On Mon, Feb 18, 2013 at 05:01:40PM -0600, Scott Wood wrote: > On 02/18/2013 06:21:59 AM, Gleb Natapov wrote: > >Copying Christoffer since ARM has in kernel irq chip too. > > > >On Wed, Feb 13, 2013 at 11:49:15PM -0600, Scott Wood wrote: > >> Currently, devices that are emulated inside KVM are configured in a > >> hardcoded manner based on an assumption that any given architecture > >> only has one way to do it. If there's any need to access device > >state, > >> it is done through inflexible one-purpose-only IOCTLs (e.g. > >> KVM_GET/SET_LAPIC). Defining new IOCTLs for every little thing is > >> cumbersome and depletes a limited numberspace. Creation of ioctl has advantages. It makes explicit what the data contains and how it should be interpreted. This means that for example, policy control can be performed at ioctl level (can group, by ioctl number, which ioctl calls are allowed after VM creation, for example). It also makes it explicit that its a userspace interface which applications depend. With a single 'set device attribute' ioctl its more difficult to do that. Abstracting the details also makes it cumbersome to read strace output :-) > >> This API provides a mechanism to instantiate a device of a certain > >> type, returning an ID that can be used to set/get attributes of the > >> device. Attributes may include configuration parameters (e.g. > >> register base address), device state, operational commands, etc. It > >> is similar to the ONE_REG API, except that it acts on devices rather > >> than vcpus. > >You are not only provide different way to create in kernel irq > >chip you > >also use an alternate way to trigger interrupt lines. Before going > >into > >interface specifics lets think about whether it is really worth it? > > Which "it" do you mean here? > > The ability to set/get attributes is needed. Sorry, but "get or set > one blob of data, up to 512 bytes, for the entire irqchip" is just > not good enough -- assuming you don't want us to start sticking > pointers and commands in *that* data. :-) Why not? Is it necessary to constantly read/write attributes? > If you mean the way to inject interrupts, it's simpler this way. Are you injecting interrupts via this new SET_DEVICE_ATTRIBUTE ioctl? > Why go out of our way to inject common glue code into a > communication path between hw/kvm/mpic.c in QEMU and > arch/powerpc/kvm/mpic.c in KVM? Or rather, why make that common > glue be specific to this one function when we could reuse the same > communication glue used for other things, such as device state? > > And that's just for regular interrupts. MSIs are vastly simpler on > MPIC than what x86 does. > > >x86 obviously support old way and will have to for some, very > >long, time. > > Sure. > > >ARM vGIC code, that is ready to go upstream, uses old way too. So > >it will > >be 2 archs against one. > > I wasn't aware that that's how it worked. :-P > > I was trying to be considerate by not making the entire thing > gratuitously PPC or MPIC specific, as some others seem inclined to > do (e.g. see irqfd and APIC). We already had a discussion on ARM's > "set address" ioctl and rather than extend *that* interface, they > preferred to just stick something ARM-specific in ASAP with the > understanding that it would be replaced (or more accurately, kept > around as a thin wrapper around the new stuff) later. > > >Christoffer do you think the proposed way it > >better for your needs. Are you willing to make vGIC use it? > > > >Scott, what other devices are you planning to support with this > >interface? > > At the moment I do not have plans for other devices, though what > does it hurt for the capability to be there? > > -Scott -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
