| Message ID | 1447539130-4613-2-git-send-email-m.smarduch@samsung.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 14/11/15 22:12, Mario Smarduch wrote: > This patch adds vcpu fields to track lazy state, save host FPEXC, and > offsets to fields. > > Signed-off-by: Mario Smarduch <m.smarduch@samsung.com> > --- > arch/arm/include/asm/kvm_host.h | 6 ++++++ > arch/arm/kernel/asm-offsets.c | 2 ++ > 2 files changed, 8 insertions(+) > > diff --git a/arch/arm/include/asm/kvm_host.h b/arch/arm/include/asm/kvm_host.h > index 3df1e97..f1bf551 100644 > --- a/arch/arm/include/asm/kvm_host.h > +++ b/arch/arm/include/asm/kvm_host.h > @@ -107,6 +107,12 @@ struct kvm_vcpu_arch { > /* Interrupt related fields */ > u32 irq_lines; /* IRQ and FIQ levels */ > > + /* fp/simd dirty flag true if guest accessed register file */ > + bool vfp_dirty; I think we do not need this bool, because it is already represented by the state of the trapping bits. > + > + /* Save host FPEXC register to later restore on vcpu put */ > + u32 host_fpexc; > + > /* Exception Information */ > struct kvm_vcpu_fault_info fault; > > diff --git a/arch/arm/kernel/asm-offsets.c b/arch/arm/kernel/asm-offsets.c > index 871b826..9f79712 100644 > --- a/arch/arm/kernel/asm-offsets.c > +++ b/arch/arm/kernel/asm-offsets.c > @@ -186,6 +186,8 @@ int main(void) > DEFINE(VCPU_CPSR, offsetof(struct kvm_vcpu, arch.regs.usr_regs.ARM_cpsr)); > DEFINE(VCPU_HCR, offsetof(struct kvm_vcpu, arch.hcr)); > DEFINE(VCPU_IRQ_LINES, offsetof(struct kvm_vcpu, arch.irq_lines)); > + DEFINE(VCPU_VFP_DIRTY, offsetof(struct kvm_vcpu, arch.vfp_dirty)); > + DEFINE(VCPU_VFP_HOST_FPEXC, offsetof(struct kvm_vcpu, arch.host_fpexc)); > DEFINE(VCPU_HSR, offsetof(struct kvm_vcpu, arch.fault.hsr)); > DEFINE(VCPU_HxFAR, offsetof(struct kvm_vcpu, arch.fault.hxfar)); > DEFINE(VCPU_HPFAR, offsetof(struct kvm_vcpu, arch.fault.hpfar)); > Thanks, M.
On 12/3/2015 7:46 AM, Marc Zyngier wrote: > On 14/11/15 22:12, Mario Smarduch wrote: >> This patch adds vcpu fields to track lazy state, save host FPEXC, and >> offsets to fields. >> >> Signed-off-by: Mario Smarduch <m.smarduch@samsung.com> >> --- >> arch/arm/include/asm/kvm_host.h | 6 ++++++ >> arch/arm/kernel/asm-offsets.c | 2 ++ >> 2 files changed, 8 insertions(+) >> >> diff --git a/arch/arm/include/asm/kvm_host.h b/arch/arm/include/asm/kvm_host.h >> index 3df1e97..f1bf551 100644 >> --- a/arch/arm/include/asm/kvm_host.h >> +++ b/arch/arm/include/asm/kvm_host.h >> @@ -107,6 +107,12 @@ struct kvm_vcpu_arch { >> /* Interrupt related fields */ >> u32 irq_lines; /* IRQ and FIQ levels */ >> >> + /* fp/simd dirty flag true if guest accessed register file */ >> + bool vfp_dirty; > > I think we do not need this bool, because it is already represented by > the state of the trapping bits. The trapping bit state is lost on exit since they're cleared, no? > >> + >> + /* Save host FPEXC register to later restore on vcpu put */ >> + u32 host_fpexc; >> + >> /* Exception Information */ >> struct kvm_vcpu_fault_info fault; >> >> diff --git a/arch/arm/kernel/asm-offsets.c b/arch/arm/kernel/asm-offsets.c >> index 871b826..9f79712 100644 >> --- a/arch/arm/kernel/asm-offsets.c >> +++ b/arch/arm/kernel/asm-offsets.c >> @@ -186,6 +186,8 @@ int main(void) >> DEFINE(VCPU_CPSR, offsetof(struct kvm_vcpu, arch.regs.usr_regs.ARM_cpsr)); >> DEFINE(VCPU_HCR, offsetof(struct kvm_vcpu, arch.hcr)); >> DEFINE(VCPU_IRQ_LINES, offsetof(struct kvm_vcpu, arch.irq_lines)); >> + DEFINE(VCPU_VFP_DIRTY, offsetof(struct kvm_vcpu, arch.vfp_dirty)); >> + DEFINE(VCPU_VFP_HOST_FPEXC, offsetof(struct kvm_vcpu, arch.host_fpexc)); >> DEFINE(VCPU_HSR, offsetof(struct kvm_vcpu, arch.fault.hsr)); >> DEFINE(VCPU_HxFAR, offsetof(struct kvm_vcpu, arch.fault.hxfar)); >> DEFINE(VCPU_HPFAR, offsetof(struct kvm_vcpu, arch.fault.hpfar)); >> > > Thanks, > > M. >
On 03/12/15 19:21, Mario Smarduch wrote: > > > On 12/3/2015 7:46 AM, Marc Zyngier wrote: >> On 14/11/15 22:12, Mario Smarduch wrote: >>> This patch adds vcpu fields to track lazy state, save host FPEXC, and >>> offsets to fields. >>> >>> Signed-off-by: Mario Smarduch <m.smarduch@samsung.com> >>> --- >>> arch/arm/include/asm/kvm_host.h | 6 ++++++ >>> arch/arm/kernel/asm-offsets.c | 2 ++ >>> 2 files changed, 8 insertions(+) >>> >>> diff --git a/arch/arm/include/asm/kvm_host.h b/arch/arm/include/asm/kvm_host.h >>> index 3df1e97..f1bf551 100644 >>> --- a/arch/arm/include/asm/kvm_host.h >>> +++ b/arch/arm/include/asm/kvm_host.h >>> @@ -107,6 +107,12 @@ struct kvm_vcpu_arch { >>> /* Interrupt related fields */ >>> u32 irq_lines; /* IRQ and FIQ levels */ >>> >>> + /* fp/simd dirty flag true if guest accessed register file */ >>> + bool vfp_dirty; >> >> I think we do not need this bool, because it is already represented by >> the state of the trapping bits. > > The trapping bit state is lost on exit since they're cleared, no? But that's what should actually be preserved, no? At the moment, you maintain some side state to reflect what the trapping state is. You might as well keep it around all the time. Thanks, M.
On 12/3/2015 11:24 AM, Marc Zyngier wrote: > On 03/12/15 19:21, Mario Smarduch wrote: >> >> >> On 12/3/2015 7:46 AM, Marc Zyngier wrote: >>> On 14/11/15 22:12, Mario Smarduch wrote: >>>> This patch adds vcpu fields to track lazy state, save host FPEXC, and >>>> offsets to fields. >>>> >>>> Signed-off-by: Mario Smarduch <m.smarduch@samsung.com> >>>> --- >>>> arch/arm/include/asm/kvm_host.h | 6 ++++++ >>>> arch/arm/kernel/asm-offsets.c | 2 ++ >>>> 2 files changed, 8 insertions(+) >>>> >>>> diff --git a/arch/arm/include/asm/kvm_host.h b/arch/arm/include/asm/kvm_host.h >>>> index 3df1e97..f1bf551 100644 >>>> --- a/arch/arm/include/asm/kvm_host.h >>>> +++ b/arch/arm/include/asm/kvm_host.h >>>> @@ -107,6 +107,12 @@ struct kvm_vcpu_arch { >>>> /* Interrupt related fields */ >>>> u32 irq_lines; /* IRQ and FIQ levels */ >>>> >>>> + /* fp/simd dirty flag true if guest accessed register file */ >>>> + bool vfp_dirty; >>> >>> I think we do not need this bool, because it is already represented by >>> the state of the trapping bits. >> >> The trapping bit state is lost on exit since they're cleared, no? > > But that's what should actually be preserved, no? At the moment, you > maintain some side state to reflect what the trapping state is. You > might as well keep it around all the time. Ok I see, you should be able to preserve and use the trap registers. I'll rework it. > > Thanks, > > M. >
diff --git a/arch/arm/include/asm/kvm_host.h b/arch/arm/include/asm/kvm_host.h index 3df1e97..f1bf551 100644 --- a/arch/arm/include/asm/kvm_host.h +++ b/arch/arm/include/asm/kvm_host.h @@ -107,6 +107,12 @@ struct kvm_vcpu_arch { /* Interrupt related fields */ u32 irq_lines; /* IRQ and FIQ levels */ + /* fp/simd dirty flag true if guest accessed register file */ + bool vfp_dirty; + + /* Save host FPEXC register to later restore on vcpu put */ + u32 host_fpexc; + /* Exception Information */ struct kvm_vcpu_fault_info fault; diff --git a/arch/arm/kernel/asm-offsets.c b/arch/arm/kernel/asm-offsets.c index 871b826..9f79712 100644 --- a/arch/arm/kernel/asm-offsets.c +++ b/arch/arm/kernel/asm-offsets.c @@ -186,6 +186,8 @@ int main(void) DEFINE(VCPU_CPSR, offsetof(struct kvm_vcpu, arch.regs.usr_regs.ARM_cpsr)); DEFINE(VCPU_HCR, offsetof(struct kvm_vcpu, arch.hcr)); DEFINE(VCPU_IRQ_LINES, offsetof(struct kvm_vcpu, arch.irq_lines)); + DEFINE(VCPU_VFP_DIRTY, offsetof(struct kvm_vcpu, arch.vfp_dirty)); + DEFINE(VCPU_VFP_HOST_FPEXC, offsetof(struct kvm_vcpu, arch.host_fpexc)); DEFINE(VCPU_HSR, offsetof(struct kvm_vcpu, arch.fault.hsr)); DEFINE(VCPU_HxFAR, offsetof(struct kvm_vcpu, arch.fault.hxfar)); DEFINE(VCPU_HPFAR, offsetof(struct kvm_vcpu, arch.fault.hpfar));
This patch adds vcpu fields to track lazy state, save host FPEXC, and offsets to fields. Signed-off-by: Mario Smarduch <m.smarduch@samsung.com> --- arch/arm/include/asm/kvm_host.h | 6 ++++++ arch/arm/kernel/asm-offsets.c | 2 ++ 2 files changed, 8 insertions(+)