Re: [Qemu-devel] [virtio-dev] [PATCH v3 0/7] Vhost-pci for inter-VM communication

[Stefan Hajnoczi]

On Mon, Dec 11, 2017 at 01:53:40PM +0000, Wang, Wei W wrote:
> On Monday, December 11, 2017 7:12 PM, Stefan Hajnoczi wrote:
> > On Sat, Dec 09, 2017 at 04:23:17PM +0000, Wang, Wei W wrote:
> > > On Friday, December 8, 2017 4:34 PM, Stefan Hajnoczi wrote:
> > > > On Fri, Dec 8, 2017 at 6:43 AM, Wei Wang <address@hidden>
> > wrote:
> > > > > On 12/08/2017 07:54 AM, Michael S. Tsirkin wrote:
> > > > >>
> > > > >> On Thu, Dec 07, 2017 at 06:28:19PM +0000, Stefan Hajnoczi wrote:
> > > > >>>
> > > > >>> On Thu, Dec 7, 2017 at 5:38 PM, Michael S. Tsirkin 
> > > > >>> <address@hidden>
> > > > > Thanks Stefan and Michael for the sharing and discussion. I 
> > > > > think above 3 and 4 are debatable (e.g. whether it is simpler 
> > > > > really depends). 1 and 2 are implementations, I think both 
> > > > > approaches could implement the device that way. We originally 
> > > > > thought about one device and driver to support all types (called 
> > > > > it transformer sometimes :-) ), that would look interesting from 
> > > > > research point of view, but from real usage point of view, I 
> > > > > think it would be better to have them separated,
> > > > because:
> > > > > - different device types have different driver logic, mixing 
> > > > > them together would cause the driver to look messy. Imagine that 
> > > > > a networking driver developer has to go over the block related 
> > > > > code to debug, that also increases the difficulty.
> > > >
> > > > I'm not sure I understand where things get messy because:
> > > > 1. The vhost-pci device implementation in QEMU relays messages but 
> > > > has no device logic, so device-specific messages like 
> > > > VHOST_USER_NET_SET_MTU are trivial at this layer.
> > > > 2. vhost-user slaves only handle certain vhost-user protocol messages.
> > > > They handle device-specific messages for their device type only.
> > > > This is like vhost drivers today where the ioctl() function 
> > > > returns an error if the ioctl is not supported by the device.  It's not 
> > > > messy.
> > > >
> > > > Where are you worried about messy driver logic?
> > >
> > > Probably I didn’t explain well, please let me summarize my thought a 
> > > little
> > bit, from the perspective of the control path and data path.
> > >
> > > Control path: the vhost-user messages - I would prefer just have the 
> > > interaction between QEMUs, instead of relaying to the GuestSlave, 
> > > because
> > > 1) I think the claimed advantage (easier to debug and develop) 
> > > doesn’t seem very convincing
> > 
> > You are defining a mapping from the vhost-user protocol to a custom 
> > virtio device interface.  Every time the vhost-user protocol (feature 
> > bits, messages,
> > etc) is extended it will be necessary to map this new extension to the 
> > virtio device interface.
> > 
> > That's non-trivial.  Mistakes are possible when designing the mapping.
> > Using the vhost-user protocol as the device interface minimizes the 
> > effort and risk of mistakes because most messages are relayed 1:1.
> > 
> > > 2) some messages can be directly answered by QemuSlave , and some
> > messages are not useful to give to the GuestSlave (inside the VM), 
> > e.g. fds, VhostUserMemoryRegion from SET_MEM_TABLE msg (the device 
> > first maps the master memory and gives the offset (in terms of the 
> > bar, i.e., where does it sit in the bar) of the mapped gpa to the 
> > guest. if we give the raw VhostUserMemoryRegion to the guest, that wouldn’t 
> > be usable).
> > 
> > I agree that QEMU has to handle some of messages, but it should still 
> > relay all (possibly modified) messages to the guest.
> > 
> > The point of using the vhost-user protocol is not just to use a 
> > familiar binary encoding, it's to match the semantics of vhost-user 
> > 100%.  That way the vhost-user software stack can work either in host 
> > userspace or with vhost-pci without significant changes.
> > 
> > Using the vhost-user protocol as the device interface doesn't seem any 
> > harder than defining a completely new virtio device interface.  It has 
> > the advantages that I've pointed out:
> > 
> > 1. Simple 1:1 mapping for most that is easy to maintain as the
> >    vhost-user protocol grows.
> > 
> > 2. Compatible with vhost-user so slaves can run in host userspace
> >    or the guest.
> > 
> > I don't see why it makes sense to define new device interfaces for 
> > each device type and create a software stack that is incompatible with 
> > vhost-user.
> 
> 
> I think this 1:1 mapping wouldn't be easy:
> 
> 1) We will have 2 Qemu side slaves to achieve this bidirectional relaying, 
> that is, the working model will be 
> - master to slave: Master->QemuSlave1->GuestSlave; and
> - slave to master: GuestSlave->QemuSlave2->Master
> QemuSlave1 and QemuSlave2 can't be the same piece of code, because QemuSlave1 
> needs to do some setup with some messages, and QemuSlave2 is more likely to 
> be a true "relayer" (receive and directly pass on)

I mostly agree with this.  Some messages cannot be passed through.  QEMU
needs to process some messages so that makes it both a slave (on the
host) and a master (to the guest).

> 2) poor re-usability of the QemuSlave and GuestSlave
> We couldn’t reuse much of the QemuSlave handling code for GuestSlave.
> For example, for the VHOST_USER_SET_MEM_TABLE msg, all the QemuSlave handling 
> code (please see the vp_slave_set_mem_table function), won't be used by 
> GuestSlave. On the other hand, GuestSlave needs an implementation to reply 
> back to the QEMU device, and this implementation isn't needed by QemuSlave.
>  If we want to run the same piece of the slave code in both QEMU and guest, 
> then we may need "if (QemuSlave) else" in each msg handling entry to choose 
> the code path for QemuSlave and GuestSlave separately.
> So, ideally we wish to run (reuse) one slave implementation in both QEMU and 
> guest. In practice, we will still need to handle them each case by case, 
> which is no different than maintaining two separate slaves for QEMU and 
> guest, and I'm afraid this would be much more complex.

Are you saying QEMU's vhost-pci code cannot be reused by guest slaves?
If so, I agree and it was not my intention to run the same slave code in
QEMU and the guest.

When I referred to reusing the vhost-user software stack I meant
something else:

1. contrib/libvhost-user/ is a vhost-user slave library.  QEMU itself
does not use it but external programs may use it to avoid reimplementing
vhost-user and vrings.  Currently this code handles the vhost-user
protocol over UNIX domain sockets, but it's possible to add vfio
vhost-pci support.  Programs using libvhost-user would be able to take
advantage of vhost-pci easily (no big changes required).

2. DPDK and other codebases that implement custom vhost-user slaves are
also easy to update for vhost-pci since the same protocol is used.  Only
the lowest layer of vhost-user slave code needs to be touched.

Stefan

signature.asc
Description: PGP signature