| Message ID | 20170217073625.13418-3-jgross@suse.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 02/17/2017 02:36 AM, Juergen Gross wrote: > When running as pv domain xen_cpuid() is being used instead of > native_cpuid(). In xen_cpuid() the aperf/mperf feature is indicated > as not being present by special casing the related cpuid leaf. > > Instead of delivering fake cpuid values clear the cpu capability bit > for aperf/mperf instead. > > Signed-off-by: Juergen Gross <jgross@suse.com> > --- > arch/x86/xen/enlighten.c | 11 +++-------- > 1 file changed, 3 insertions(+), 8 deletions(-) > > diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c > index 83399ce..0eebb75 100644 > --- a/arch/x86/xen/enlighten.c > +++ b/arch/x86/xen/enlighten.c > @@ -301,9 +301,6 @@ xen_running_on_version_or_later(unsigned int major, unsigned int minor) > return false; > } > > -#define CPUID_THERM_POWER_LEAF 6 > -#define APERFMPERF_PRESENT 0 > - > static __read_mostly unsigned int cpuid_leaf1_edx_mask = ~0; > static __read_mostly unsigned int cpuid_leaf1_ecx_mask = ~0; > > @@ -337,11 +334,6 @@ static void xen_cpuid(unsigned int *ax, unsigned int *bx, > *dx = cpuid_leaf5_edx_val; > return; > > - case CPUID_THERM_POWER_LEAF: > - /* Disabling APERFMPERF for kernel usage */ > - maskecx = ~(1 << APERFMPERF_PRESENT); > - break; > - But now APERF/MPERF will be reported as supported by CPUID, won't it? > case 0xb: > /* Suppress extended topology stuff */ > maskebx = 0; > @@ -462,6 +454,9 @@ static void __init xen_init_cpuid_mask(void) > if (xen_check_mwait()) > cpuid_leaf1_ecx_set_mask = (1 << (X86_FEATURE_MWAIT % 32)); > > + /* Disable APERFMPERF feature. */ > + setup_clear_cpu_cap(X86_FEATURE_APERFMPERF); > + > /* Disable DCA feature. */ > setup_clear_cpu_cap(X86_FEATURE_DCA); I think both of those can go to xen_set_cpu_features(). -boris
On 17/02/17 15:05, Boris Ostrovsky wrote: > > > On 02/17/2017 02:36 AM, Juergen Gross wrote: >> When running as pv domain xen_cpuid() is being used instead of >> native_cpuid(). In xen_cpuid() the aperf/mperf feature is indicated >> as not being present by special casing the related cpuid leaf. >> >> Instead of delivering fake cpuid values clear the cpu capability bit >> for aperf/mperf instead. >> >> Signed-off-by: Juergen Gross <jgross@suse.com> >> --- >> arch/x86/xen/enlighten.c | 11 +++-------- >> 1 file changed, 3 insertions(+), 8 deletions(-) >> >> diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c >> index 83399ce..0eebb75 100644 >> --- a/arch/x86/xen/enlighten.c >> +++ b/arch/x86/xen/enlighten.c >> @@ -301,9 +301,6 @@ xen_running_on_version_or_later(unsigned int >> major, unsigned int minor) >> return false; >> } >> >> -#define CPUID_THERM_POWER_LEAF 6 >> -#define APERFMPERF_PRESENT 0 >> - >> static __read_mostly unsigned int cpuid_leaf1_edx_mask = ~0; >> static __read_mostly unsigned int cpuid_leaf1_ecx_mask = ~0; >> >> @@ -337,11 +334,6 @@ static void xen_cpuid(unsigned int *ax, unsigned >> int *bx, >> *dx = cpuid_leaf5_edx_val; >> return; >> >> - case CPUID_THERM_POWER_LEAF: >> - /* Disabling APERFMPERF for kernel usage */ >> - maskecx = ~(1 << APERFMPERF_PRESENT); >> - break; >> - > > > But now APERF/MPERF will be reported as supported by CPUID, won't it? Yes. But this shouldn't be a problem as the kernel is always testing X86_FEATURE_APERFMPERF for testing the support. >> case 0xb: >> /* Suppress extended topology stuff */ >> maskebx = 0; >> @@ -462,6 +454,9 @@ static void __init xen_init_cpuid_mask(void) >> if (xen_check_mwait()) >> cpuid_leaf1_ecx_set_mask = (1 << (X86_FEATURE_MWAIT % 32)); >> >> + /* Disable APERFMPERF feature. */ >> + setup_clear_cpu_cap(X86_FEATURE_APERFMPERF); >> + >> /* Disable DCA feature. */ >> setup_clear_cpu_cap(X86_FEATURE_DCA); > > > I think both of those can go to xen_set_cpu_features(). Okay. I'll move them. I think we can convert some of the remaining cpuid bit modifications to cpu capabilities as well. Juergen
On 02/17/2017 09:19 AM, Juergen Gross wrote: > On 17/02/17 15:05, Boris Ostrovsky wrote: >> >> >> On 02/17/2017 02:36 AM, Juergen Gross wrote: >>> When running as pv domain xen_cpuid() is being used instead of >>> native_cpuid(). In xen_cpuid() the aperf/mperf feature is indicated >>> as not being present by special casing the related cpuid leaf. >>> >>> Instead of delivering fake cpuid values clear the cpu capability bit >>> for aperf/mperf instead. >>> >>> Signed-off-by: Juergen Gross <jgross@suse.com> >>> --- >>> arch/x86/xen/enlighten.c | 11 +++-------- >>> 1 file changed, 3 insertions(+), 8 deletions(-) >>> >>> diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c >>> index 83399ce..0eebb75 100644 >>> --- a/arch/x86/xen/enlighten.c >>> +++ b/arch/x86/xen/enlighten.c >>> @@ -301,9 +301,6 @@ xen_running_on_version_or_later(unsigned int >>> major, unsigned int minor) >>> return false; >>> } >>> >>> -#define CPUID_THERM_POWER_LEAF 6 >>> -#define APERFMPERF_PRESENT 0 >>> - >>> static __read_mostly unsigned int cpuid_leaf1_edx_mask = ~0; >>> static __read_mostly unsigned int cpuid_leaf1_ecx_mask = ~0; >>> >>> @@ -337,11 +334,6 @@ static void xen_cpuid(unsigned int *ax, unsigned >>> int *bx, >>> *dx = cpuid_leaf5_edx_val; >>> return; >>> >>> - case CPUID_THERM_POWER_LEAF: >>> - /* Disabling APERFMPERF for kernel usage */ >>> - maskecx = ~(1 << APERFMPERF_PRESENT); >>> - break; >>> - >> >> >> But now APERF/MPERF will be reported as supported by CPUID, won't it? > > Yes. But this shouldn't be a problem as the kernel is always testing > X86_FEATURE_APERFMPERF for testing the support. But X86_FEATURE_APERFMPERF cap is set based on CPUID query. -boris > >>> case 0xb: >>> /* Suppress extended topology stuff */ >>> maskebx = 0; >>> @@ -462,6 +454,9 @@ static void __init xen_init_cpuid_mask(void) >>> if (xen_check_mwait()) >>> cpuid_leaf1_ecx_set_mask = (1 << (X86_FEATURE_MWAIT % 32)); >>> >>> + /* Disable APERFMPERF feature. */ >>> + setup_clear_cpu_cap(X86_FEATURE_APERFMPERF); >>> + >>> /* Disable DCA feature. */ >>> setup_clear_cpu_cap(X86_FEATURE_DCA); >> >> >> I think both of those can go to xen_set_cpu_features(). > > Okay. I'll move them. > > I think we can convert some of the remaining cpuid bit modifications to > cpu capabilities as well. > > > Juergen >
On 17/02/17 15:30, Boris Ostrovsky wrote: > > > On 02/17/2017 09:19 AM, Juergen Gross wrote: >> On 17/02/17 15:05, Boris Ostrovsky wrote: >>> >>> >>> On 02/17/2017 02:36 AM, Juergen Gross wrote: >>>> When running as pv domain xen_cpuid() is being used instead of >>>> native_cpuid(). In xen_cpuid() the aperf/mperf feature is indicated >>>> as not being present by special casing the related cpuid leaf. >>>> >>>> Instead of delivering fake cpuid values clear the cpu capability bit >>>> for aperf/mperf instead. >>>> >>>> Signed-off-by: Juergen Gross <jgross@suse.com> >>>> --- >>>> arch/x86/xen/enlighten.c | 11 +++-------- >>>> 1 file changed, 3 insertions(+), 8 deletions(-) >>>> >>>> diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c >>>> index 83399ce..0eebb75 100644 >>>> --- a/arch/x86/xen/enlighten.c >>>> +++ b/arch/x86/xen/enlighten.c >>>> @@ -301,9 +301,6 @@ xen_running_on_version_or_later(unsigned int >>>> major, unsigned int minor) >>>> return false; >>>> } >>>> >>>> -#define CPUID_THERM_POWER_LEAF 6 >>>> -#define APERFMPERF_PRESENT 0 >>>> - >>>> static __read_mostly unsigned int cpuid_leaf1_edx_mask = ~0; >>>> static __read_mostly unsigned int cpuid_leaf1_ecx_mask = ~0; >>>> >>>> @@ -337,11 +334,6 @@ static void xen_cpuid(unsigned int *ax, unsigned >>>> int *bx, >>>> *dx = cpuid_leaf5_edx_val; >>>> return; >>>> >>>> - case CPUID_THERM_POWER_LEAF: >>>> - /* Disabling APERFMPERF for kernel usage */ >>>> - maskecx = ~(1 << APERFMPERF_PRESENT); >>>> - break; >>>> - >>> >>> >>> But now APERF/MPERF will be reported as supported by CPUID, won't it? >> >> Yes. But this shouldn't be a problem as the kernel is always testing >> X86_FEATURE_APERFMPERF for testing the support. > > > But X86_FEATURE_APERFMPERF cap is set based on CPUID query. Right. And setup_clear_cpu_cap() will result in removing the feature from the capabilities (removes it from boot cpu capabilities at once and is setting it in the mask which is negated and anded into the capabilities after issuing the cpuid queries). Juergen
diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 83399ce..0eebb75 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -301,9 +301,6 @@ xen_running_on_version_or_later(unsigned int major, unsigned int minor) return false; } -#define CPUID_THERM_POWER_LEAF 6 -#define APERFMPERF_PRESENT 0 - static __read_mostly unsigned int cpuid_leaf1_edx_mask = ~0; static __read_mostly unsigned int cpuid_leaf1_ecx_mask = ~0; @@ -337,11 +334,6 @@ static void xen_cpuid(unsigned int *ax, unsigned int *bx, *dx = cpuid_leaf5_edx_val; return; - case CPUID_THERM_POWER_LEAF: - /* Disabling APERFMPERF for kernel usage */ - maskecx = ~(1 << APERFMPERF_PRESENT); - break; - case 0xb: /* Suppress extended topology stuff */ maskebx = 0; @@ -462,6 +454,9 @@ static void __init xen_init_cpuid_mask(void) if (xen_check_mwait()) cpuid_leaf1_ecx_set_mask = (1 << (X86_FEATURE_MWAIT % 32)); + /* Disable APERFMPERF feature. */ + setup_clear_cpu_cap(X86_FEATURE_APERFMPERF); + /* Disable DCA feature. */ setup_clear_cpu_cap(X86_FEATURE_DCA); }
When running as pv domain xen_cpuid() is being used instead of native_cpuid(). In xen_cpuid() the aperf/mperf feature is indicated as not being present by special casing the related cpuid leaf. Instead of delivering fake cpuid values clear the cpu capability bit for aperf/mperf instead. Signed-off-by: Juergen Gross <jgross@suse.com> --- arch/x86/xen/enlighten.c | 11 +++-------- 1 file changed, 3 insertions(+), 8 deletions(-)