Message ID | 20210819004305.20203-1-deepak.sharma@amd.com (mailing list archive) |
---|---|
State | Superseded, archived |
Headers | show |
Series | x86/ACPI/State: Optimize C3 entry on AMD CPUs | expand |
On Thu, Aug 19, 2021 at 2:43 AM Deepak Sharma <deepak.sharma@amd.com> wrote: > > AMD CPU which support C3 shares cache. Its not necessary to flush the > caches in software before entering C3. This will cause performance drop > for the cores which share some caches. ARB_DIS is not used with current > AMD C state implementation. So set related flags correctly. > > Signed-off-by: Deepak Sharma <deepak.sharma@amd.com> Applied as 5.15 material under the edited subject "x86: ACPI: cstate: Optimize C3 entry on AMD CPUs", thanks! > --- > arch/x86/kernel/acpi/cstate.c | 15 +++++++++++++++ > 1 file changed, 15 insertions(+) > > diff --git a/arch/x86/kernel/acpi/cstate.c b/arch/x86/kernel/acpi/cstate.c > index 7de599eba7f0..62a5986d625a 100644 > --- a/arch/x86/kernel/acpi/cstate.c > +++ b/arch/x86/kernel/acpi/cstate.c > @@ -79,6 +79,21 @@ void acpi_processor_power_init_bm_check(struct acpi_processor_flags *flags, > */ > flags->bm_control = 0; > } > + if (c->x86_vendor == X86_VENDOR_AMD) { > + /* > + * For all AMD CPUs that support C3, caches should not be > + * flushed by software while entering C3 type state. Set > + * bm->check to 1 so that kernel doesn't need to execute > + * cache flush operation. > + */ > + flags->bm_check = 1; > + /* > + * In current AMD C state implementation ARB_DIS is no longer > + * used. So set bm_control to zero to indicate ARB_DIS is not > + * required while entering C3 type state. > + */ > + flags->bm_control = 0; > + } > } > EXPORT_SYMBOL(acpi_processor_power_init_bm_check); > > -- > 2.25.1 >
On Wed, Aug 18 2021 at 17:43, Deepak Sharma wrote: > AMD CPU which support C3 shares cache. Its not necessary to flush the > caches in software before entering C3. This will cause performance drop > for the cores which share some caches. ARB_DIS is not used with current > AMD C state implementation. So set related flags correctly. > > Signed-off-by: Deepak Sharma <deepak.sharma@amd.com> > --- > arch/x86/kernel/acpi/cstate.c | 15 +++++++++++++++ > 1 file changed, 15 insertions(+) > > diff --git a/arch/x86/kernel/acpi/cstate.c b/arch/x86/kernel/acpi/cstate.c > index 7de599eba7f0..62a5986d625a 100644 > --- a/arch/x86/kernel/acpi/cstate.c > +++ b/arch/x86/kernel/acpi/cstate.c > @@ -79,6 +79,21 @@ void acpi_processor_power_init_bm_check(struct acpi_processor_flags *flags, > */ > flags->bm_control = 0; > } > + if (c->x86_vendor == X86_VENDOR_AMD) { > + /* > + * For all AMD CPUs that support C3, caches should not be > + * flushed by software while entering C3 type state. Set > + * bm->check to 1 so that kernel doesn't need to execute > + * cache flush operation. > + */ > + flags->bm_check = 1; > + /* > + * In current AMD C state implementation ARB_DIS is no longer Fine for current implementations, but what about older implementations? Thanks, tglx
On 8/26/21 4:04 PM, Thomas Gleixner wrote: > On Wed, Aug 18 2021 at 17:43, Deepak Sharma wrote: > >> AMD CPU which support C3 shares cache. Its not necessary to flush the >> caches in software before entering C3. This will cause performance drop >> for the cores which share some caches. ARB_DIS is not used with current >> AMD C state implementation. So set related flags correctly. >> >> Signed-off-by: Deepak Sharma <deepak.sharma@amd.com> >> --- >> arch/x86/kernel/acpi/cstate.c | 15 +++++++++++++++ >> 1 file changed, 15 insertions(+) >> >> diff --git a/arch/x86/kernel/acpi/cstate.c b/arch/x86/kernel/acpi/cstate.c >> index 7de599eba7f0..62a5986d625a 100644 >> --- a/arch/x86/kernel/acpi/cstate.c >> +++ b/arch/x86/kernel/acpi/cstate.c >> @@ -79,6 +79,21 @@ void acpi_processor_power_init_bm_check(struct acpi_processor_flags *flags, >> */ >> flags->bm_control = 0; >> } >> + if (c->x86_vendor == X86_VENDOR_AMD) { >> + /* >> + * For all AMD CPUs that support C3, caches should not be >> + * flushed by software while entering C3 type state. Set >> + * bm->check to 1 so that kernel doesn't need to execute >> + * cache flush operation. >> + */ >> + flags->bm_check = 1; >> + /* >> + * In current AMD C state implementation ARB_DIS is no longer > Fine for current implementations, but what about older implementations? We are internally discussing about its validity on much older implementations. Will send subsequent patch based on the conclusion. Thanks, Deepak
On 8/25/21 11:07 AM, Rafael J. Wysocki wrote: > On Thu, Aug 19, 2021 at 2:43 AM Deepak Sharma <deepak.sharma@amd.com> wrote: >> AMD CPU which support C3 shares cache. Its not necessary to flush the >> caches in software before entering C3. This will cause performance drop >> for the cores which share some caches. ARB_DIS is not used with current >> AMD C state implementation. So set related flags correctly. >> >> Signed-off-by: Deepak Sharma <deepak.sharma@amd.com> > Applied as 5.15 material under the edited subject "x86: ACPI: cstate: > Optimize C3 entry on AMD CPUs", thanks! I might need to send subsequent patch for this. Can you please point me to git and branch where this has been merged. >> --- >> arch/x86/kernel/acpi/cstate.c | 15 +++++++++++++++ >> 1 file changed, 15 insertions(+) >> >> diff --git a/arch/x86/kernel/acpi/cstate.c b/arch/x86/kernel/acpi/cstate.c >> index 7de599eba7f0..62a5986d625a 100644 >> --- a/arch/x86/kernel/acpi/cstate.c >> +++ b/arch/x86/kernel/acpi/cstate.c >> @@ -79,6 +79,21 @@ void acpi_processor_power_init_bm_check(struct acpi_processor_flags *flags, >> */ >> flags->bm_control = 0; >> } >> + if (c->x86_vendor == X86_VENDOR_AMD) { >> + /* >> + * For all AMD CPUs that support C3, caches should not be >> + * flushed by software while entering C3 type state. Set >> + * bm->check to 1 so that kernel doesn't need to execute >> + * cache flush operation. >> + */ >> + flags->bm_check = 1; >> + /* >> + * In current AMD C state implementation ARB_DIS is no longer >> + * used. So set bm_control to zero to indicate ARB_DIS is not >> + * required while entering C3 type state. >> + */ >> + flags->bm_control = 0; >> + } >> } >> EXPORT_SYMBOL(acpi_processor_power_init_bm_check); >> >> -- >> 2.25.1 >>
On Wed, Sep 1, 2021 at 4:14 AM Deepak Sharma <deesharm@amd.com> wrote: > > > On 8/25/21 11:07 AM, Rafael J. Wysocki wrote: > > On Thu, Aug 19, 2021 at 2:43 AM Deepak Sharma <deepak.sharma@amd.com> wrote: > >> AMD CPU which support C3 shares cache. Its not necessary to flush the > >> caches in software before entering C3. This will cause performance drop > >> for the cores which share some caches. ARB_DIS is not used with current > >> AMD C state implementation. So set related flags correctly. > >> > >> Signed-off-by: Deepak Sharma <deepak.sharma@amd.com> > > Applied as 5.15 material under the edited subject "x86: ACPI: cstate: > > Optimize C3 entry on AMD CPUs", thanks! > > I might need to send subsequent patch for this. Can you please point me > to git and branch where this has been merged. git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm.git linux-next > >> --- > >> arch/x86/kernel/acpi/cstate.c | 15 +++++++++++++++ > >> 1 file changed, 15 insertions(+) > >> > >> diff --git a/arch/x86/kernel/acpi/cstate.c b/arch/x86/kernel/acpi/cstate.c > >> index 7de599eba7f0..62a5986d625a 100644 > >> --- a/arch/x86/kernel/acpi/cstate.c > >> +++ b/arch/x86/kernel/acpi/cstate.c > >> @@ -79,6 +79,21 @@ void acpi_processor_power_init_bm_check(struct acpi_processor_flags *flags, > >> */ > >> flags->bm_control = 0; > >> } > >> + if (c->x86_vendor == X86_VENDOR_AMD) { > >> + /* > >> + * For all AMD CPUs that support C3, caches should not be > >> + * flushed by software while entering C3 type state. Set > >> + * bm->check to 1 so that kernel doesn't need to execute > >> + * cache flush operation. > >> + */ > >> + flags->bm_check = 1; > >> + /* > >> + * In current AMD C state implementation ARB_DIS is no longer > >> + * used. So set bm_control to zero to indicate ARB_DIS is not > >> + * required while entering C3 type state. > >> + */ > >> + flags->bm_control = 0; > >> + } > >> } > >> EXPORT_SYMBOL(acpi_processor_power_init_bm_check); > >> > >> -- > >> 2.25.1 > >>
Hi Rafael, On 9/1/2021 5:45 AM, Rafael J. Wysocki wrote: > On Wed, Sep 1, 2021 at 4:14 AM Deepak Sharma <deesharm@amd.com> wrote: >> >> On 8/25/21 11:07 AM, Rafael J. Wysocki wrote: >>> On Thu, Aug 19, 2021 at 2:43 AM Deepak Sharma <deepak.sharma@amd.com> wrote: >>>> AMD CPU which support C3 shares cache. Its not necessary to flush the >>>> caches in software before entering C3. This will cause performance drop >>>> for the cores which share some caches. ARB_DIS is not used with current >>>> AMD C state implementation. So set related flags correctly. >>>> >>>> Signed-off-by: Deepak Sharma <deepak.sharma@amd.com> >>> Applied as 5.15 material under the edited subject "x86: ACPI: cstate: >>> Optimize C3 entry on AMD CPUs", thanks! >> I might need to send subsequent patch for this. Can you please point me >> to git and branch where this has been merged. > git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm.git linux-next Still I am not able to see this patch merged on linux-next branch. Thanks, Deepak >>>> --- >>>> arch/x86/kernel/acpi/cstate.c | 15 +++++++++++++++ >>>> 1 file changed, 15 insertions(+) >>>> >>>> diff --git a/arch/x86/kernel/acpi/cstate.c b/arch/x86/kernel/acpi/cstate.c >>>> index 7de599eba7f0..62a5986d625a 100644 >>>> --- a/arch/x86/kernel/acpi/cstate.c >>>> +++ b/arch/x86/kernel/acpi/cstate.c >>>> @@ -79,6 +79,21 @@ void acpi_processor_power_init_bm_check(struct acpi_processor_flags *flags, >>>> */ >>>> flags->bm_control = 0; >>>> } >>>> + if (c->x86_vendor == X86_VENDOR_AMD) { >>>> + /* >>>> + * For all AMD CPUs that support C3, caches should not be >>>> + * flushed by software while entering C3 type state. Set >>>> + * bm->check to 1 so that kernel doesn't need to execute >>>> + * cache flush operation. >>>> + */ >>>> + flags->bm_check = 1; >>>> + /* >>>> + * In current AMD C state implementation ARB_DIS is no longer >>>> + * used. So set bm_control to zero to indicate ARB_DIS is not >>>> + * required while entering C3 type state. >>>> + */ >>>> + flags->bm_control = 0; >>>> + } >>>> } >>>> EXPORT_SYMBOL(acpi_processor_power_init_bm_check); >>>> >>>> -- >>>> 2.25.1 >>>>
On Wed, Sep 22, 2021 at 5:50 AM Sharma, Deepak <deesharm@amd.com> wrote: > > Hi Rafael, > > On 9/1/2021 5:45 AM, Rafael J. Wysocki wrote: > > On Wed, Sep 1, 2021 at 4:14 AM Deepak Sharma <deesharm@amd.com> wrote: > >> > >> On 8/25/21 11:07 AM, Rafael J. Wysocki wrote: > >>> On Thu, Aug 19, 2021 at 2:43 AM Deepak Sharma <deepak.sharma@amd.com> wrote: > >>>> AMD CPU which support C3 shares cache. Its not necessary to flush the > >>>> caches in software before entering C3. This will cause performance drop > >>>> for the cores which share some caches. ARB_DIS is not used with current > >>>> AMD C state implementation. So set related flags correctly. > >>>> > >>>> Signed-off-by: Deepak Sharma <deepak.sharma@amd.com> > >>> Applied as 5.15 material under the edited subject "x86: ACPI: cstate: > >>> Optimize C3 entry on AMD CPUs", thanks! > >> I might need to send subsequent patch for this. Can you please point me > >> to git and branch where this has been merged. > > git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm.git linux-next > > Still I am not able to see this patch merged on linux-next branch. It was there, but got dropped before the merge window due to the unclear next steps. I'd rather see this resent along with the subsequent patch you were talking about.
On 9/22/2021 5:47 AM, Rafael J. Wysocki wrote: > On Wed, Sep 22, 2021 at 5:50 AM Sharma, Deepak <deesharm@amd.com> wrote: >> Hi Rafael, >> >> On 9/1/2021 5:45 AM, Rafael J. Wysocki wrote: >>> On Wed, Sep 1, 2021 at 4:14 AM Deepak Sharma <deesharm@amd.com> wrote: >>>> On 8/25/21 11:07 AM, Rafael J. Wysocki wrote: >>>>> On Thu, Aug 19, 2021 at 2:43 AM Deepak Sharma <deepak.sharma@amd.com> wrote: >>>>>> AMD CPU which support C3 shares cache. Its not necessary to flush the >>>>>> caches in software before entering C3. This will cause performance drop >>>>>> for the cores which share some caches. ARB_DIS is not used with current >>>>>> AMD C state implementation. So set related flags correctly. >>>>>> >>>>>> Signed-off-by: Deepak Sharma <deepak.sharma@amd.com> >>>>> Applied as 5.15 material under the edited subject "x86: ACPI: cstate: >>>>> Optimize C3 entry on AMD CPUs", thanks! >>>> I might need to send subsequent patch for this. Can you please point me >>>> to git and branch where this has been merged. >>> git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm.git linux-next >> Still I am not able to see this patch merged on linux-next branch. > It was there, but got dropped before the merge window due to the > unclear next steps. > > I'd rather see this resent along with the subsequent patch you were > talking about. Thanks for the clarification. I will resend patch with subsequent changes which will have support for zen onward CPU's.
diff --git a/arch/x86/kernel/acpi/cstate.c b/arch/x86/kernel/acpi/cstate.c index 7de599eba7f0..62a5986d625a 100644 --- a/arch/x86/kernel/acpi/cstate.c +++ b/arch/x86/kernel/acpi/cstate.c @@ -79,6 +79,21 @@ void acpi_processor_power_init_bm_check(struct acpi_processor_flags *flags, */ flags->bm_control = 0; } + if (c->x86_vendor == X86_VENDOR_AMD) { + /* + * For all AMD CPUs that support C3, caches should not be + * flushed by software while entering C3 type state. Set + * bm->check to 1 so that kernel doesn't need to execute + * cache flush operation. + */ + flags->bm_check = 1; + /* + * In current AMD C state implementation ARB_DIS is no longer + * used. So set bm_control to zero to indicate ARB_DIS is not + * required while entering C3 type state. + */ + flags->bm_control = 0; + } } EXPORT_SYMBOL(acpi_processor_power_init_bm_check);
AMD CPU which support C3 shares cache. Its not necessary to flush the caches in software before entering C3. This will cause performance drop for the cores which share some caches. ARB_DIS is not used with current AMD C state implementation. So set related flags correctly. Signed-off-by: Deepak Sharma <deepak.sharma@amd.com> --- arch/x86/kernel/acpi/cstate.c | 15 +++++++++++++++ 1 file changed, 15 insertions(+)