KVM: arm64: Fixes for SMC64 SYSTEM_RESET2 calls
cover KVM: arm64: Fixes for SMC64 SYSTEM_RESET2 calls
This series addresses a couple of issues with how KVM exposes SMC64 calls to its guest. It is currently possible for an AArch32 guest to discover the SMC64 SYSTEM_RESET2 function (via PSCI_1_0_FN_PSCI_FEATURES) and even make a call to it. SMCCC does not allow for 64 bit calls to be made from a 32 bit state.
Patch 1 cleans up the way we filter SMC64 calls in PSCI. Using a switch with case statements for each possibly-filtered function is asking for trouble. Instead, pivot off of the bit that indicates the desired calling convention. This plugs the PSCI_FEATURES hole for SYSTEM_RESET2.
Patch 2 adds a check to the PSCI v1.x call handler in KVM, bailing out early if the guest is not allowed to use a particular function. This closes the door on calls to 64-bit SYSTEM_RESET2 from AArch32.
My first crack at this [1] was missing the fix for direct calls to SYSTEM_RESET2. Taking the patch out of that series and sending separately.
Applies on top of today's kvmarm pull, commit:
21ea45784275 ("KVM: arm64: fix typos in comments")
[1]: https://patchwork.kernel.org/project/kvm/patch/[email protected]/
Oliver Upton (2):
KVM: arm64: Generally disallow SMC64 for AArch32 guests
KVM: arm64: Actually prevent SMC64 SYSTEM_RESET2 from AArch32
arch/arm64/kvm/psci.c | 19 ++++++++++--------- 1 file changed, 10 insertions(+), 9 deletions(-)
Signature · 1 line
2.35.1.894.gb6a874cedc-goog
PATCH 1/2 KVM: arm64: Generally disallow SMC64 for AArch32 guests
The only valid calling SMC calling convention from an AArch32 state is SMC32. Disallow any PSCI function that sets the SMC64 function ID bit when called from AArch32 rather than comparing against known SMC64 PSCI functions.
Note that without this change KVM advertises the SMC64 flavor of SYSTEM_RESET2 to AArch32 guests.
--- arch/arm64/kvm/psci.c | 14 +++++--------- 1 file changed, 5 insertions(+), 9 deletions(-)
arch/arm64/kvm/psci.c +5 -9
@@ -216,15 +216,11 @@ static void kvm_psci_narrow_to_32bit(struct kvm_vcpu *vcpu)static unsigned long kvm_psci_check_allowed_function (struct kvm_vcpu *vcpu,u32 fn) { -switch (fn) { -case PSCI_0_2_FN64_CPU_SUSPEND : -case PSCI_0_2_FN64_CPU_ON : -case PSCI_0_2_FN64_AFFINITY_INFO : -/* Disallow these functions for 32bit guests */ -if (vcpu_mode_is_32bit (vcpu)) -return PSCI_RET_NOT_SUPPORTED ; -break ; - } +/* +* Prevent 32 bit guests from calling 64 bit PSCI functions. +*/ +if ((fn &PSCI_0_2_64BIT ) &&vcpu_mode_is_32bit (vcpu)) +return PSCI_RET_NOT_SUPPORTED ;return 0 ; }
Signature · 1 line
2.35.1.894.gb6a874cedc-goog
PATCH 2/2 KVM: arm64: Actually prevent SMC64 SYSTEM_RESET2 from AArch32
In reply to Oliver UptonThe SMCCC does not allow the SMC64 calling convention to be used from AArch32. While KVM checks to see if the calling convention is allowed in PSCI_1_0_FN_PSCI_FEATURES, it does not actually prevent calls to unadvertised PSCI v1.0+ functions.
Check to see if the requested function is allowed from the guest's execution state. Deny the call if it is not.
--- arch/arm64/kvm/psci.c | 5 +++++ 1 file changed, 5 insertions(+)
arch/arm64/kvm/psci.c +5 -0
@@ -318,6 +318,10 @@ static int kvm_psci_1_x_call(struct kvm_vcpu *vcpu, u32 minor)if (minor >1 )return -EINVAL ; + val =kvm_psci_check_allowed_function (vcpu, psci_fn); +if (val) + goto out; +switch (psci_fn) {case PSCI_0_2_FN_PSCI_VERSION : val = minor ==0 ?KVM_ARM_PSCI_1_0 :KVM_ARM_PSCI_1_1 ;
@@ -378,6 +382,7 @@ static int kvm_psci_1_x_call(struct kvm_vcpu *vcpu, u32 minor)return kvm_psci_0_2_call (vcpu); } +out:smccc_set_retval (vcpu, val,0 ,0 ,0 );return ret; }
Signature · 1 line
2.35.1.894.gb6a874cedc-goog
Re: [PATCH 2/2] KVM: arm64: Actually prevent SMC64 SYSTEM_RESET2 from AArch32
Quoted 11 lines · Oliver Upton
The SMCCC does not allow the SMC64 calling convention to be used from AArch32. While KVM checks to see if the calling convention is allowed in
PSCI_1_0_FN_PSCI_FEATURES, it does not actually prevent calls to unadvertised PSCI v1.0+ functions.Check to see if the requested function is allowed from the guest's execution state. Deny the call if it is not.
Fixes d43583b890e7 ("KVM: arm64: Expose PSCI SYSTEM_RESET2 call to the guest")Cc Will Deacon <[email protected]>Signed-off-by Oliver Upton <[email protected]>
BTW, considering the new kvm_psci_check_allowed_function()implementation in the patch-1, it might be better to call kvm_psci_check_allowed_function() from kvm_psci_call() instead? Then, we could avoid the similar issue next time we support a newer PSCI version.
Thanks,
Reiji
Quoted 27 lines
--- arch/arm64/kvm/psci.c | 5 +++++ 1 file changed, 5 insertions(+)arch/arm64/kvm/psci.c +5
-0@@ -318,6 +318,10 @@ static int kvm_psci_1_x_call(struct kvm_vcpu *vcpu, u32 minor)if (minor >1 )return -EINVAL ; + val =kvm_psci_check_allowed_function (vcpu, psci_fn); +if (val) + goto out; +switch (psci_fn) {case PSCI_0_2_FN_PSCI_VERSION : val = minor ==0 ?KVM_ARM_PSCI_1_0 :KVM_ARM_PSCI_1_1 ;@@ -378,6 +382,7 @@ static int kvm_psci_1_x_call(struct kvm_vcpu *vcpu, u32 minor)return kvm_psci_0_2_call (vcpu); } +out:smccc_set_retval (vcpu, val,0 ,0 ,0 );return ret; }
Re: [PATCH 2/2] KVM: arm64: Actually prevent SMC64 SYSTEM_RESET2 from AArch32
Quoted 6 lines · Reiji Watanabe
Quoted 5 lines · Oliver Upton
The SMCCC does not allow the SMC64 calling convention to be used from AArch32. While KVM checks to see if the calling convention is allowed in
PSCI_1_0_FN_PSCI_FEATURES, it does not actually prevent calls to unadvertised PSCI v1.0+ functions.
Quoted 8 lines
Quoted 6 lines
Check to see if the requested function is allowed from the guest's execution state. Deny the call if it is not.
Fixes d43583b890e7 ("KVM: arm64: Expose PSCI SYSTEM_RESET2 call to the guest")Cc Will Deacon <[email protected]>Signed-off-by Oliver Upton <[email protected]>Reviewed-by Reiji Watanabe <[email protected]>
Appreciated :-)
Quoted 4 lines
BTW, considering the new kvm_psci_check_allowed_function()implementation in the patch-1, it might be better to call
kvm_psci_check_allowed_function()fromkvm_psci_call()instead? Then, we could avoid the similar issue next time we support a newer PSCI version.
Good point. If Marc doesn't bite in the next day or two I'll address this with a new spin, otherwise I'll do a separate cleanup. Just want to avoid spamming on this topic since I already replied with yet another patch [1].
Thanks!
[1] https://lore.kernel.org/kvmarm/[email protected]
--
Oliver
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainerRe: [PATCH 2/2] KVM: arm64: Actually prevent SMC64 SYSTEM_RESET2 from AArch32
Quoted 24 lines · Oliver Upton
Quoted 6 lines · Reiji Watanabe
Quoted 5 lines · Oliver Upton
The SMCCC does not allow the SMC64 calling convention to be used from AArch32. While KVM checks to see if the calling convention is allowed in
PSCI_1_0_FN_PSCI_FEATURES, it does not actually prevent calls to unadvertised PSCI v1.0+ functions.Quoted 8 lines
Quoted 6 lines
Check to see if the requested function is allowed from the guest's execution state. Deny the call if it is not.
Fixes d43583b890e7 ("KVM: arm64: Expose PSCI SYSTEM_RESET2 call to the guest")Cc Will Deacon <[email protected]>Signed-off-by Oliver Upton <[email protected]>Reviewed-by Reiji Watanabe <[email protected]>Appreciated :-)
Quoted 4 lines
BTW, considering the new kvm_psci_check_allowed_function()implementation in the patch-1, it might be better to call
kvm_psci_check_allowed_function()fromkvm_psci_call()instead? Then, we could avoid the similar issue next time we support a newer PSCI version.
Quoted 4 lines
Good point. If Marc doesn't bite in the next day or two I'll address this with a new spin, otherwise I'll do a separate cleanup. Just want to avoid spamming on this topic since I already replied with yet another patch [1].
Please do, and I'll queue that for -rc1.
Thanks,
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
Quoted 23 lines · Oliver Upton
This series addresses a couple of issues with how KVM exposes SMC64 calls to its guest. It is currently possible for an AArch32 guest to discover the SMC64
SYSTEM_RESET2function (viaPSCI_1_0_FN_PSCI_FEATURES) and even make a call to it. SMCCC does not allow for 64 bit calls to be made from a 32 bit state.Patch 1 cleans up the way we filter SMC64 calls in PSCI. Using a switch with case statements for each possibly-filtered function is asking for trouble. Instead, pivot off of the bit that indicates the desired calling convention. This plugs the
PSCI_FEATUREShole forSYSTEM_RESET2.Patch 2 adds a check to the PSCI v1.x call handler in KVM, bailing out early if the guest is not allowed to use a particular function. This closes the door on calls to 64-bit
SYSTEM_RESET2from AArch32.My first crack at this [1] was missing the fix for direct calls to
SYSTEM_RESET2. Taking the patch out of that series and sending separately.Applies on top of today's kvmarm pull, commit:
21ea45784275("KVM: arm64: fix typos in comments")
Quoted 8 lines
[1]: https://patchwork.kernel.org/project/kvm/patch/[email protected]/
Oliver Upton (2):
KVM: arm64: Generally disallow SMC64 for AArch32 guests
KVM: arm64: Actually prevent SMC64SYSTEM_RESET2from AArch32arch/arm64/kvm/psci.c | 19 ++++++++++--------- 1 file changed, 10 insertions(+), 9 deletions(-)
For both patches:
Thanks for fixing this!
Will
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
PATCH 1/1 KVM: arm64: Drop unneeded minor version check from PSCI v1.x handler
In reply to Oliver UptonWe already sanitize the guest's PSCI version when it is being written by userspace, rejecting unsupported version numbers. Additionally, the 'minor' parameter to kvm_psci_1_x_call() is a constant known at compile time for all callsites.
Though it is benign, the additional check against the
PSCI kvm_psci_1_x_call() is unnecessary and likely to be missed the next time KVM raises its maximum PSCI version. Drop the check altogether and rely on sanitization when the PSCI version is set by userspace.
No functional change intended.
---
Sorry for not sending this with the other ones. I took another read and do not believe this check is necessary + might hurt when we raise the PSCI version again.
Applies on top of the series [1], which itself is based on kvmarm/next at commit:
21ea45784275 ("KVM: arm64: fix typos in comments")
[1]: http://lore.kernel.org/r/[email protected]
arch/arm64/kvm/psci.c | 3 --- 1 file changed, 3 deletions(-)
arch/arm64/kvm/psci.c +0 -3
@@ -315,9 +315,6 @@ static int kvm_psci_1_x_call(struct kvm_vcpu *vcpu, u32 minor)unsigned long val;int ret =1 ; -if (minor >1 ) -return -EINVAL ; - val =kvm_psci_check_allowed_function (vcpu, psci_fn);if (val) goto out;
Signature · 1 line
2.35.1.894.gb6a874cedc-goog