| Message ID | 20190109135435.178664-3-marc.zyngier@arm.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | arm64: KVM: Allow direct function calls on VHE | expand |
On Wed, Jan 09, 2019 at 01:54:34PM +0000, Marc Zyngier wrote: > When running VHE, there is no need to jump via some stub to perform > a "HYP" function call, as there is a single address space. > > Let's thus change kvm_call_hyp() and co to perform a direct call > in this case. Although this results in a bit of code expansion, > it allows the compiler to check for type compatibility, something > that we are missing so far. > > Signed-off-by: Marc Zyngier <marc.zyngier@arm.com> > --- > arch/arm64/include/asm/kvm_host.h | 32 +++++++++++++++++++++++++++++-- > 1 file changed, 30 insertions(+), 2 deletions(-) > > diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h > index e54cb7c88a4e..df32edbadd69 100644 > --- a/arch/arm64/include/asm/kvm_host.h > +++ b/arch/arm64/include/asm/kvm_host.h > @@ -370,8 +370,36 @@ void kvm_arm_halt_guest(struct kvm *kvm); > void kvm_arm_resume_guest(struct kvm *kvm); > > u64 __kvm_call_hyp(void *hypfn, ...); > -#define kvm_call_hyp(f, ...) __kvm_call_hyp(kvm_ksym_ref(f), ##__VA_ARGS__) > -#define kvm_call_hyp_ret(f, ...) kvm_call_hyp(f, ##__VA_ARGS__) > + > +/* > + * The couple of isb() below are there to guarantee the same behaviour > + * on VHE as on !VHE, where the eret to EL1 acts as a context > + * synchronization event. > + */ > +#define kvm_call_hyp(f, ...) \ > + do { \ > + if (cpus_have_const_cap(ARM64_HAS_VIRT_HOST_EXTN)) { \ > + f(__VA_ARGS__); \ > + isb(); \ > + } else { \ > + __kvm_call_hyp(kvm_ksym_ref(f), ##__VA_ARGS__); \ > + } \ > + } while(0) > + > +#define kvm_call_hyp_ret(f, ...) \ > + ({ \ > + u64 ret; \ > + \ > + if (cpus_have_const_cap(ARM64_HAS_VIRT_HOST_EXTN)) { \ > + ret = f(__VA_ARGS__); \ __kvm_get_mdcr_el2 and __kvm_vcpu_run_nvhe don't return u64 type, they return a smaller type. I guess any issues would be picked up when compiling, but should the name of the macro be clearer as to the assumptions it makes? Or perhaps take an argument which is the type of ret? Andrew Murray > + isb(); \ > + } else { \ > + ret = __kvm_call_hyp(kvm_ksym_ref(f), \ > + ##__VA_ARGS__); \ > + } \ > + \ > + ret; \ > + }) > > void force_vm_exit(const cpumask_t *mask); > void kvm_mmu_wp_memory_region(struct kvm *kvm, int slot); > -- > 2.20.1 > > _______________________________________________ > kvmarm mailing list > kvmarm@lists.cs.columbia.edu > https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Hi Andrew, On 09/01/2019 14:24, Andrew Murray wrote: > On Wed, Jan 09, 2019 at 01:54:34PM +0000, Marc Zyngier wrote: >> When running VHE, there is no need to jump via some stub to perform >> a "HYP" function call, as there is a single address space. >> >> Let's thus change kvm_call_hyp() and co to perform a direct call >> in this case. Although this results in a bit of code expansion, >> it allows the compiler to check for type compatibility, something >> that we are missing so far. >> >> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com> >> --- >> arch/arm64/include/asm/kvm_host.h | 32 +++++++++++++++++++++++++++++-- >> 1 file changed, 30 insertions(+), 2 deletions(-) >> >> diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h >> index e54cb7c88a4e..df32edbadd69 100644 >> --- a/arch/arm64/include/asm/kvm_host.h >> +++ b/arch/arm64/include/asm/kvm_host.h >> @@ -370,8 +370,36 @@ void kvm_arm_halt_guest(struct kvm *kvm); >> void kvm_arm_resume_guest(struct kvm *kvm); >> >> u64 __kvm_call_hyp(void *hypfn, ...); >> -#define kvm_call_hyp(f, ...) __kvm_call_hyp(kvm_ksym_ref(f), ##__VA_ARGS__) >> -#define kvm_call_hyp_ret(f, ...) kvm_call_hyp(f, ##__VA_ARGS__) >> + >> +/* >> + * The couple of isb() below are there to guarantee the same behaviour >> + * on VHE as on !VHE, where the eret to EL1 acts as a context >> + * synchronization event. >> + */ >> +#define kvm_call_hyp(f, ...) \ >> + do { \ >> + if (cpus_have_const_cap(ARM64_HAS_VIRT_HOST_EXTN)) { \ >> + f(__VA_ARGS__); \ >> + isb(); \ >> + } else { \ >> + __kvm_call_hyp(kvm_ksym_ref(f), ##__VA_ARGS__); \ >> + } \ >> + } while(0) >> + >> +#define kvm_call_hyp_ret(f, ...) \ >> + ({ \ >> + u64 ret; \ >> + \ >> + if (cpus_have_const_cap(ARM64_HAS_VIRT_HOST_EXTN)) { \ >> + ret = f(__VA_ARGS__); \ > > __kvm_get_mdcr_el2 and __kvm_vcpu_run_nvhe don't return u64 type, they > return a smaller type. I guess any issues would be picked up when compiling, > but should the name of the macro be clearer as to the assumptions it makes? > Or perhaps take an argument which is the type of ret? kvm_call_hyp has always returned a u64, so no semantic has changed here. Otherwise, your suggestion of specifying a return type is interesting, but it also gives the programmer another chance to shoot itself in the foot by not providing the return type corresponding to the function that is called. Unless we can extract the return type by pure magic, I'm not sure we gain much. Thanks, M.
On 09/01/2019 14:45, Marc Zyngier wrote: > Hi Andrew, > > On 09/01/2019 14:24, Andrew Murray wrote: >> On Wed, Jan 09, 2019 at 01:54:34PM +0000, Marc Zyngier wrote: >>> When running VHE, there is no need to jump via some stub to perform >>> a "HYP" function call, as there is a single address space. >>> >>> Let's thus change kvm_call_hyp() and co to perform a direct call >>> in this case. Although this results in a bit of code expansion, >>> it allows the compiler to check for type compatibility, something >>> that we are missing so far. >>> >>> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com> >>> --- >>> arch/arm64/include/asm/kvm_host.h | 32 +++++++++++++++++++++++++++++-- >>> 1 file changed, 30 insertions(+), 2 deletions(-) >>> >>> diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h >>> index e54cb7c88a4e..df32edbadd69 100644 >>> --- a/arch/arm64/include/asm/kvm_host.h >>> +++ b/arch/arm64/include/asm/kvm_host.h >>> @@ -370,8 +370,36 @@ void kvm_arm_halt_guest(struct kvm *kvm); >>> void kvm_arm_resume_guest(struct kvm *kvm); >>> >>> u64 __kvm_call_hyp(void *hypfn, ...); >>> -#define kvm_call_hyp(f, ...) __kvm_call_hyp(kvm_ksym_ref(f), ##__VA_ARGS__) >>> -#define kvm_call_hyp_ret(f, ...) kvm_call_hyp(f, ##__VA_ARGS__) >>> + >>> +/* >>> + * The couple of isb() below are there to guarantee the same behaviour >>> + * on VHE as on !VHE, where the eret to EL1 acts as a context >>> + * synchronization event. >>> + */ >>> +#define kvm_call_hyp(f, ...) \ >>> + do { \ >>> + if (cpus_have_const_cap(ARM64_HAS_VIRT_HOST_EXTN)) { \ >>> + f(__VA_ARGS__); \ >>> + isb(); \ >>> + } else { \ >>> + __kvm_call_hyp(kvm_ksym_ref(f), ##__VA_ARGS__); \ >>> + } \ >>> + } while(0) >>> + >>> +#define kvm_call_hyp_ret(f, ...) \ >>> + ({ \ >>> + u64 ret; \ >>> + \ >>> + if (cpus_have_const_cap(ARM64_HAS_VIRT_HOST_EXTN)) { \ >>> + ret = f(__VA_ARGS__); \ >> >> __kvm_get_mdcr_el2 and __kvm_vcpu_run_nvhe don't return u64 type, they >> return a smaller type. I guess any issues would be picked up when compiling, >> but should the name of the macro be clearer as to the assumptions it makes? >> Or perhaps take an argument which is the type of ret? > > kvm_call_hyp has always returned a u64, so no semantic has changed here. > > Otherwise, your suggestion of specifying a return type is interesting, > but it also gives the programmer another chance to shoot itself in the > foot by not providing the return type corresponding to the function that > is called. > > Unless we can extract the return type by pure magic, I'm not sure we > gain much. > Would the following work? typeof(f(__VA_ARGS__)) ret; If typeof works anything like sizeof, I'd expect it would evaluate stuff passed as argument and we'd have the return type of the function. Cheers,
On 09/01/2019 14:51, Julien Thierry wrote: > > > On 09/01/2019 14:45, Marc Zyngier wrote: >> Hi Andrew, >> >> On 09/01/2019 14:24, Andrew Murray wrote: >>> On Wed, Jan 09, 2019 at 01:54:34PM +0000, Marc Zyngier wrote: >>>> When running VHE, there is no need to jump via some stub to perform >>>> a "HYP" function call, as there is a single address space. >>>> >>>> Let's thus change kvm_call_hyp() and co to perform a direct call >>>> in this case. Although this results in a bit of code expansion, >>>> it allows the compiler to check for type compatibility, something >>>> that we are missing so far. >>>> >>>> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com> >>>> --- >>>> arch/arm64/include/asm/kvm_host.h | 32 +++++++++++++++++++++++++++++-- >>>> 1 file changed, 30 insertions(+), 2 deletions(-) >>>> >>>> diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h >>>> index e54cb7c88a4e..df32edbadd69 100644 >>>> --- a/arch/arm64/include/asm/kvm_host.h >>>> +++ b/arch/arm64/include/asm/kvm_host.h >>>> @@ -370,8 +370,36 @@ void kvm_arm_halt_guest(struct kvm *kvm); >>>> void kvm_arm_resume_guest(struct kvm *kvm); >>>> >>>> u64 __kvm_call_hyp(void *hypfn, ...); >>>> -#define kvm_call_hyp(f, ...) __kvm_call_hyp(kvm_ksym_ref(f), ##__VA_ARGS__) >>>> -#define kvm_call_hyp_ret(f, ...) kvm_call_hyp(f, ##__VA_ARGS__) >>>> + >>>> +/* >>>> + * The couple of isb() below are there to guarantee the same behaviour >>>> + * on VHE as on !VHE, where the eret to EL1 acts as a context >>>> + * synchronization event. >>>> + */ >>>> +#define kvm_call_hyp(f, ...) \ >>>> + do { \ >>>> + if (cpus_have_const_cap(ARM64_HAS_VIRT_HOST_EXTN)) { \ >>>> + f(__VA_ARGS__); \ >>>> + isb(); \ >>>> + } else { \ >>>> + __kvm_call_hyp(kvm_ksym_ref(f), ##__VA_ARGS__); \ >>>> + } \ >>>> + } while(0) >>>> + >>>> +#define kvm_call_hyp_ret(f, ...) \ >>>> + ({ \ >>>> + u64 ret; \ >>>> + \ >>>> + if (cpus_have_const_cap(ARM64_HAS_VIRT_HOST_EXTN)) { \ >>>> + ret = f(__VA_ARGS__); \ >>> >>> __kvm_get_mdcr_el2 and __kvm_vcpu_run_nvhe don't return u64 type, they >>> return a smaller type. I guess any issues would be picked up when compiling, >>> but should the name of the macro be clearer as to the assumptions it makes? >>> Or perhaps take an argument which is the type of ret? >> >> kvm_call_hyp has always returned a u64, so no semantic has changed here. >> >> Otherwise, your suggestion of specifying a return type is interesting, >> but it also gives the programmer another chance to shoot itself in the >> foot by not providing the return type corresponding to the function that >> is called. >> >> Unless we can extract the return type by pure magic, I'm not sure we >> gain much. >> > > Would the following work? > > typeof(f(__VA_ARGS__)) ret; > > If typeof works anything like sizeof, I'd expect it would evaluate stuff it wouldn't* > passed as argument and we'd have the return type of the function. > > Cheers, >
On 09/01/2019 14:51, Julien Thierry wrote: > > > On 09/01/2019 14:45, Marc Zyngier wrote: >> Hi Andrew, >> >> On 09/01/2019 14:24, Andrew Murray wrote: >>> On Wed, Jan 09, 2019 at 01:54:34PM +0000, Marc Zyngier wrote: >>>> When running VHE, there is no need to jump via some stub to perform >>>> a "HYP" function call, as there is a single address space. >>>> >>>> Let's thus change kvm_call_hyp() and co to perform a direct call >>>> in this case. Although this results in a bit of code expansion, >>>> it allows the compiler to check for type compatibility, something >>>> that we are missing so far. >>>> >>>> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com> >>>> --- >>>> arch/arm64/include/asm/kvm_host.h | 32 +++++++++++++++++++++++++++++-- >>>> 1 file changed, 30 insertions(+), 2 deletions(-) >>>> >>>> diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h >>>> index e54cb7c88a4e..df32edbadd69 100644 >>>> --- a/arch/arm64/include/asm/kvm_host.h >>>> +++ b/arch/arm64/include/asm/kvm_host.h >>>> @@ -370,8 +370,36 @@ void kvm_arm_halt_guest(struct kvm *kvm); >>>> void kvm_arm_resume_guest(struct kvm *kvm); >>>> >>>> u64 __kvm_call_hyp(void *hypfn, ...); >>>> -#define kvm_call_hyp(f, ...) __kvm_call_hyp(kvm_ksym_ref(f), ##__VA_ARGS__) >>>> -#define kvm_call_hyp_ret(f, ...) kvm_call_hyp(f, ##__VA_ARGS__) >>>> + >>>> +/* >>>> + * The couple of isb() below are there to guarantee the same behaviour >>>> + * on VHE as on !VHE, where the eret to EL1 acts as a context >>>> + * synchronization event. >>>> + */ >>>> +#define kvm_call_hyp(f, ...) \ >>>> + do { \ >>>> + if (cpus_have_const_cap(ARM64_HAS_VIRT_HOST_EXTN)) { \ >>>> + f(__VA_ARGS__); \ >>>> + isb(); \ >>>> + } else { \ >>>> + __kvm_call_hyp(kvm_ksym_ref(f), ##__VA_ARGS__); \ >>>> + } \ >>>> + } while(0) >>>> + >>>> +#define kvm_call_hyp_ret(f, ...) \ >>>> + ({ \ >>>> + u64 ret; \ >>>> + \ >>>> + if (cpus_have_const_cap(ARM64_HAS_VIRT_HOST_EXTN)) { \ >>>> + ret = f(__VA_ARGS__); \ >>> >>> __kvm_get_mdcr_el2 and __kvm_vcpu_run_nvhe don't return u64 type, they >>> return a smaller type. I guess any issues would be picked up when compiling, >>> but should the name of the macro be clearer as to the assumptions it makes? >>> Or perhaps take an argument which is the type of ret? >> >> kvm_call_hyp has always returned a u64, so no semantic has changed here. >> >> Otherwise, your suggestion of specifying a return type is interesting, >> but it also gives the programmer another chance to shoot itself in the >> foot by not providing the return type corresponding to the function that >> is called. >> >> Unless we can extract the return type by pure magic, I'm not sure we >> gain much. >> > > Would the following work? > > typeof(f(__VA_ARGS__)) ret; > > If typeof works anything like sizeof, I'd expect it would evaluate stuff > passed as argument and we'd have the return type of the function. And it actually works! Thanks for the awful tip! ;-) M.
On Wed, Jan 09, 2019 at 04:01:56PM +0000, Marc Zyngier wrote: > On 09/01/2019 14:51, Julien Thierry wrote: > > > > > > On 09/01/2019 14:45, Marc Zyngier wrote: > >> Hi Andrew, > >> > >> On 09/01/2019 14:24, Andrew Murray wrote: > >>> On Wed, Jan 09, 2019 at 01:54:34PM +0000, Marc Zyngier wrote: > >>>> When running VHE, there is no need to jump via some stub to perform > >>>> a "HYP" function call, as there is a single address space. > >>>> > >>>> Let's thus change kvm_call_hyp() and co to perform a direct call > >>>> in this case. Although this results in a bit of code expansion, > >>>> it allows the compiler to check for type compatibility, something > >>>> that we are missing so far. > >>>> > >>>> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com> > >>>> --- > >>>> arch/arm64/include/asm/kvm_host.h | 32 +++++++++++++++++++++++++++++-- > >>>> 1 file changed, 30 insertions(+), 2 deletions(-) > >>>> > >>>> diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h > >>>> index e54cb7c88a4e..df32edbadd69 100644 > >>>> --- a/arch/arm64/include/asm/kvm_host.h > >>>> +++ b/arch/arm64/include/asm/kvm_host.h > >>>> @@ -370,8 +370,36 @@ void kvm_arm_halt_guest(struct kvm *kvm); > >>>> void kvm_arm_resume_guest(struct kvm *kvm); > >>>> > >>>> u64 __kvm_call_hyp(void *hypfn, ...); > >>>> -#define kvm_call_hyp(f, ...) __kvm_call_hyp(kvm_ksym_ref(f), ##__VA_ARGS__) > >>>> -#define kvm_call_hyp_ret(f, ...) kvm_call_hyp(f, ##__VA_ARGS__) > >>>> + > >>>> +/* > >>>> + * The couple of isb() below are there to guarantee the same behaviour > >>>> + * on VHE as on !VHE, where the eret to EL1 acts as a context > >>>> + * synchronization event. > >>>> + */ > >>>> +#define kvm_call_hyp(f, ...) \ > >>>> + do { \ > >>>> + if (cpus_have_const_cap(ARM64_HAS_VIRT_HOST_EXTN)) { \ > >>>> + f(__VA_ARGS__); \ > >>>> + isb(); \ > >>>> + } else { \ > >>>> + __kvm_call_hyp(kvm_ksym_ref(f), ##__VA_ARGS__); \ > >>>> + } \ > >>>> + } while(0) > >>>> + > >>>> +#define kvm_call_hyp_ret(f, ...) \ > >>>> + ({ \ > >>>> + u64 ret; \ > >>>> + \ > >>>> + if (cpus_have_const_cap(ARM64_HAS_VIRT_HOST_EXTN)) { \ > >>>> + ret = f(__VA_ARGS__); \ > >>> > >>> __kvm_get_mdcr_el2 and __kvm_vcpu_run_nvhe don't return u64 type, they > >>> return a smaller type. I guess any issues would be picked up when compiling, > >>> but should the name of the macro be clearer as to the assumptions it makes? > >>> Or perhaps take an argument which is the type of ret? > >> > >> kvm_call_hyp has always returned a u64, so no semantic has changed here. Ah I missed that! > >> > >> Otherwise, your suggestion of specifying a return type is interesting, > >> but it also gives the programmer another chance to shoot itself in the > >> foot by not providing the return type corresponding to the function that > >> is called. > >> > >> Unless we can extract the return type by pure magic, I'm not sure we > >> gain much. > >> > > > > Would the following work? > > > > typeof(f(__VA_ARGS__)) ret; > > > > If typeof works anything like sizeof, I'd expect it would evaluate stuff > > passed as argument and we'd have the return type of the function. > And it actually works! Thanks for the awful tip! ;-) Awesome. Thanks, Andrew Murray > > M. > -- > Jazz is not dead. It just smells funny...
diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h index e54cb7c88a4e..df32edbadd69 100644 --- a/arch/arm64/include/asm/kvm_host.h +++ b/arch/arm64/include/asm/kvm_host.h @@ -370,8 +370,36 @@ void kvm_arm_halt_guest(struct kvm *kvm); void kvm_arm_resume_guest(struct kvm *kvm); u64 __kvm_call_hyp(void *hypfn, ...); -#define kvm_call_hyp(f, ...) __kvm_call_hyp(kvm_ksym_ref(f), ##__VA_ARGS__) -#define kvm_call_hyp_ret(f, ...) kvm_call_hyp(f, ##__VA_ARGS__) + +/* + * The couple of isb() below are there to guarantee the same behaviour + * on VHE as on !VHE, where the eret to EL1 acts as a context + * synchronization event. + */ +#define kvm_call_hyp(f, ...) \ + do { \ + if (cpus_have_const_cap(ARM64_HAS_VIRT_HOST_EXTN)) { \ + f(__VA_ARGS__); \ + isb(); \ + } else { \ + __kvm_call_hyp(kvm_ksym_ref(f), ##__VA_ARGS__); \ + } \ + } while(0) + +#define kvm_call_hyp_ret(f, ...) \ + ({ \ + u64 ret; \ + \ + if (cpus_have_const_cap(ARM64_HAS_VIRT_HOST_EXTN)) { \ + ret = f(__VA_ARGS__); \ + isb(); \ + } else { \ + ret = __kvm_call_hyp(kvm_ksym_ref(f), \ + ##__VA_ARGS__); \ + } \ + \ + ret; \ + }) void force_vm_exit(const cpumask_t *mask); void kvm_mmu_wp_memory_region(struct kvm *kvm, int slot);
When running VHE, there is no need to jump via some stub to perform a "HYP" function call, as there is a single address space. Let's thus change kvm_call_hyp() and co to perform a direct call in this case. Although this results in a bit of code expansion, it allows the compiler to check for type compatibility, something that we are missing so far. Signed-off-by: Marc Zyngier <marc.zyngier@arm.com> --- arch/arm64/include/asm/kvm_host.h | 32 +++++++++++++++++++++++++++++-- 1 file changed, 30 insertions(+), 2 deletions(-)