i386: disable PMU CPUID leaf by default
cover i386: disable PMU CPUID leaf by default
Changes v1 -> v2:
* compat property is now named "pmu" instead of "pmu-passthrough"
Eduardo Habkost (2):
i386: pass X86CPU object to cpu_x86_find_by_name()
i386: disable PMU CPUID leaves by default
include/hw/i386/pc.h | 4 ++++ target-i386/cpu-qom.h | 7 +++++++ target-i386/cpu.c | 16 +++++++++++++--- 3 files changed, 24 insertions(+), 3 deletions(-)
Signature · 1 line
1.8.3.1
PATCH v2 1/2 i386: pass X86CPU object to cpu_x86_find_by_name()
This will help us change the initialization code to not require carrying some intermediate values in a x86_def_t struct (and eventually kill the x86_def_t struct entirely).
--- target-i386/cpu.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)
target-i386/cpu.c +3 -2
@@ -1475,7 +1475,8 @@ static void x86_cpu_get_feature_words(Object *obj, Visitor *v, void *opaque,error_propagate (errp, err); } -static int cpu_x86_find_by_name (x86_def_t *x86_cpu_def,const char *name) +static int cpu_x86_find_by_name (X86CPU *cpu,x86_def_t *x86_cpu_def, +const char *name) {x86_def_t *def;int i;
@@ -1742,7 +1743,7 @@ static void cpu_x86_register(X86CPU *cpu, const char *name, Error **errp)memset (def,0 ,sizeof (*def)); -if (cpu_x86_find_by_name (def, name) <0 ) { +if (cpu_x86_find_by_name (cpu, def, name) <0 ) {error_setg (errp,"Unable to find CPU definition: %s" , name);return ; }
Signature · 1 line
1.8.3.1
PATCH v2 2/2 i386: disable PMU CPUID leaf by default
In reply to Eduardo HabkostBug description: QEMU currently gets all bits from GET_SUPPORTED_CPUID for CPUID leaf 0xA and passes them directly to the guest. This makes the guest ABI depend on host kernel and host CPU capabilities, and breaks live migration if we migrate between host with different capabilities (e.g. different number of PMU counters).
This patch adds a "pmu" property to X86CPU, and set it to true only on "-cpu host", or on pc-*-1.5 and older machine-types.
By now, setting pmu=on will enable the current passthrough mode that doesn't have any ABI stability guarantees, but in the future we may implement a mode where the PMU CPUID bits are stable and configurable.
--- include/hw/i386/pc.h | 4 ++++ target-i386/cpu-qom.h | 7 +++++++ target-i386/cpu.c | 11 ++++++++++- 3 files changed, 21 insertions(+), 1 deletion(-)
include/hw/i386/pc.h +4 -0
@@ -235,6 +235,10 @@ int e820_add_entry(uint64_t, uint64_t, uint32_t); .driver = "virtio-net-pci",\ .property = "any_layout",\ .value = "off",\ + },{\ + .driver = TYPE_X86_CPU,\ + .property = "pmu",\ + .value = "on",\ }#define PC_COMPAT_1_4 \
target-i386/cpu-qom.h +7 -0
@@ -68,6 +68,13 @@ typedef struct X86CPU {/* Features that were filtered out because of missing host capabilities */ uint32_t filtered_features[FEATURE_WORDS ]; + +/* Enable PMU CPUID bits. This can't be enabled by default yet because +* it doesn't have ABI stability guarantees, as it passes all PMU CPUID +* bits returned by GET_SUPPORTED_CPUID (that depend on host CPU and kernel +* capabilities) directly to the guest. +*/ +bool enable_pmu; }X86CPU ;static inline X86CPU *x86_env_get_cpu (CPUX86State *env)
target-i386/cpu.c +10 -1
@@ -1475,17 +1475,25 @@ static void x86_cpu_get_feature_words(Object *obj, Visitor *v, void *opaque,error_propagate (errp, err); } +static Property cpu_x86_properties[] = { +DEFINE_PROP_BOOL ("pmu" ,X86CPU , enable_pmu, false), +DEFINE_PROP_END_OF_LIST (), +}; +static int cpu_x86_find_by_name (X86CPU *cpu,x86_def_t *x86_cpu_def,const char *name) {x86_def_t *def;int i; +Error *err =NULL ;if (name ==NULL ) {return -1 ; }if (kvm_enabled () &&strcmp (name,"host" ) ==0 ) {kvm_cpu_fill_host (x86_cpu_def); +object_property_set_bool (OBJECT (cpu), true,"pmu" , &err); +assert_no_error (err);return 0 ; }
@@ -2017,7 +2025,7 @@ void cpu_x86_cpuid(CPUX86State *env, uint32_t index, uint32_t count,break ;case 0xA :/* Architectural Performance Monitoring Leaf */ -if (kvm_enabled ()) { +if (kvm_enabled () && cpu->enable_pmu) {KVMState *s = cs->kvm_state; *eax =kvm_arch_get_supported_cpuid (s,0xA , count,R_EAX );
@@ -2530,6 +2538,7 @@ static void x86_cpu_common_class_init(ObjectClass *oc, void *data) xcc->parent_realize = dc->realize; dc->realize = x86_cpu_realizefn; dc->bus_type =TYPE_ICC_BUS ; + dc->props = cpu_x86_properties; xcc->parent_reset = cc->reset; cc->reset = x86_cpu_reset;
Signature · 1 line
1.8.3.1
Am 26.07.2013 22:09, schrieb Eduardo Habkost:
Quoted 11 lines
Changes v1 -> v2:
* compat property is now named "pmu" instead of "pmu-passthrough"Eduardo Habkost (2):
i386: pass X86CPU object tocpu_x86_find_by_name()
i386: disable PMU CPUID leaves by defaultinclude/hw/i386/pc.h | 4 ++++ target-i386/cpu-qom.h | 7 +++++++ target-i386/cpu.c | 16 +++++++++++++--- 3 files changed, 24 insertions(+), 3 deletions(-)
Technically the patch looks okay for 1.6. Waiting for functional acks.
Andreas
Signature · 2 lines
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg
Am 26.07.2013 22:15, schrieb Andreas Färber:
Quoted 8 lines
Am 26.07.2013 22:09, schrieb Eduardo Habkost:
Quoted 7 lines
Changes v1 -> v2:
* compat property is now named "pmu" instead of "pmu-passthrough"Eduardo Habkost (2):
i386: pass X86CPU object tocpu_x86_find_by_name()
i386: disable PMU CPUID leaves by default
Quoted 6 lines
Quoted 4 lines
include/hw/i386/pc.h | 4 ++++ target-i386/cpu-qom.h | 7 +++++++ target-i386/cpu.c | 16 +++++++++++++--- 3 files changed, 24 insertions(+), 3 deletions(-)Technically the patch looks okay for 1.6. Waiting for functional acks.
Since time is running out and the name requested by Paolo was adopted, I have queued this on qom-cpu with minor adjustments, most notably moving and renaming the property array:
https://github.com/afaerber/qemu-cpu/commits/qom-cpu
Paolo, if there's still corrections necessary, let me know whether to unqueue or whether we can follow-up during Hard Freeze.
Thanks,
Andreas
Signature · 2 lines
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg