| Message ID | 1458526442-2125-1-git-send-email-aik@ozlabs.ru (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Mon, Mar 21, 2016 at 01:14:02PM +1100, Alexey Kardashevskiy wrote: > ePAPR defines "hcall-instructions" device-tree property which contains > code to call hypercalls in ePAPR paravirtualized guests. In general > pseries guests should not be using this facility (as there is sPAPR > interface) and this property should not be present in the device tree > for pseries guests. > > However a while ago this interface was chosen to implement a hypervisor > interface to speed up PR KVM guests and since then the property is always > present in the device tree. All KVM guests use it at least to read features > via the KVM_HC_FEATURES hypercall. > > The property is populated by the code returned from the KVM's > KVM_PPC_GET_PVINFO ioctl; if not implemented in the KVM, QEMU writes there > "li r3, -1; nop; nop; nop". If QEMU does not create the property, and > the guest kernel is compiled with CONFIG_EPAPR_PARAVIRT (which is > normally the case), there is exactly the same stub at > @epapr_hypercall_start already. > > Rather than maintaining the property (which used to be BE only; then was > fixed to be endian-agnostic) and confusing the guest (which might think > there is ePAPR host while there is none), it makes sense not to create > the property in the device tree in the first place if the host kernel > does not implement it. > > This changes kvmppc_get_hypercall() to return 1 if the host kernel > does not implement KVM_CAP_PPC_GET_PVINFO. The caller can use it to decide > on whether to create the property or not. > > This changes the pseries machine to not create the property if KVM does > not implement KVM_PPC_GET_PVINFO. In practice this means that from now > on the property will not be created if either HV KVM or TCG is used. > > Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> Further reworded the commit message and merged to ppc-for-2.6.
diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c index d0bb423..d43d6d9 100644 --- a/hw/ppc/spapr.c +++ b/hw/ppc/spapr.c @@ -497,10 +497,11 @@ static void *spapr_create_fdt_skel(hwaddr initrd_base, * Older KVM versions with older guest kernels were broken with the * magic page, don't allow the guest to map it. */ - kvmppc_get_hypercall(first_cpu->env_ptr, hypercall, - sizeof(hypercall)); - _FDT((fdt_property(fdt, "hcall-instructions", hypercall, - sizeof(hypercall)))); + if (!kvmppc_get_hypercall(first_cpu->env_ptr, hypercall, + sizeof(hypercall))) { + _FDT((fdt_property(fdt, "hcall-instructions", hypercall, + sizeof(hypercall)))); + } } _FDT((fdt_end_node(fdt))); } diff --git a/target-ppc/kvm.c b/target-ppc/kvm.c index 776336b..e5183db 100644 --- a/target-ppc/kvm.c +++ b/target-ppc/kvm.c @@ -2001,7 +2001,7 @@ int kvmppc_get_hypercall(CPUPPCState *env, uint8_t *buf, int buf_len) hc[2] = cpu_to_be32(0x48000008); hc[3] = cpu_to_be32(bswap32(0x3860ffff)); - return 0; + return 1; } static inline int kvmppc_enable_hcall(KVMState *s, target_ulong hcall)
ePAPR defines "hcall-instructions" device-tree property which contains code to call hypercalls in ePAPR paravirtualized guests. In general pseries guests should not be using this facility (as there is sPAPR interface) and this property should not be present in the device tree for pseries guests. However a while ago this interface was chosen to implement a hypervisor interface to speed up PR KVM guests and since then the property is always present in the device tree. All KVM guests use it at least to read features via the KVM_HC_FEATURES hypercall. The property is populated by the code returned from the KVM's KVM_PPC_GET_PVINFO ioctl; if not implemented in the KVM, QEMU writes there "li r3, -1; nop; nop; nop". If QEMU does not create the property, and the guest kernel is compiled with CONFIG_EPAPR_PARAVIRT (which is normally the case), there is exactly the same stub at @epapr_hypercall_start already. Rather than maintaining the property (which used to be BE only; then was fixed to be endian-agnostic) and confusing the guest (which might think there is ePAPR host while there is none), it makes sense not to create the property in the device tree in the first place if the host kernel does not implement it. This changes kvmppc_get_hypercall() to return 1 if the host kernel does not implement KVM_CAP_PPC_GET_PVINFO. The caller can use it to decide on whether to create the property or not. This changes the pseries machine to not create the property if KVM does not implement KVM_PPC_GET_PVINFO. In practice this means that from now on the property will not be created if either HV KVM or TCG is used. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> --- Changes: v2: * changed the commit log --- hw/ppc/spapr.c | 9 +++++---- target-ppc/kvm.c | 2 +- 2 files changed, 6 insertions(+), 5 deletions(-)