For these reasons, let's replace the spinlock with a mutex that gives us more flexibility.

On 2020/8/4 ??5:00, Michael S. Tsirkin wrote: > 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:

On 2020/8/4 ??5:00, Michael S. Tsirkin wrote: > 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:

VDPA sim stores config space as native endian, but that is wrong: modern guests expect LE. I coded up the following to fix it up, but it is wrong too: vdpasim_create is called before guest features are known. So what should we do? New ioctl to specify the interface used? More ideas? Signed-off-by: Michael S. Tsirkin <mst at redhat.com> --- drivers/vdpa/vdpa_sim/vdpa_sim.c | 22

VDPA sim stores config space as native endian, but that is wrong: modern guests expect LE. I coded up the following to fix it up, but it is wrong too: vdpasim_create is called before guest features are known. So what should we do? New ioctl to specify the interface used? More ideas? Signed-off-by: Michael S. Tsirkin <mst at redhat.com> --- drivers/vdpa/vdpa_sim/vdpa_sim.c | 22

Config space endian-ness is currently a mess: fields are not tagged with the correct endian-ness so it's easy to make mistakes like instanciating config space in native endian-ness. The following patches adding sparse tagging are currently in my tree. Lightly tested. As a follow-up, I plan to add new APIs that handle modern config space in a more efficient way (bypassing the version check).