| Message ID | 20220124102417.3741427-3-npiggin@gmail.com (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | KVM: PPC: Book3S PR: SCV fixes | expand |
| Context | Check | Description |
|---|---|---|
| snowpatch_ozlabs/github-powerpc_ppctests | success | Successfully ran 8 jobs. |
| snowpatch_ozlabs/github-powerpc_selftests | success | Successfully ran 8 jobs. |
| snowpatch_ozlabs/github-powerpc_sparse | success | Successfully ran 4 jobs. |
| snowpatch_ozlabs/github-powerpc_kernel_qemu | success | Successfully ran 24 jobs. |
| snowpatch_ozlabs/github-powerpc_clang | success | Successfully ran 7 jobs. |
Nicholas Piggin <npiggin@gmail.com> writes: > KVM PR does not implement address translation modes on interrupt, so it > must not allow H_SET_MODE to succeed. > > This is not compatible with QEMU behaviour. The solution might be to > have a cap-ail for this, but now it's broken either way so fix it in > KVM to start with. > > This allows PR Linux guests that are using the SCV facility to boot and > run, because Linux disables the use of SCV if AIL can not be set to 3. > This isn't a real fix because Linux or another OS could implement real > mode SCV vectors and try to enable it. The right solution is for KVM to > emulate scv interrupts from the facility unavailable interrupt. > > Signed-off-by: Nicholas Piggin <npiggin@gmail.com> > --- Reviewed-by: Fabiano Rosas <farosas@linux.ibm.com> > arch/powerpc/kvm/book3s_pr_papr.c | 20 ++++++++++++++++++++ > 1 file changed, 20 insertions(+) > > diff --git a/arch/powerpc/kvm/book3s_pr_papr.c b/arch/powerpc/kvm/book3s_pr_papr.c > index 1f10e7dfcdd0..dc4f51ac84bc 100644 > --- a/arch/powerpc/kvm/book3s_pr_papr.c > +++ b/arch/powerpc/kvm/book3s_pr_papr.c > @@ -281,6 +281,22 @@ static int kvmppc_h_pr_logical_ci_store(struct kvm_vcpu *vcpu) > return EMULATE_DONE; > } > > +static int kvmppc_h_pr_set_mode(struct kvm_vcpu *vcpu) > +{ > + unsigned long mflags = kvmppc_get_gpr(vcpu, 4); > + unsigned long resource = kvmppc_get_gpr(vcpu, 5); > + > + if (resource == H_SET_MODE_RESOURCE_ADDR_TRANS_MODE) { > + /* KVM PR does not provide AIL!=0 to guests */ > + if (mflags == 0) > + kvmppc_set_gpr(vcpu, 3, H_SUCCESS); > + else > + kvmppc_set_gpr(vcpu, 3, H_UNSUPPORTED_FLAG_START - 63); > + return EMULATE_DONE; > + } > + return EMULATE_FAIL; > +} > + > #ifdef CONFIG_SPAPR_TCE_IOMMU > static int kvmppc_h_pr_put_tce(struct kvm_vcpu *vcpu) > { > @@ -384,6 +400,8 @@ int kvmppc_h_pr(struct kvm_vcpu *vcpu, unsigned long cmd) > return kvmppc_h_pr_logical_ci_load(vcpu); > case H_LOGICAL_CI_STORE: > return kvmppc_h_pr_logical_ci_store(vcpu); > + case H_SET_MODE: > + return kvmppc_h_pr_set_mode(vcpu); > case H_XIRR: > case H_CPPR: > case H_EOI: > @@ -421,6 +439,7 @@ int kvmppc_hcall_impl_pr(unsigned long cmd) > case H_CEDE: > case H_LOGICAL_CI_LOAD: > case H_LOGICAL_CI_STORE: > + case H_SET_MODE: > #ifdef CONFIG_KVM_XICS > case H_XIRR: > case H_CPPR: > @@ -447,6 +466,7 @@ static unsigned int default_hcall_list[] = { > H_BULK_REMOVE, > H_PUT_TCE, > H_CEDE, > + H_SET_MODE, > #ifdef CONFIG_KVM_XICS > H_XIRR, > H_CPPR,
diff --git a/arch/powerpc/kvm/book3s_pr_papr.c b/arch/powerpc/kvm/book3s_pr_papr.c index 1f10e7dfcdd0..dc4f51ac84bc 100644 --- a/arch/powerpc/kvm/book3s_pr_papr.c +++ b/arch/powerpc/kvm/book3s_pr_papr.c @@ -281,6 +281,22 @@ static int kvmppc_h_pr_logical_ci_store(struct kvm_vcpu *vcpu) return EMULATE_DONE; } +static int kvmppc_h_pr_set_mode(struct kvm_vcpu *vcpu) +{ + unsigned long mflags = kvmppc_get_gpr(vcpu, 4); + unsigned long resource = kvmppc_get_gpr(vcpu, 5); + + if (resource == H_SET_MODE_RESOURCE_ADDR_TRANS_MODE) { + /* KVM PR does not provide AIL!=0 to guests */ + if (mflags == 0) + kvmppc_set_gpr(vcpu, 3, H_SUCCESS); + else + kvmppc_set_gpr(vcpu, 3, H_UNSUPPORTED_FLAG_START - 63); + return EMULATE_DONE; + } + return EMULATE_FAIL; +} + #ifdef CONFIG_SPAPR_TCE_IOMMU static int kvmppc_h_pr_put_tce(struct kvm_vcpu *vcpu) { @@ -384,6 +400,8 @@ int kvmppc_h_pr(struct kvm_vcpu *vcpu, unsigned long cmd) return kvmppc_h_pr_logical_ci_load(vcpu); case H_LOGICAL_CI_STORE: return kvmppc_h_pr_logical_ci_store(vcpu); + case H_SET_MODE: + return kvmppc_h_pr_set_mode(vcpu); case H_XIRR: case H_CPPR: case H_EOI: @@ -421,6 +439,7 @@ int kvmppc_hcall_impl_pr(unsigned long cmd) case H_CEDE: case H_LOGICAL_CI_LOAD: case H_LOGICAL_CI_STORE: + case H_SET_MODE: #ifdef CONFIG_KVM_XICS case H_XIRR: case H_CPPR: @@ -447,6 +466,7 @@ static unsigned int default_hcall_list[] = { H_BULK_REMOVE, H_PUT_TCE, H_CEDE, + H_SET_MODE, #ifdef CONFIG_KVM_XICS H_XIRR, H_CPPR,
KVM PR does not implement address translation modes on interrupt, so it must not allow H_SET_MODE to succeed. This is not compatible with QEMU behaviour. The solution might be to have a cap-ail for this, but now it's broken either way so fix it in KVM to start with. This allows PR Linux guests that are using the SCV facility to boot and run, because Linux disables the use of SCV if AIL can not be set to 3. This isn't a real fix because Linux or another OS could implement real mode SCV vectors and try to enable it. The right solution is for KVM to emulate scv interrupts from the facility unavailable interrupt. Signed-off-by: Nicholas Piggin <npiggin@gmail.com> --- arch/powerpc/kvm/book3s_pr_papr.c | 20 ++++++++++++++++++++ 1 file changed, 20 insertions(+)