On 10/26/2012 06:33 PM, KONRAD Frédéric wrote: > On 26/10/2012 10:16, Evgeny Voevodin wrote: >> On 10/26/2012 12:10 PM, KONRAD Frédéric wrote: >>> On 26/10/2012 05:15, Evgeny Voevodin wrote: >>>> On 10/25/2012 09:05 PM, Peter Maydell wrote: >>>>> On 25 October 2012 16:00, Mark Burton <mark.burton@xxxxxxxxxxxxx> >>>>> wrote: >>>>>> Peter, I guess it's easier to review on the GIT, though we should >>>>>> send >>>>>> out the full updated patch set to the QEMU-dev list, once you've had >>>>>> the opportunity to review it on here…. >>>>> Yes, that sounds like a good idea. I haven't had time for a full >>>>> review, but my initial thoughts: >>>>> * not convinced about the whole transport link thing, need to >>>>> think about whether this is the right way to model this in QOM >>>> That was an initial mechanism to connect transport devices. It was >>>> planned to >>>> remove it by QOM <link>, but my hands didn't reach this (and also I >>>> wasn't sure what's the >>>> difference between QOM <link> and transport list I used.) )) >>> Is there any documentation about that ? >> >> You can ask Anthony about QOM links. Also you can look at qemu's bus >> implementation in qdev.c (he said that )) >> > Ok thanks, I'll look into qdev.c > >> >>> >>>> >>>>> * the patches which create virtio-pci-new.c as a modified copy >>>>> of virtio-pci.c and then at the end copy it over into >>>>> virtio-pci.c >>>>> are basically unreviewable. We need to find a better way of >>>>> arranging these changes into patches. >>>> I agree. Any future changes in virtio-pci.c will force you to update >>>> virtio-pci-new.c. And nobody >>>> can be sure that you did it correctly until compare two files. >>>> >>> Yes, sure I'll look at it. >>> > > If I compare virtio-pci.c and virtio-pci-new.c the patch : > - replace VirtioPCIProxy to VirtioPCI I'm not sure that we must replace VirtioPCIProxy by VirtioPC. Maybe we just need to modify VirtioPCIProxy itself. But to me it looks more nice to have VirtioPCI. > - add the virtio-pci device. > - add the virtio_pci_transport_cb(..). > > I think there are two problems for reviewing : the creation of the > virtio-pci-new.c, and the addition of device specific code ( for > example : the switch case in virtio_pci_transport_cb(..) and in the > exit function ). > > So what I can do is modifying directly the virtio-pci.c file ( with no > virtio-pci-new.c file ) Sure, that's what must be done. > , and then for each virtio-x-pci in separate patch : > - remove the old virtio-x-pci device. ( it is what is done for the > moment ) > - add the case in the virtio_pci_transport_cb(..). > - add the code in the exit function. > It will be easier to compare the old code and the new code as it will > be in the same patch. > > What do you think ? > > Thanks, > > Fred. > -- Kind regards, Evgeny Voevodin, Technical Leader, Mobile Group, Samsung Moscow Research Center, e-mail: e.voevodin@xxxxxxxxxxx _______________________________________________ kvmarm mailing list kvmarm@xxxxxxxxxxxxxxxxxxxxx https://lists.cs.columbia.edu/cucslists/listinfo/kvmarm
