Message ID | 20250221210200.244405-1-prsampat@amd.com |
---|---|
Headers | show |
Series | Basic SEV-SNP Selftests | expand |
On 21/02/2025 21:01, Pratik R. Sampat wrote: > During platform init, SNP initialization may fail for several reasons, > such as firmware command failures and incompatible versions. However, > the KVM capability may continue to advertise support for it. Export this > information to KVM and withdraw SEV-SNP support if has not been > successfully initialized. > > Fixes: 1dfe571c12cf ("KVM: SEV: Add initial SEV-SNP support") > Suggested-by: Sean Christopherson <seanjc@google.com> > Signed-off-by: Pratik R. Sampat <prsampat@amd.com> > --- > v6..v7: > > * Replace FW version check with sev->snp_initialized (Sean) > --- > arch/x86/kvm/svm/sev.c | 4 +++- > drivers/crypto/ccp/sev-dev.c | 8 ++++++++ > include/linux/psp-sev.h | 3 +++ > 3 files changed, 14 insertions(+), 1 deletion(-) > > diff --git a/arch/x86/kvm/svm/sev.c b/arch/x86/kvm/svm/sev.c > index 0dbb25442ec1..87b5d63a5817 100644 > --- a/arch/x86/kvm/svm/sev.c > +++ b/arch/x86/kvm/svm/sev.c > @@ -3050,7 +3050,9 @@ void __init sev_hardware_setup(void) > sev_es_asid_count = min_sev_asid - 1; > WARN_ON_ONCE(misc_cg_set_capacity(MISC_CG_RES_SEV_ES, sev_es_asid_count)); > sev_es_supported = true; > - sev_snp_supported = sev_snp_enabled && cc_platform_has(CC_ATTR_HOST_SEV_SNP); > + sev_snp_supported = (sev_snp_enabled && > + cc_platform_has(CC_ATTR_HOST_SEV_SNP) && > + snp_initialized()); > > out: > if (boot_cpu_has(X86_FEATURE_SEV)) > diff --git a/drivers/crypto/ccp/sev-dev.c b/drivers/crypto/ccp/sev-dev.c > index 2e87ca0e292a..8d2cf8552bc2 100644 > --- a/drivers/crypto/ccp/sev-dev.c > +++ b/drivers/crypto/ccp/sev-dev.c > @@ -1352,6 +1352,14 @@ int sev_platform_init(struct sev_platform_init_args *args) > } > EXPORT_SYMBOL_GPL(sev_platform_init); > > +bool snp_initialized(void) > +{ > + struct sev_device *sev = psp_master->sev_data; Should check psp_master isn't NULL before accessing just in case (particularly for future potential callers). (e.g. see ccb88e9549e7 ("crypto: ccp - Fix null pointer dereference in __sev_platform_shutdown_locked") > + > + return sev->snp_initialized; > +} > +EXPORT_SYMBOL_GPL(snp_initialized); > + > static int __sev_platform_shutdown_locked(int *error) > { > struct psp_device *psp = psp_master; > diff --git a/include/linux/psp-sev.h b/include/linux/psp-sev.h > index f3cad182d4ef..d34068c87a28 100644 > --- a/include/linux/psp-sev.h > +++ b/include/linux/psp-sev.h > @@ -954,6 +954,7 @@ int sev_do_cmd(int cmd, void *data, int *psp_ret); > void *psp_copy_user_blob(u64 uaddr, u32 len); > void *snp_alloc_firmware_page(gfp_t mask); > void snp_free_firmware_page(void *addr); > +bool snp_initialized(void); > > #else /* !CONFIG_CRYPTO_DEV_SP_PSP */ > > @@ -988,6 +989,8 @@ static inline void *snp_alloc_firmware_page(gfp_t mask) > > static inline void snp_free_firmware_page(void *addr) { } > > +static inline bool snp_initialized(void) { return false; } > + > #endif /* CONFIG_CRYPTO_DEV_SP_PSP */ > > #endif /* __PSP_SEV_H__ */
On 2/21/25 15:01, Pratik R. Sampat wrote: > During platform init, SNP initialization may fail for several reasons, > such as firmware command failures and incompatible versions. However, > the KVM capability may continue to advertise support for it. Export this > information to KVM and withdraw SEV-SNP support if has not been > successfully initialized. Hmmm... rather than creating a new API, can you just issue an SNP_PLATFORM_STATUS command and see if the SNP is not in the UNINIT state? Thanks, Tom > > Fixes: 1dfe571c12cf ("KVM: SEV: Add initial SEV-SNP support") > Suggested-by: Sean Christopherson <seanjc@google.com> > Signed-off-by: Pratik R. Sampat <prsampat@amd.com> > --- > v6..v7: > > * Replace FW version check with sev->snp_initialized (Sean) > --- > arch/x86/kvm/svm/sev.c | 4 +++- > drivers/crypto/ccp/sev-dev.c | 8 ++++++++ > include/linux/psp-sev.h | 3 +++ > 3 files changed, 14 insertions(+), 1 deletion(-) > > diff --git a/arch/x86/kvm/svm/sev.c b/arch/x86/kvm/svm/sev.c > index 0dbb25442ec1..87b5d63a5817 100644 > --- a/arch/x86/kvm/svm/sev.c > +++ b/arch/x86/kvm/svm/sev.c > @@ -3050,7 +3050,9 @@ void __init sev_hardware_setup(void) > sev_es_asid_count = min_sev_asid - 1; > WARN_ON_ONCE(misc_cg_set_capacity(MISC_CG_RES_SEV_ES, sev_es_asid_count)); > sev_es_supported = true; > - sev_snp_supported = sev_snp_enabled && cc_platform_has(CC_ATTR_HOST_SEV_SNP); > + sev_snp_supported = (sev_snp_enabled && > + cc_platform_has(CC_ATTR_HOST_SEV_SNP) && > + snp_initialized()); > > out: > if (boot_cpu_has(X86_FEATURE_SEV)) > diff --git a/drivers/crypto/ccp/sev-dev.c b/drivers/crypto/ccp/sev-dev.c > index 2e87ca0e292a..8d2cf8552bc2 100644 > --- a/drivers/crypto/ccp/sev-dev.c > +++ b/drivers/crypto/ccp/sev-dev.c > @@ -1352,6 +1352,14 @@ int sev_platform_init(struct sev_platform_init_args *args) > } > EXPORT_SYMBOL_GPL(sev_platform_init); > > +bool snp_initialized(void) > +{ > + struct sev_device *sev = psp_master->sev_data; > + > + return sev->snp_initialized; > +} > +EXPORT_SYMBOL_GPL(snp_initialized); > + > static int __sev_platform_shutdown_locked(int *error) > { > struct psp_device *psp = psp_master; > diff --git a/include/linux/psp-sev.h b/include/linux/psp-sev.h > index f3cad182d4ef..d34068c87a28 100644 > --- a/include/linux/psp-sev.h > +++ b/include/linux/psp-sev.h > @@ -954,6 +954,7 @@ int sev_do_cmd(int cmd, void *data, int *psp_ret); > void *psp_copy_user_blob(u64 uaddr, u32 len); > void *snp_alloc_firmware_page(gfp_t mask); > void snp_free_firmware_page(void *addr); > +bool snp_initialized(void); > > #else /* !CONFIG_CRYPTO_DEV_SP_PSP */ > > @@ -988,6 +989,8 @@ static inline void *snp_alloc_firmware_page(gfp_t mask) > > static inline void snp_free_firmware_page(void *addr) { } > > +static inline bool snp_initialized(void) { return false; } > + > #endif /* CONFIG_CRYPTO_DEV_SP_PSP */ > > #endif /* __PSP_SEV_H__ */
Hi Tom, On 2/24/2025 3:28 PM, Tom Lendacky wrote: > On 2/21/25 15:01, Pratik R. Sampat wrote: >> During platform init, SNP initialization may fail for several reasons, >> such as firmware command failures and incompatible versions. However, >> the KVM capability may continue to advertise support for it. Export this >> information to KVM and withdraw SEV-SNP support if has not been >> successfully initialized. > > Hmmm... rather than creating a new API, can you just issue an > SNP_PLATFORM_STATUS command and see if the SNP is not in the UNINIT state? > Although reading sev->snp_initialized is probably cheaper to do, it is cleaner to query the platform status. Querying SNP_PLATFORM_STATUS requires the pages to transition to firmware-owned and back, and the helpers for it are implemented within sev-dev.c. So, similar to sev_platform_status(), I'm thinking it is probably better to create the snp_platform_status() API as well and use that within KVM to check the state. Thanks! Pratik
Hi Liam, Thanks for review! On 2/24/2025 1:01 PM, Liam Merwick wrote: > > > On 21/02/2025 21:01, Pratik R. Sampat wrote: >> During platform init, SNP initialization may fail for several reasons, >> such as firmware command failures and incompatible versions. However, >> the KVM capability may continue to advertise support for it. Export this >> information to KVM and withdraw SEV-SNP support if has not been >> successfully initialized. >> >> Fixes: 1dfe571c12cf ("KVM: SEV: Add initial SEV-SNP support") >> Suggested-by: Sean Christopherson <seanjc@google.com> >> Signed-off-by: Pratik R. Sampat <prsampat@amd.com> >> --- >> v6..v7: >> >> * Replace FW version check with sev->snp_initialized (Sean) >> --- >> arch/x86/kvm/svm/sev.c | 4 +++- >> drivers/crypto/ccp/sev-dev.c | 8 ++++++++ >> include/linux/psp-sev.h | 3 +++ >> 3 files changed, 14 insertions(+), 1 deletion(-) >> >> diff --git a/arch/x86/kvm/svm/sev.c b/arch/x86/kvm/svm/sev.c >> index 0dbb25442ec1..87b5d63a5817 100644 >> --- a/arch/x86/kvm/svm/sev.c >> +++ b/arch/x86/kvm/svm/sev.c >> @@ -3050,7 +3050,9 @@ void __init sev_hardware_setup(void) >> sev_es_asid_count = min_sev_asid - 1; >> WARN_ON_ONCE(misc_cg_set_capacity(MISC_CG_RES_SEV_ES, >> sev_es_asid_count)); >> sev_es_supported = true; >> - sev_snp_supported = sev_snp_enabled && >> cc_platform_has(CC_ATTR_HOST_SEV_SNP); >> + sev_snp_supported = (sev_snp_enabled && >> + cc_platform_has(CC_ATTR_HOST_SEV_SNP) && >> + snp_initialized()); >> out: >> if (boot_cpu_has(X86_FEATURE_SEV)) >> diff --git a/drivers/crypto/ccp/sev-dev.c b/drivers/crypto/ccp/sev-dev.c >> index 2e87ca0e292a..8d2cf8552bc2 100644 >> --- a/drivers/crypto/ccp/sev-dev.c >> +++ b/drivers/crypto/ccp/sev-dev.c >> @@ -1352,6 +1352,14 @@ int sev_platform_init(struct >> sev_platform_init_args *args) >> } >> EXPORT_SYMBOL_GPL(sev_platform_init); >> +bool snp_initialized(void) >> +{ >> + struct sev_device *sev = psp_master->sev_data; > > > Should check psp_master isn't NULL before accessing just in case > (particularly for future potential callers). > > (e.g. see ccb88e9549e7 ("crypto: ccp - Fix null pointer dereference in > __sev_platform_shutdown_locked") > Thanks for pointing this out, if I end up using this interface, I'll put the NULL check in. Thanks! Pratik > > > >> + >> + return sev->snp_initialized; >> +} >> +EXPORT_SYMBOL_GPL(snp_initialized); >> + >> static int __sev_platform_shutdown_locked(int *error) >> { >> struct psp_device *psp = psp_master; >> diff --git a/include/linux/psp-sev.h b/include/linux/psp-sev.h >> index f3cad182d4ef..d34068c87a28 100644 >> --- a/include/linux/psp-sev.h >> +++ b/include/linux/psp-sev.h >> @@ -954,6 +954,7 @@ int sev_do_cmd(int cmd, void *data, int *psp_ret); >> void *psp_copy_user_blob(u64 uaddr, u32 len); >> void *snp_alloc_firmware_page(gfp_t mask); >> void snp_free_firmware_page(void *addr); >> +bool snp_initialized(void); >> #else /* !CONFIG_CRYPTO_DEV_SP_PSP */ >> @@ -988,6 +989,8 @@ static inline void >> *snp_alloc_firmware_page(gfp_t mask) >> static inline void snp_free_firmware_page(void *addr) { } >> +static inline bool snp_initialized(void) { return false; } >> + >> #endif /* CONFIG_CRYPTO_DEV_SP_PSP */ >> #endif /* __PSP_SEV_H__ */ >
On 2/25/2025 10:41 AM, Pratik R. Sampat wrote: > Hi Tom, > > On 2/24/2025 3:28 PM, Tom Lendacky wrote: >> On 2/21/25 15:01, Pratik R. Sampat wrote: >>> During platform init, SNP initialization may fail for several reasons, >>> such as firmware command failures and incompatible versions. However, >>> the KVM capability may continue to advertise support for it. Export this >>> information to KVM and withdraw SEV-SNP support if has not been >>> successfully initialized. >> >> Hmmm... rather than creating a new API, can you just issue an >> SNP_PLATFORM_STATUS command and see if the SNP is not in the UNINIT state? >> > > Although reading sev->snp_initialized is probably cheaper to do, it is > cleaner to query the platform status. > > Querying SNP_PLATFORM_STATUS requires the pages to transition to > firmware-owned and back, and the helpers for it are implemented within > sev-dev.c. So, similar to sev_platform_status(), I'm thinking it is > probably better to create the snp_platform_status() API as well and use > that within KVM to check the state. > Although I am guessing the initial intent was to not have an API exposed at all from CCP and only make the SNP_PLATFORM_STATUS call instead? Since that may not be cleanly possible (we have helpers for page state conversions such as rmp_mark_pages_firmware() in ccp) without duplicating functionality in KVM as well, I guess the question really boils down to whether we export the cheaper snp_initialized() or the snp_platform_status() API instead? Thanks again! Pratik
On 2/25/2025 1:09 PM, Tom Lendacky wrote: > On 2/25/25 11:45, Pratik R. Sampat wrote: >> On 2/25/2025 10:41 AM, Pratik R. Sampat wrote: >>> Hi Tom, >>> >>> On 2/24/2025 3:28 PM, Tom Lendacky wrote: >>>> On 2/21/25 15:01, Pratik R. Sampat wrote: >>>>> During platform init, SNP initialization may fail for several reasons, >>>>> such as firmware command failures and incompatible versions. However, >>>>> the KVM capability may continue to advertise support for it. Export this >>>>> information to KVM and withdraw SEV-SNP support if has not been >>>>> successfully initialized. >>>> >>>> Hmmm... rather than creating a new API, can you just issue an >>>> SNP_PLATFORM_STATUS command and see if the SNP is not in the UNINIT state? >>>> >>> >>> Although reading sev->snp_initialized is probably cheaper to do, it is >>> cleaner to query the platform status. >>> >>> Querying SNP_PLATFORM_STATUS requires the pages to transition to >>> firmware-owned and back, and the helpers for it are implemented within >>> sev-dev.c. So, similar to sev_platform_status(), I'm thinking it is >>> probably better to create the snp_platform_status() API as well and use >>> that within KVM to check the state. >>> >> >> Although I am guessing the initial intent was to not have an API exposed >> at all from CCP and only make the SNP_PLATFORM_STATUS call instead? >> >> Since that may not be cleanly possible (we have helpers for page state >> conversions such as rmp_mark_pages_firmware() in ccp) without >> duplicating functionality in KVM as well, I guess the question really >> boils down to whether we export the cheaper snp_initialized() or the >> snp_platform_status() API instead? > > Taking a closer look, we do already have APIs that KVM uses to allocate > firmware pages (output pages for SNP APIs) that can be used: > snp_alloc_firmware_page() and snp_free_firmware_page(). > > I think that should be enough to use sev_do_cmd() to perform the > SEV_CMD_SNP_PLATFORM_STATUS command without exposing a new API. > Ah, I had missed that! Calling the SNP_PLATFORM_STATUS this way works. Pratik > Thanks, > Tom > >> >> Thanks again! >> Pratik