All of lore.kernel.org
 help / color / mirror / Atom feed
* [Qemu-devel] [PATCH 0/2] ppc/kvm: Always allow transactional memory on POWER8 KVM-HV
@ 2016-09-29 10:48 Thomas Huth
  2016-09-29 10:48 ` [Qemu-devel] [PATCH 1/2] target-ppc/kvm: Add a wrapper function to check for KVM-PR Thomas Huth
  2016-09-29 10:48 ` [Qemu-devel] [PATCH 2/2] target-ppc/kvm: Enable transactional memory on POWER8 with KVM-HV, too Thomas Huth
  0 siblings, 2 replies; 5+ messages in thread
From: Thomas Huth @ 2016-09-29 10:48 UTC (permalink / raw)
  To: David Gibson, qemu-ppc; +Cc: Alexander Graf, qemu-devel

KVM-HV also supports transactional memory on POWER8 with older
kernels that do not support the KVM_CAP_PPC_HTM extension yet.
The first patch introduces a proper function for checking whether
we're running on KVM-PR or KVM-HV, which is then used in the
second patch to check that we're running on POWER8 with KVM-HV,
so we can allow TM there, too.

Thomas Huth (2):
  target-ppc/kvm: Add a wrapper function to check for KVM-PR
  target-ppc/kvm: Enable transactional memory on POWER8 with KVM-HV, too

 hw/ppc/spapr.c   |  2 +-
 target-ppc/kvm.c | 32 +++++++++++++++++++++-----------
 2 files changed, 22 insertions(+), 12 deletions(-)

-- 
1.8.3.1

^ permalink raw reply	[flat|nested] 5+ messages in thread

* [Qemu-devel] [PATCH 1/2] target-ppc/kvm: Add a wrapper function to check for KVM-PR
  2016-09-29 10:48 [Qemu-devel] [PATCH 0/2] ppc/kvm: Always allow transactional memory on POWER8 KVM-HV Thomas Huth
@ 2016-09-29 10:48 ` Thomas Huth
  2016-09-29 23:23   ` David Gibson
  2016-09-29 10:48 ` [Qemu-devel] [PATCH 2/2] target-ppc/kvm: Enable transactional memory on POWER8 with KVM-HV, too Thomas Huth
  1 sibling, 1 reply; 5+ messages in thread
From: Thomas Huth @ 2016-09-29 10:48 UTC (permalink / raw)
  To: David Gibson, qemu-ppc; +Cc: Alexander Graf, qemu-devel

It makes more sense if we have a proper function to check
for KVM-PR than to check for the GET_PVINFO extension all
over the place.

Signed-off-by: Thomas Huth <thuth@redhat.com>
---
 target-ppc/kvm.c | 23 +++++++++++++----------
 1 file changed, 13 insertions(+), 10 deletions(-)

diff --git a/target-ppc/kvm.c b/target-ppc/kvm.c
index e9a9faf..dca50bc 100644
--- a/target-ppc/kvm.c
+++ b/target-ppc/kvm.c
@@ -102,6 +102,13 @@ static void kvm_kick_cpu(void *opaque)
     qemu_cpu_kick(CPU(cpu));
 }
 
+/* Check whether we are running with KVM-PR (instead of KVM-HV) */
+static bool kvmppc_is_pr(KVMState *ks)
+{
+    /* Assume KVM-PR if the GET_PVINFO capability is available */
+    return kvm_check_extension(ks, KVM_CAP_PPC_GET_PVINFO) != 0;
+}
+
 static int kvm_ppc_register_host_cpu_type(void);
 
 int kvm_arch_init(MachineState *ms, KVMState *s)
@@ -223,10 +230,9 @@ static void kvm_get_fallback_smmu_info(PowerPCCPU *cpu,
      *
      * For that to work we make a few assumptions:
      *
-     * - If KVM_CAP_PPC_GET_PVINFO is supported we are running "PR"
-     *   KVM which only supports 4K and 16M pages, but supports them
-     *   regardless of the backing store characteritics. We also don't
-     *   support 1T segments.
+     * - Check whether we are running "PR" KVM which only supports 4K
+     *   and 16M pages, but supports them regardless of the backing
+     *   store characteritics. We also don't support 1T segments.
      *
      *   This is safe as if HV KVM ever supports that capability or PR
      *   KVM grows supports for more page/segment sizes, those versions
@@ -241,7 +247,7 @@ static void kvm_get_fallback_smmu_info(PowerPCCPU *cpu,
      *   implements KVM_CAP_PPC_GET_SMMU_INFO and thus doesn't hit
      *   this fallback.
      */
-    if (kvm_check_extension(cs->kvm_state, KVM_CAP_PPC_GET_PVINFO)) {
+    if (kvmppc_is_pr(cs->kvm_state)) {
         /* No flags */
         info->flags = 0;
         info->slb_size = 64;
@@ -2270,11 +2276,8 @@ int kvmppc_reset_htab(int shift_hint)
 
     /* We have a kernel that predates the htab reset calls.  For PR
      * KVM, we need to allocate the htab ourselves, for an HV KVM of
-     * this era, it has allocated a 16MB fixed size hash table
-     * already.  Kernels of this era have the GET_PVINFO capability
-     * only on PR, so we use this hack to determine the right
-     * answer */
-    if (kvm_check_extension(kvm_state, KVM_CAP_PPC_GET_PVINFO)) {
+     * this era, it has allocated a 16MB fixed size hash table already. */
+    if (kvmppc_is_pr(kvm_state)) {
         /* PR - tell caller to allocate htab */
         return 0;
     } else {
-- 
1.8.3.1

^ permalink raw reply related	[flat|nested] 5+ messages in thread

* [Qemu-devel] [PATCH 2/2] target-ppc/kvm: Enable transactional memory on POWER8 with KVM-HV, too
  2016-09-29 10:48 [Qemu-devel] [PATCH 0/2] ppc/kvm: Always allow transactional memory on POWER8 KVM-HV Thomas Huth
  2016-09-29 10:48 ` [Qemu-devel] [PATCH 1/2] target-ppc/kvm: Add a wrapper function to check for KVM-PR Thomas Huth
@ 2016-09-29 10:48 ` Thomas Huth
  2016-09-29 23:26   ` David Gibson
  1 sibling, 1 reply; 5+ messages in thread
From: Thomas Huth @ 2016-09-29 10:48 UTC (permalink / raw)
  To: David Gibson, qemu-ppc; +Cc: Alexander Graf, qemu-devel

Transactional memory is also supported on POWER8 KVM-HV if the
KVM_CAP_PPC_HTM is not available in the kernel yet, so add a hack
to allow TM here, too.

Signed-off-by: Thomas Huth <thuth@redhat.com>
---
 hw/ppc/spapr.c   | 2 +-
 target-ppc/kvm.c | 9 ++++++++-
 2 files changed, 9 insertions(+), 2 deletions(-)

diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
index fc37145..2f4e818 100644
--- a/hw/ppc/spapr.c
+++ b/hw/ppc/spapr.c
@@ -584,7 +584,7 @@ static void spapr_populate_pa_features(CPUPPCState *env, void *fdt, int offset)
          */
         pa_features[3] |= 0x20;
     }
-    if (kvmppc_has_cap_htm()) {
+    if (kvmppc_has_cap_htm() && pa_size > 24) {
         pa_features[24] |= 0x80;    /* Transactional memory support */
     }
 
diff --git a/target-ppc/kvm.c b/target-ppc/kvm.c
index dca50bc..e990cb7 100644
--- a/target-ppc/kvm.c
+++ b/target-ppc/kvm.c
@@ -567,11 +567,18 @@ int kvm_arch_init_vcpu(CPUState *cs)
 
     idle_timer = timer_new_ns(QEMU_CLOCK_VIRTUAL, kvm_kick_cpu, cpu);
 
-    /* Some targets support access to KVM's guest TLB. */
     switch (cenv->mmu_model) {
     case POWERPC_MMU_BOOKE206:
+        /* This target supports access to KVM's guest TLB */
         ret = kvm_booke206_tlb_init(cpu);
         break;
+    case POWERPC_MMU_2_07:
+        if (!cap_htm && !kvmppc_is_pr(cs->kvm_state)) {
+            /* KVM-HV has transactional memory on POWER8 also without the
+             * KVM_CAP_PPC_HTM extension, so enable it here instead. */
+            cap_htm = true;
+        }
+        break;
     default:
         break;
     }
-- 
1.8.3.1

^ permalink raw reply related	[flat|nested] 5+ messages in thread

* Re: [Qemu-devel] [PATCH 1/2] target-ppc/kvm: Add a wrapper function to check for KVM-PR
  2016-09-29 10:48 ` [Qemu-devel] [PATCH 1/2] target-ppc/kvm: Add a wrapper function to check for KVM-PR Thomas Huth
@ 2016-09-29 23:23   ` David Gibson
  0 siblings, 0 replies; 5+ messages in thread
From: David Gibson @ 2016-09-29 23:23 UTC (permalink / raw)
  To: Thomas Huth; +Cc: qemu-ppc, Alexander Graf, qemu-devel

[-- Attachment #1: Type: text/plain, Size: 3211 bytes --]

On Thu, Sep 29, 2016 at 12:48:06PM +0200, Thomas Huth wrote:
> It makes more sense if we have a proper function to check
> for KVM-PR than to check for the GET_PVINFO extension all
> over the place.
> 
> Signed-off-by: Thomas Huth <thuth@redhat.com>

Applied to ppc-for-2.8.  I expanded your comment to say this should
only be used for fallback tests, though.

> ---
>  target-ppc/kvm.c | 23 +++++++++++++----------
>  1 file changed, 13 insertions(+), 10 deletions(-)
> 
> diff --git a/target-ppc/kvm.c b/target-ppc/kvm.c
> index e9a9faf..dca50bc 100644
> --- a/target-ppc/kvm.c
> +++ b/target-ppc/kvm.c
> @@ -102,6 +102,13 @@ static void kvm_kick_cpu(void *opaque)
>      qemu_cpu_kick(CPU(cpu));
>  }
>  
> +/* Check whether we are running with KVM-PR (instead of KVM-HV) */
> +static bool kvmppc_is_pr(KVMState *ks)
> +{
> +    /* Assume KVM-PR if the GET_PVINFO capability is available */
> +    return kvm_check_extension(ks, KVM_CAP_PPC_GET_PVINFO) != 0;
> +}
> +
>  static int kvm_ppc_register_host_cpu_type(void);
>  
>  int kvm_arch_init(MachineState *ms, KVMState *s)
> @@ -223,10 +230,9 @@ static void kvm_get_fallback_smmu_info(PowerPCCPU *cpu,
>       *
>       * For that to work we make a few assumptions:
>       *
> -     * - If KVM_CAP_PPC_GET_PVINFO is supported we are running "PR"
> -     *   KVM which only supports 4K and 16M pages, but supports them
> -     *   regardless of the backing store characteritics. We also don't
> -     *   support 1T segments.
> +     * - Check whether we are running "PR" KVM which only supports 4K
> +     *   and 16M pages, but supports them regardless of the backing
> +     *   store characteritics. We also don't support 1T segments.
>       *
>       *   This is safe as if HV KVM ever supports that capability or PR
>       *   KVM grows supports for more page/segment sizes, those versions
> @@ -241,7 +247,7 @@ static void kvm_get_fallback_smmu_info(PowerPCCPU *cpu,
>       *   implements KVM_CAP_PPC_GET_SMMU_INFO and thus doesn't hit
>       *   this fallback.
>       */
> -    if (kvm_check_extension(cs->kvm_state, KVM_CAP_PPC_GET_PVINFO)) {
> +    if (kvmppc_is_pr(cs->kvm_state)) {
>          /* No flags */
>          info->flags = 0;
>          info->slb_size = 64;
> @@ -2270,11 +2276,8 @@ int kvmppc_reset_htab(int shift_hint)
>  
>      /* We have a kernel that predates the htab reset calls.  For PR
>       * KVM, we need to allocate the htab ourselves, for an HV KVM of
> -     * this era, it has allocated a 16MB fixed size hash table
> -     * already.  Kernels of this era have the GET_PVINFO capability
> -     * only on PR, so we use this hack to determine the right
> -     * answer */
> -    if (kvm_check_extension(kvm_state, KVM_CAP_PPC_GET_PVINFO)) {
> +     * this era, it has allocated a 16MB fixed size hash table already. */
> +    if (kvmppc_is_pr(kvm_state)) {
>          /* PR - tell caller to allocate htab */
>          return 0;
>      } else {

-- 
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

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [Qemu-devel] [PATCH 2/2] target-ppc/kvm: Enable transactional memory on POWER8 with KVM-HV, too
  2016-09-29 10:48 ` [Qemu-devel] [PATCH 2/2] target-ppc/kvm: Enable transactional memory on POWER8 with KVM-HV, too Thomas Huth
@ 2016-09-29 23:26   ` David Gibson
  0 siblings, 0 replies; 5+ messages in thread
From: David Gibson @ 2016-09-29 23:26 UTC (permalink / raw)
  To: Thomas Huth; +Cc: qemu-ppc, Alexander Graf, qemu-devel

[-- Attachment #1: Type: text/plain, Size: 2110 bytes --]

On Thu, Sep 29, 2016 at 12:48:07PM +0200, Thomas Huth wrote:
> Transactional memory is also supported on POWER8 KVM-HV if the
> KVM_CAP_PPC_HTM is not available in the kernel yet, so add a hack
> to allow TM here, too.
> 
> Signed-off-by: Thomas Huth <thuth@redhat.com>

Applied to ppc-for-2.8.

> ---
>  hw/ppc/spapr.c   | 2 +-
>  target-ppc/kvm.c | 9 ++++++++-
>  2 files changed, 9 insertions(+), 2 deletions(-)
> 
> diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> index fc37145..2f4e818 100644
> --- a/hw/ppc/spapr.c
> +++ b/hw/ppc/spapr.c
> @@ -584,7 +584,7 @@ static void spapr_populate_pa_features(CPUPPCState *env, void *fdt, int offset)
>           */
>          pa_features[3] |= 0x20;
>      }
> -    if (kvmppc_has_cap_htm()) {
> +    if (kvmppc_has_cap_htm() && pa_size > 24) {

It looks like this is an unrelated fix to the original TM test.  I'll
fold it into the original patch.

>          pa_features[24] |= 0x80;    /* Transactional memory support */
>      }
>  
> diff --git a/target-ppc/kvm.c b/target-ppc/kvm.c
> index dca50bc..e990cb7 100644
> --- a/target-ppc/kvm.c
> +++ b/target-ppc/kvm.c
> @@ -567,11 +567,18 @@ int kvm_arch_init_vcpu(CPUState *cs)
>  
>      idle_timer = timer_new_ns(QEMU_CLOCK_VIRTUAL, kvm_kick_cpu, cpu);
>  
> -    /* Some targets support access to KVM's guest TLB. */
>      switch (cenv->mmu_model) {
>      case POWERPC_MMU_BOOKE206:
> +        /* This target supports access to KVM's guest TLB */
>          ret = kvm_booke206_tlb_init(cpu);
>          break;
> +    case POWERPC_MMU_2_07:
> +        if (!cap_htm && !kvmppc_is_pr(cs->kvm_state)) {
> +            /* KVM-HV has transactional memory on POWER8 also without the
> +             * KVM_CAP_PPC_HTM extension, so enable it here instead. */
> +            cap_htm = true;
> +        }
> +        break;
>      default:
>          break;
>      }

-- 
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

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2016-09-29 23:37 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2016-09-29 10:48 [Qemu-devel] [PATCH 0/2] ppc/kvm: Always allow transactional memory on POWER8 KVM-HV Thomas Huth
2016-09-29 10:48 ` [Qemu-devel] [PATCH 1/2] target-ppc/kvm: Add a wrapper function to check for KVM-PR Thomas Huth
2016-09-29 23:23   ` David Gibson
2016-09-29 10:48 ` [Qemu-devel] [PATCH 2/2] target-ppc/kvm: Enable transactional memory on POWER8 with KVM-HV, too Thomas Huth
2016-09-29 23:26   ` David Gibson

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.