Re: [virtio-comment] [PATCH V2 2/2] virtio: introduce STOP status bit

[Stefan Hajnoczi <stefanha@redhat.com>]

On Fri, Jul 16, 2021 at 10:03:17AM +0800, Jason Wang wrote:
> 
> å 2021/7/15 äå6:01, Stefan Hajnoczi åé:
> > On Thu, Jul 15, 2021 at 09:35:13AM +0800, Jason Wang wrote:
> > > å 2021/7/14 äå11:07, Stefan Hajnoczi åé:
> > > > On Wed, Jul 14, 2021 at 06:29:28PM +0800, Jason Wang wrote:
> > > > > å 2021/7/14 äå5:53, Stefan Hajnoczi åé:
> > > > > > On Tue, Jul 13, 2021 at 08:16:35PM +0800, Jason Wang wrote:
> > > > > > > å 2021/7/13 äå6:00, Stefan Hajnoczi åé:
> > > > > > > > On Tue, Jul 13, 2021 at 11:27:03AM +0800, Jason Wang wrote:
> > > > > > > > > å 2021/7/12 äå5:57, Stefan Hajnoczi åé:
> > > > > > > > > > On Mon, Jul 12, 2021 at 12:00:39PM +0800, Jason Wang wrote:
> > > > > > > > > > > å 2021/7/11 äå4:36, Michael S. Tsirkin åé:
> > > > > > > > > > > > On Fri, Jul 09, 2021 at 07:23:33PM +0200, Eugenio Perez Martin wrote:
> > > > > > > > > > > > > > >         If I understand correctly, this is all
> > > > > > > > > > > > > > > driven from the driver inside the guest, so for this to work
> > > > > > > > > > > > > > > the guest must be running and already have initialised the driver.
> > > > > > > > > > > > > > Yes.
> > > > > > > > > > > > > > 
> > > > > > > > > > > > > As I see it, the feature can be driven entirely by the VMM as long as
> > > > > > > > > > > > > it intercept the relevant configuration space (PCI, MMIO, etc) from
> > > > > > > > > > > > > guest's reads and writes, and present it as coherent and transparent
> > > > > > > > > > > > > for the guest. Some use cases I can imagine with a physical device (or
> > > > > > > > > > > > > vp_vpda device) with VIRTIO_F_STOP:
> > > > > > > > > > > > > 
> > > > > > > > > > > > > 1) The VMM chooses not to pass the feature flag. The guest cannot stop
> > > > > > > > > > > > > the device, so any write to this flag is an error/undefined.
> > > > > > > > > > > > > 2) The VMM passes the flag to the guest. The guest can stop the device.
> > > > > > > > > > > > > 2.1) The VMM stops the device to perform a live migration, and the
> > > > > > > > > > > > > guest does not write to STOP in any moment of the LM. It resets the
> > > > > > > > > > > > > destination device with the state, and then initializes the device.
> > > > > > > > > > > > > 2.2) The guest stops the device and, when STOP(32) is set, the source
> > > > > > > > > > > > > VMM migrates the device status. The destination VMM realizes the bit,
> > > > > > > > > > > > > so it sets the bit in the destination too after device initialization.
> > > > > > > > > > > > > 2.3) The device is not initialized by the guest so it doesn't matter
> > > > > > > > > > > > > what bit has the HW, but the VM can be migrated.
> > > > > > > > > > > > > 
> > > > > > > > > > > > > Am I missing something?
> > > > > > > > > > > > > 
> > > > > > > > > > > > > Thanks!
> > > > > > > > > > > > It's doable like this. It's all a lot of hoops to jump through though.
> > > > > > > > > > > > It's also not easy for devices to implement.
> > > > > > > > > > > It just requires a new status bit. Anything that makes you think it's hard
> > > > > > > > > > > to implement?
> > > > > > > > > > > 
> > > > > > > > > > > E.g for networking device, it should be sufficient to use this bit + the
> > > > > > > > > > > virtqueue state.
> > > > > > > > > > > 
> > > > > > > > > > > 
> > > > > > > > > > > > Why don't we design the feature in a way that is useable by VMMs
> > > > > > > > > > > > and implementable by devices in a simple way?
> > > > > > > > > > > It use the common technology like register shadowing without any further
> > > > > > > > > > > stuffs.
> > > > > > > > > > > 
> > > > > > > > > > > Or do you have any other ideas?
> > > > > > > > > > > 
> > > > > > > > > > > (I think we all know migration will be very hard if we simply pass through
> > > > > > > > > > > those state registers).
> > > > > > > > > > If an admin virtqueue is used instead of the STOP Device Status field
> > > > > > > > > > bit then there's no need to re-read the Device Status field in a loop
> > > > > > > > > > until the device has stopped.
> > > > > > > > > Probably not. Let me to clarify several points:
> > > > > > > > > 
> > > > > > > > > - This proposal has nothing to do with admin virtqueue. Actually, admin
> > > > > > > > > virtqueue could be used for carrying any basic device facility like status
> > > > > > > > > bit. E.g I'm going to post patches that use admin virtqueue as a "transport"
> > > > > > > > > for device slicing at virtio level.
> > > > > > > > > - Even if we had introduced admin virtqueue, we still need a per function
> > > > > > > > > interface for this. This is a must for nested virtualization, we can't
> > > > > > > > > always expect things like PF can be assigned to L1 guest.
> > > > > > > > > - According to the proposal, there's no need for the device to complete all
> > > > > > > > > the consumed buffers, device can choose to expose those inflight descriptors
> > > > > > > > > in a device specific way and set the STOP bit. This means, if we have the
> > > > > > > > > device specific in-flight descriptor reporting facility, the device can
> > > > > > > > > almost set the STOP bit immediately.
> > > > > > > > > - If we don't go with the basic device facility but using the admin
> > > > > > > > > virtqueue specific method, we still need to clarify how it works with the
> > > > > > > > > device status state machine, it will be some kind of sub-states which looks
> > > > > > > > > much more complicated than the current proposal.
> > > > > > > > > 
> > > > > > > > > 
> > > > > > > > > > When migrating a guest with many VIRTIO devices a busy waiting approach
> > > > > > > > > > extends downtime if implemented sequentially (stopping one device at a
> > > > > > > > > > time).
> > > > > > > > > Well. You need some kinds of waiting for sure, the device/DMA needs sometime
> > > > > > > > > to be stopped. The downtime is determined by a specific virtio
> > > > > > > > > implementation which is hard to be restricted at the spec level. We can
> > > > > > > > > clarify that the device must set the STOP bit in e.g 100ms.
> > > > > > > > > 
> > > > > > > > > 
> > > > > > > > > >       It can be implemented concurrently (setting the STOP bit on all
> > > > > > > > > > devices and then looping until all their Device Status fields have the
> > > > > > > > > > bit set), but this becomes more complex to implement.
> > > > > > > > > I still don't get what kind of complexity did you worry here.
> > > > > > > > > 
> > > > > > > > > 
> > > > > > > > > > I'm a little worried about adding a new bit that requires busy
> > > > > > > > > > waiting...
> > > > > > > > > Busy wait is not something that is introduced in this patch:
> > > > > > > > > 
> > > > > > > > > 4.1.4.3.2 Driver Requirements: Common configuration structure layout
> > > > > > > > > 
> > > > > > > > > After writing 0 to device_status, the driver MUST wait for a read of
> > > > > > > > > device_status to return 0 before reinitializing the device.
> > > > > > > > > 
> > > > > > > > > Since it was required for at least one transport. We need do something
> > > > > > > > > similar to when introducing basic facility.
> > > > > > > > Adding the STOP but as a Device Status bit is a small and clean VIRTIO
> > > > > > > > spec change. I like that.
> > > > > > > > 
> > > > > > > > On the other hand, devices need time to stop and that time can be
> > > > > > > > unbounded. For example, software virtio-blk/scsi implementations since
> > > > > > > > cannot immediately cancel in-flight I/O requests on Linux hosts.
> > > > > > > > 
> > > > > > > > The natural interface for long-running operations is virtqueue requests.
> > > > > > > > That's why I mentioned the alternative of using an admin virtqueue
> > > > > > > > instead of a Device Status bit.
> > > > > > > So I'm not against the admin virtqueue. As said before, admin virtqueue
> > > > > > > could be used for carrying the device status bit.
> > > > > > > 
> > > > > > > Send a command to set STOP status bit to admin virtqueue. Device will make
> > > > > > > the command buffer used after it has successfully stopped the device.
> > > > > > > 
> > > > > > > AFAIK, they are not mutually exclusive, since they are trying to solve
> > > > > > > different problems.
> > > > > > > 
> > > > > > > Device status - basic device facility
> > > > > > > 
> > > > > > > Admin virtqueue - transport/device specific way to implement (part of) the
> > > > > > > device facility
> > > > > > > 
> > > > > > > > Although you mentioned that the stopped state needs to be reflected in
> > > > > > > > the Device Status field somehow, I'm not sure about that since the
> > > > > > > > driver typically doesn't need to know whether the device is being
> > > > > > > > migrated.
> > > > > > > The guest won't see the real device status bit. VMM will shadow the device
> > > > > > > status bit in this case.
> > > > > > > 
> > > > > > > E.g with the current vhost-vDPA, vDPA behave like a vhost device, guest is
> > > > > > > unaware of the migration.
> > > > > > > 
> > > > > > > STOP status bit is set by Qemu to real virtio hardware. But guest will only
> > > > > > > see the DRIVER_OK without STOP.
> > > > > > > 
> > > > > > > It's not hard to implement the nested on top, see the discussion initiated
> > > > > > > by Eugenio about how expose VIRTIO_F_STOP to guest for nested live
> > > > > > > migration.
> > > > > > > 
> > > > > > > 
> > > > > > > >      In fact, the VMM would need to hide this bit and it's safer to
> > > > > > > > keep it out-of-band instead of risking exposing it by accident.
> > > > > > > See above, VMM may choose to hide or expose the capability. It's useful for
> > > > > > > migrating a nested guest.
> > > > > > > 
> > > > > > > If we design an interface that can be used in the nested environment, it's
> > > > > > > not an ideal interface.
> > > > > > > 
> > > > > > > 
> > > > > > > > In addition, stateful devices need to load/save non-trivial amounts of
> > > > > > > > data. They need DMA to do this efficiently, so an admin virtqueue is a
> > > > > > > > good fit again.
> > > > > > > I don't get the point here. You still need to address the exact the similar
> > > > > > > issues for admin virtqueue: the unbound time in freezing the device, the
> > > > > > > interaction with the virtio device status state machine.
> > > > > > Device state state can be large so a register interface would be a
> > > > > > bottleneck. DMA is needed. I think a virtqueue is a good fit for
> > > > > > saving/loading device state.
> > > > > So this patch doesn't mandate a register interface, isn't it?
> > > > You're right, not this patch. I mentioned it because your other patch
> > > > series ("[PATCH] virtio-pci: implement VIRTIO_F_RING_STATE") implements
> > > > it a register interface.
> > > > 
> > > > > And DMA
> > > > > doesn't means a virtqueue, it could be a transport specific method.
> > > > Yes, although virtqueues are a pretty good interface that works across
> > > > transports (PCI/MMIO/etc) thanks to the standard vring memory layout.
> > > > 
> > > > > I think we need to start from defining the state of one specific device and
> > > > > see what is the best interface.
> > > > virtio-blk might be the simplest. I think virtio-net has more device
> > > > state and virtio-scsi is definitely more complext than virtio-blk.
> > > > 
> > > > First we need agreement on whether "device state" encompasses the full
> > > > state of the device or just state that is unknown to the VMM.
> > > 
> > > I think we've discussed this in the past. It can't work since:
> > > 
> > > 1) The state and its format must be clearly defined in the spec
> > > 2) We need to maintain migration compatibility and debug-ability
> > Some devices need implementation-specific state. They should still be
> > able to live migrate even if it means cross-implementation migration and
> > debug-ability is not possible.
> 
> 
> I think we need to re-visit this conclusion. Migration compatibility is
> pretty important, especially consider the software stack has spent a huge
> mount of effort in maintaining them.
> 
> Say a virtio hardware would break this, this mean we will lose all the
> advantages of being a standard device.
> 
> If we can't do live migration among:
> 
> 1) different backends, e.g migrate from virtio hardware to migrate software
> 2) different vendors
> 
> We failed to say as a standard device and the customer is in fact locked by
> the vendor implicitly.

My virtiofs device implementation is backed by an in-memory file system.
The device state includes the contents of each file.

Your virtiofs device implementation uses Linux file handles to keep
track of open files. The device state includes Linux file handles (but
not the contents of each file) so the destination host can access the
same files on shared storage.

Cornelia's virtiofs device implementation is backed by an object storage
HTTP API. The device state includes API object IDs.

The device state is implementation-dependent. There is no standard
representation and it's not possible to migrate between device
implementations. How are they supposed to migrate?

This is why I think it's necessarily to allow implementation-specific
device state representations.

> > 
> > > 3) Not a proper uAPI desgin
> > I never understood this argument. The Linux uAPI passes through lots of
> > opaque data from devices to userspace. Allowing an
> > implementation-specific device state representation is nothing new. VFIO
> > already does it.
> 
> 
> I think we've already had a lots of discussion for VFIO but without a
> conclusion. Maybe we need the verdict from Linus or Greg (not sure if it's
> too late). But that's not related to virito and this thread.
> 
> What you propose here is kind of conflict with the efforts of virtio. I
> think we all aggree that we should define the state in the spec. Assuming
> this is correct:
> 
> 1) why do we still offer opaque migration state to userspace?

See above. Stateful devices may require an implementation-defined device
state representation.

> 2) how can it be integrated into the current VMM (Qemu) virtio devices'
> migration bytes stream?

Opaque data like D-Bus VMState:
https://qemu.readthedocs.io/en/latest/interop/dbus-vmstate.html

> > 
> > > 
> > > > That's
> > > > basically the difference between the vhost/vDPA's selective passthrough
> > > > approach and VFIO's full passthrough approach.
> > > 
> > > We can't do VFIO full pasthrough for migration anyway, some kind of mdev is
> > > required but it's duplicated with the current vp_vdpa driver.
> > I'm not sure that's true. Generic VFIO PCI migration can probably be
> > achieved without mdev:
> > 1. Define a migration PCI Capability that indicates support for
> >     VFIO_REGION_TYPE_MIGRATION. This allows the PCI device to implement
> >     the migration interface in hardware instead of an mdev driver.
> 
> 
> So I think it still depend on the driver to implement migrate state which is
> vendor specific.

The current VFIO migration interface depends on a device-specific
software mdev driver but here I'm showing that the physical device can
implement the migration interface so that no device-specific driver code
is needed.

> 
> Note that it's just an uAPI definition not something defined in the PCI
> spec.

Yes, that's why I mentioned Changpeng Liu's idea to turn the uAPI into a
standard PCI Capability to eliminate the need for device-specific
drivers.

> 
> Out of curiosity, the patch is merged without any real users in the Linux.
> This is very bad since we lose the change to audit the whole design.

I agree. It would have helped to have a complete vision for how live
migration should work along with demos. I don't see any migration code
in samples/vfio-mdev/ :(.

> > 2. The VMM either uses the migration PCI Capability directly from
> >     userspace or core VFIO PCI code advertises VFIO_REGION_TYPE_MIGRATION
> >     to userspace so migration can proceed in the same way as with
> >     VFIO/mdev drivers.
> > 3. The PCI Capability is not passed through to the guest.
> 
> 
> This brings troubles in the nested environment.

It depends on the device splitting/management design. If L0 wishes to
let L1 manage the VFs then it would need to expose a management device.
Since the migration interface is generic (not device-specific) a generic
management device solves this for all devices.

Stefan

Attachment: signature.asc
Description: PGP signature