On Wed, Sep 28, 2016 at 01:16:30PM +0200, Thomas Huth wrote: > KVM-PR currently does not support transactional memory, and the > implementation in TCG is just a fake. We should not announce TM > support in the ibm,pa-features property when running on such a > system, so disable it by default and only enable it if the KVM > implementation supports it (i.e. recent versions of KVM-HV). > These changes are based on some earlier work from Anton Blanchard > (thanks!). > > Signed-off-by: Thomas Huth <th...@redhat.com>
So, I've applied this, but I do have one concern. KVM HV has, IIUC,
supported TM in guests since basically forever, but the
KVM_CAP_PPC_HTM tag is pretty new. So this change will break TM for
guests on KVM HV older than the capability flag.
So, I think we want a fallback which will set cap_htm if we detect KVM
HV using the usual KVM_CAP_PPC_GET_PVINFO hack.
I think we can do that as a follow up patch, but we definitely should
implement it before 2.8 is released.
> ---
> hw/ppc/spapr.c | 5 ++++-
> target-ppc/kvm.c | 7 +++++++
> target-ppc/kvm_ppc.h | 6 ++++++
> 3 files changed, 17 insertions(+), 1 deletion(-)
>
> diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> index 46d6b90..0bdea5b 100644
> --- a/hw/ppc/spapr.c
> +++ b/hw/ppc/spapr.c
> @@ -555,7 +555,7 @@ static void spapr_populate_pa_features(CPUPPCState *env,
> void *fdt, int offset)
> 0xf6, 0x1f, 0xc7, 0xc0, 0x80, 0xf0,
> 0x80, 0x00, 0x00, 0x00, 0x00, 0x00,
> 0x00, 0x00, 0x00, 0x00, 0x80, 0x00,
> - 0x80, 0x00, 0x80, 0x00, 0x80, 0x00 };
> + 0x80, 0x00, 0x80, 0x00, 0x00, 0x00 };
> uint8_t *pa_features;
> size_t pa_size;
>
> @@ -584,6 +584,9 @@ static void spapr_populate_pa_features(CPUPPCState *env,
> void *fdt, int offset)
> */
> pa_features[3] |= 0x20;
> }
> + if (kvmppc_has_cap_htm()) {
> + pa_features[24] |= 0x80; /* Transactional memory support */
> + }
>
> _FDT((fdt_setprop(fdt, offset, "ibm,pa-features", pa_features,
> pa_size)));
> }
> diff --git a/target-ppc/kvm.c b/target-ppc/kvm.c
> index a18d4d5..e9a9faf 100644
> --- a/target-ppc/kvm.c
> +++ b/target-ppc/kvm.c
> @@ -80,6 +80,7 @@ static int cap_ppc_watchdog;
> static int cap_papr;
> static int cap_htab_fd;
> static int cap_fixup_hcalls;
> +static int cap_htm; /* Hardware transactional memory support */
>
> static uint32_t debug_inst_opcode;
>
> @@ -122,6 +123,7 @@ int kvm_arch_init(MachineState *ms, KVMState *s)
> * only activated after this by kvmppc_set_papr() */
> cap_htab_fd = kvm_check_extension(s, KVM_CAP_PPC_HTAB_FD);
> cap_fixup_hcalls = kvm_check_extension(s, KVM_CAP_PPC_FIXUP_HCALL);
> + cap_htm = kvm_vm_check_extension(s, KVM_CAP_PPC_HTM);
>
> if (!cap_interrupt_level) {
> fprintf(stderr, "KVM: Couldn't find level irq capability. Expect the
> "
> @@ -2353,6 +2355,11 @@ bool kvmppc_has_cap_fixup_hcalls(void)
> return cap_fixup_hcalls;
> }
>
> +bool kvmppc_has_cap_htm(void)
> +{
> + return cap_htm;
> +}
> +
> static PowerPCCPUClass *ppc_cpu_get_family_class(PowerPCCPUClass *pcc)
> {
> ObjectClass *oc = OBJECT_CLASS(pcc);
> diff --git a/target-ppc/kvm_ppc.h b/target-ppc/kvm_ppc.h
> index a778184..bd1d78b 100644
> --- a/target-ppc/kvm_ppc.h
> +++ b/target-ppc/kvm_ppc.h
> @@ -55,6 +55,7 @@ void kvmppc_hash64_free_pteg(uint64_t token);
> void kvmppc_hash64_write_pte(CPUPPCState *env, target_ulong pte_index,
> target_ulong pte0, target_ulong pte1);
> bool kvmppc_has_cap_fixup_hcalls(void);
> +bool kvmppc_has_cap_htm(void);
> int kvmppc_enable_hwrng(void);
> int kvmppc_put_books_sregs(PowerPCCPU *cpu);
> PowerPCCPUClass *kvm_ppc_get_host_cpu_class(void);
> @@ -249,6 +250,11 @@ static inline bool kvmppc_has_cap_fixup_hcalls(void)
> abort();
> }
>
> +static inline bool kvmppc_has_cap_htm(void)
> +{
> + return false;
> +}
> +
> static inline int kvmppc_enable_hwrng(void)
> {
> return -1;
--
David Gibson | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_
| _way_ _around_!
http://www.ozlabs.org/~dgibson
signature.asc
Description: PGP signature
