[PATCH 2/2] KVM: Add documentation for KVM_ARM_SUITABLE_TARGET ioctl

Wed Sep 11 15:15:25 EDT 2013

On Thu, Sep 12, 2013 at 12:12:55AM +0530, Anup Patel wrote:
> On Wed, Sep 11, 2013 at 11:39 PM, Christoffer Dall
> <christoffer.dall at linaro.org> wrote:
> > On Wed, Sep 11, 2013 at 06:29:53PM +0530, Anup Patel wrote:
> >> To implement CPU=Host we have added KVM_ARM_SUITABLE_TARGET ioctl
> >> which provides a CPU target type to user space for creating VCPU
> >> matching underlying Host.
> >>
> >> This patch adds info related to this new KVM_ARM_SUITABLE_TARGET
> >> ioctl in the KVM API documentation.
> >>
> >> Signed-off-by: Anup Patel <anup.patel at linaro.org>
> >> Signed-off-by: Pranavkumar Sawargaonkar <pranavkumar at linaro.org>+
> >> ---
> >>  Documentation/virtual/kvm/api.txt |   25 ++++++++++++++++++++-----
> >>  1 file changed, 20 insertions(+), 5 deletions(-)
> >>
> >> diff --git a/Documentation/virtual/kvm/api.txt b/Documentation/virtual/kvm/api.txt
> >> index ef925ea..1ae9721 100644
> >> --- a/Documentation/virtual/kvm/api.txt
> >> +++ b/Documentation/virtual/kvm/api.txt
> >> @@ -2283,7 +2283,7 @@ current state.  "addr" is ignored.
> >>  Capability: basic
> >>  Architectures: arm, arm64
> >>  Type: vcpu ioctl
> >> -Parameters: struct struct kvm_vcpu_init (in)
> >> +Parameters: struct kvm_vcpu_init (in)
> >>  Returns: 0 on success; -1 on error
> >>  Errors:
> >>    EINVAL:    the target is unknown, or the combination of features is invalid.
> >> @@ -2303,8 +2303,24 @@ Possible features:
> >>       - KVM_ARM_VCPU_EL1_32BIT: Starts the CPU in a 32bit mode.
> >>         Depends on KVM_CAP_ARM_EL1_32BIT (arm64 only).
> >
> > unrelated change, can you put in separate patch?
> 
> Ok
> 
> >
> >>
> >> +4.83 KVM_ARM_SUITABLE_TARGET
> >
> > hmm, I'm a little uneasy about the use of the word 'suitable'.
> > KVM_ARM_PREFERRED_TARGET ?
> 
> Sure, I will rename it to KVM_ARM_PREFERRED_TARGET.
> 
> >
> >>
> >> -4.83 KVM_GET_REG_LIST
> >> +Capability: basic
> >> +Architectures: arm, arm64
> >> +Type: vcpu ioctl
> >
> > this should be a system ioctl (or at least a VM ioctl) so we can query
> > the preferred vcpu without creating one.
> 
> I was really confused on this one because the ioctl is all about VCPU
> target type but I am fine calling it system (or VM) ioctl.
> 

the 'Type' goes to which type of file descriptor the ioctl is called on.
After I wrote the comment I realized you don't technically need the type
of the CPU before creating a vcpu fd, but I think it's nicer this way,
so user space can just tell at the earliest state what this particular
KVM instance prefers.

> >
> >> +Parameters: None
> >> +Returns: 0 on success; -1 on error
> >
> > no parameters and return 0 on success?
> 
> Typo from my-side.
> It returns positive value upon success.
> 
> >
> > I would suggest returning struct kvm_vcpu_init (out) instead.  It makes
> > for a more flexible API in case we want to say something about the
> > preferred/supported vcpu features as well in the future.
> 
> I started-off with an implementation which returned struct kvm_vcpu_init
> but there will be features such POWER_OFF which be only decided by
> user space.
> 
> For e.g. SMP Guest might start VCPU0 in POWER_ON whereas other
> VCPUs in POWER_OFF. We cannot suggest anything about POWER
> state of VCPU to user space from KVM because we really dont know
> what machine model is emulated by user space.
> 
> IMHO, the only thing that seems worth returning is the VCPU target
> type hence this implementation.
> 

I agree as things are looking right now, but I wonder if we would later
add cpu features like an emulated SMC environment and we want to be able
to return that information without having to come up with yet another
ioctl.  Are there any downsides in returning the kvm_vcpu_init struct
and just specify here in the docs that the features field does not
return anything useful (yet)?

> >
> >> +Errors:
> >> +  EINVAL:    no suitable target available for the host
> >
> > EINVAL is not really a proper error code for this IMHO, because it
> > indicates the user supplied something incorrect.  How about ENODEV?
> 
> Ok, I can change this.
> 
> I kept EINVAL because I did not want to make changes in
> kvm_target_cpu().
> 

ah, ok, yeah, the EINVAL tends to be too overloaded inside the kernel
and I guess I fell into that trap as well.  If it doesn't make the code
too ugly, I think another error would be bette for a user space API.

Should be fixable with something like

return (ret == -EINVAL) ? -ENODEV : ret;

> >
> >> +
> >> +This queries KVM for suitable CPU target type which can be emulated by
> >> +KVM on underlying host. This is not a mandatory API and could be used
> >> +to create VCPUs matching underlying host.
> >
> > this last sentece caused me to trip.  Isn't the second part of the
> > sentence (after the 'and') what you are trying to say in the first
> > sentece, and what you want to say here is that it is not required to
> > call this ioctl before calling KVM_ARM_VCPU_INIT?  I'm not sure this
> > information is required at all, but it doesn't hurt I guess.
> 
> Ok, let me try to re-phrase this.
> 
> >
> >> +
> >> +The ioctl returns a target type which can be directly passed-back to
> >
> > s/passed-back/passed back/
> > or consider revising it to "... which can be used for the target type in
> > the KVM_ARM_VCPU_INIT ioctl."
> 
> Ok
> 
> >
> >> +the KVM_ARM_VCPU_INIT ioctl.
> >> +
> >> +4.84 KVM_GET_REG_LIST
> >>
> >>  Capability: basic
> >>  Architectures: arm, arm64
> >> @@ -2323,8 +2339,7 @@ struct kvm_reg_list {
> >>  This ioctl returns the guest registers that are supported for the
> >>  KVM_GET_ONE_REG/KVM_SET_ONE_REG calls.
> >>
> >> -
> >> -4.84 KVM_ARM_SET_DEVICE_ADDR
> >> +4.85 KVM_ARM_SET_DEVICE_ADDR
> >>
> >>  Capability: KVM_CAP_ARM_SET_DEVICE_ADDR
> >>  Architectures: arm, arm64
> >> @@ -2362,7 +2377,7 @@ must be called after calling KVM_CREATE_IRQCHIP, but before calling
> >>  KVM_RUN on any of the VCPUs.  Calling this ioctl twice for any of the
> >>  base addresses will return -EEXIST.
> >>
> >> -4.85 KVM_PPC_RTAS_DEFINE_TOKEN
> >> +4.86 KVM_PPC_RTAS_DEFINE_TOKEN
> >>
> >>  Capability: KVM_CAP_PPC_RTAS
> >>  Architectures: ppc
> >> --
> >> 1.7.9.5
> >>
> > _______________________________________________
> > kvmarm mailing list
> > kvmarm at lists.cs.columbia.edu
> > https://lists.cs.columbia.edu/cucslists/listinfo/kvmarm