Skip to contentSkip to conversation

Bug fixes related to secure 2 stage translation

PATCH Applied Idan Horowitz 4 participants 12 messages Mar 27, 2022 – Mar 31, 2022

lore
Share & tools
lore
#0 lore raw

cover Bug fixes related to secure 2 stage translation

Idan Horowitz (3):
target/arm: Check VSTCR.SW when assigning the stage 2 output PA space target/arm: Take VSTCR.SW, VTCR.NSW into account in final stage 2 walk target/arm: Determine final stage 2 output PA space based on original
IPA

 target/arm/helper.c | 16 ++++++++++++++--
 1 file changed, 14 insertions(+), 2 deletions(-)
Signature · 1 line
2.35.1
#1 lore raw

PATCH 1/3 target/arm: Check VSTCR.SW when assigning the stage 2 output PA space

As per the AArch64.SS2OutputPASpace() psuedo-code in the ARMv8 ARM when the PA space of the IPA is non secure, the output PA space is secure if and only if all of the bits VTCR.<NSW, NSA>, VSTCR.<SW, SA> are not set.

---
 target/arm/helper.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
target/arm/helper.c +1 -1
index 812ca591f4..d0265b760f 100644
@@ -12697,7 +12697,7 @@ bool get_phys_addr(CPUARMState *env, target_ulong address,
                 } else {
                     attrs->secure =
                         !((env->cp15.vtcr_el2.raw_tcr & (VTCR_NSA | VTCR_NSW))
-                        || (env->cp15.vstcr_el2.raw_tcr & VSTCR_SA));
+                        || (env->cp15.vstcr_el2.raw_tcr & (VSTCR_SA | VSTCR_SW)));
                 }
             }
             return 0;
Signature · 1 line
2.35.1
#2 lore raw

Re: [PATCH 1/3] target/arm: Check VSTCR.SW when assigning the stage 2 output PA space

Quoted 17 lines · Le sunnuntaina 27. maaliskuuta 2022, 12.34.26 EEST Idan Horowitz a écrit :

As per the AArch64.SS2OutputPASpace() psuedo-code in the ARMv8 ARM when the PA space of the IPA is non secure, the output PA space is secure if and only if all of the bits VTCR.<NSW, NSA>, VSTCR.<SW, SA> are not set.

---
 target/arm/helper.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
target/arm/helper.c +0 -0
index 812ca591f4..d0265b760f 100644
@@ -12697,7 +12697,7 @@ bool get_phys_addr(CPUARMState *env, target_ulong
 address, } else {
                     attrs->secure =
                         !((env->cp15.vtcr_el2.raw_tcr & (VTCR_NSA |
Quoted 3 lines

VTCR_NSW)) - || (env->cp15.vstcr_el2.raw_tcr &
VSTCR_SA)); + || (env->cp15.vstcr_el2.raw_tcr &
(VSTCR_SA | VSTCR_SW))); }

The VTCR_EL2 specification says that the NSA bit "behaves as 1 for all purposes other than reading back the value of the bit when one of the following is true (...)
* The value of VTCR_EL2.NSW is 1.
* The value of VSTCR_EL2.SA is 1."

Sorry but I don't see any reason to check the SW bit here.

Signature · 2 lines
Реми Дёни-Курмон
http://www.remlab.net/
#3 lore raw

Re: [PATCH 1/3] target/arm: Check VSTCR.SW when assigning the stage 2 output PA space

Quoted 28 lines · Rémi Denis-Courmont
Quoted 17 lines · Le sunnuntaina 27. maaliskuuta 2022, 12.34.26 EEST Idan Horowitz a écrit :

As per the AArch64.SS2OutputPASpace() psuedo-code in the ARMv8 ARM when the PA space of the IPA is non secure, the output PA space is secure if and only if all of the bits VTCR.<NSW, NSA>, VSTCR.<SW, SA> are not set.

---
  target/arm/helper.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
target/arm/helper.c +0 -0
index 812ca591f4..d0265b760f 100644
@@ -12697,7 +12697,7 @@ bool get_phys_addr(CPUARMState *env, target_ulong
 address, } else {
                      attrs->secure =
                          !((env->cp15.vtcr_el2.raw_tcr & (VTCR_NSA |
Quoted 3 lines

VTCR_NSW)) - || (env->cp15.vstcr_el2.raw_tcr &
VSTCR_SA)); + || (env->cp15.vstcr_el2.raw_tcr &
(VSTCR_SA | VSTCR_SW))); }

The VTCR_EL2 specification says that the NSA bit "behaves as 1 for all purposes other than reading back the value of the bit when one of the following is true (...)
* The value of VTCR_EL2.NSW is 1.
* The value of VSTCR_EL2.SA is 1."

Quoted 1 line

Sorry but I don't see any reason to check the SW bit here.

Because the description of SA says that it behaves as 1 if SW is 1.

r~

Reviewed-by on v1 1/3
#4 lore raw

Re: [PATCH 1/3] target/arm: Check VSTCR.SW when assigning the stage 2 output PA space

In reply to Idan Horowitz
Quoted 8 lines · Idan Horowitz

As per the AArch64.SS2OutputPASpace() psuedo-code in the ARMv8 ARM when the PA space of the IPA is non secure, the output PA space is secure if and only if all of the bits VTCR.<NSW, NSA>, VSTCR.<SW, SA> are not set.

---
  target/arm/helper.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)

r~

#5 lore raw

PATCH 2/3 target/arm: Take VSTCR.SW, VTCR.NSW into account in final stage 2 walk

In reply to Idan Horowitz

As per the AArch64.SS2InitialTTWState() psuedo-code in the ARMv8 ARM the initial PA space used for stage 2 table walks is assigned based on the SW and NSW bits of the VSTCR and VTCR registers.
This was already implemented for the recursive stage 2 page table walks in S1_ptw_translate(), but was missing for the final stage 2 walk.

---
 target/arm/helper.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
target/arm/helper.c +10 -0
index d0265b760f..e2695e846a 100644
@@ -12657,6 +12657,16 @@ bool get_phys_addr(CPUARMState *env, target_ulong address,
                 return ret;
             }
 
+            if (arm_is_secure_below_el3(env)) {
+                if (attrs->secure) {
+                    attrs->secure = !(env->cp15.vstcr_el2.raw_tcr & VSTCR_SW);
+                } else {
+                    attrs->secure = !(env->cp15.vtcr_el2.raw_tcr & VTCR_NSW);
+                }
+            } else {
+                assert(!attrs->secure);
+            }
+
             s2_mmu_idx = attrs->secure ? ARMMMUIdx_Stage2_S : ARMMMUIdx_Stage2;
             is_el0 = mmu_idx == ARMMMUIdx_E10_0 || mmu_idx == ARMMMUIdx_SE10_0;
 
Signature · 1 line
2.35.1
Reviewed-by on v1 2/3
#6 lore raw

Re: [PATCH 2/3] target/arm: Take VSTCR.SW, VTCR.NSW into account in final stage 2 walk

Quoted 10 lines · Idan Horowitz

As per the AArch64.SS2InitialTTWState() psuedo-code in the ARMv8 ARM the initial PA space used for stage 2 table walks is assigned based on the SW and NSW bits of the VSTCR and VTCR registers.
This was already implemented for the recursive stage 2 page table walks in S1_ptw_translate(), but was missing for the final stage 2 walk.

---
  target/arm/helper.c | 10 ++++++++++
  1 file changed, 10 insertions(+)

r~

#7 lore raw

PATCH 3/3 target/arm: Determine final stage 2 output PA space based on original IPA

In reply to Idan Horowitz

As per the AArch64.S2Walk() psuedo-code in the ARMv8 ARM, the final decision as to the output address's PA space based on the SA/SW/NSA/NSA bits needs to take the input IPA's PA space into account, and not the PA space of the result of the stage 2 walk itself.

---
 target/arm/helper.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)
target/arm/helper.c +5 -3
index e2695e846a..16c2628f8f 100644
@@ -12644,6 +12644,7 @@ bool get_phys_addr(CPUARMState *env, target_ulong address,
             hwaddr ipa;
             int s2_prot;
             int ret;
+            bool ipa_secure;
             ARMCacheAttrs cacheattrs2 = {};
             ARMMMUIdx s2_mmu_idx;
             bool is_el0;
@@ -12657,14 +12658,15 @@ bool get_phys_addr(CPUARMState *env, target_ulong address,
                 return ret;
             }
 
+            ipa_secure = attrs->secure;
             if (arm_is_secure_below_el3(env)) {
-                if (attrs->secure) {
+                if (ipa_secure) {
                     attrs->secure = !(env->cp15.vstcr_el2.raw_tcr & VSTCR_SW);
                 } else {
                     attrs->secure = !(env->cp15.vtcr_el2.raw_tcr & VTCR_NSW);
                 }
             } else {
-                assert(!attrs->secure);
+                assert(!ipa_secure);
             }
 
             s2_mmu_idx = attrs->secure ? ARMMMUIdx_Stage2_S : ARMMMUIdx_Stage2;
@@ -12701,7 +12703,7 @@ bool get_phys_addr(CPUARMState *env, target_ulong address,
 
             /* Check if IPA translates to secure or non-secure PA space. */
             if (arm_is_secure_below_el3(env)) {
-                if (attrs->secure) {
+                if (ipa_secure) {
                     attrs->secure =
                         !(env->cp15.vstcr_el2.raw_tcr & (VSTCR_SA | VSTCR_SW));
                 } else {
Signature · 1 line
2.35.1
Reviewed-by on v1 3/3
#8 lore raw

Re: [PATCH 3/3] target/arm: Determine final stage 2 output PA space based on original IPA

Quoted 6 lines · Idan Horowitz

As per the AArch64.S2Walk() psuedo-code in the ARMv8 ARM, the final decision as to the output address's PA space based on the SA/SW/NSA/NSA bits needs to take the input IPA's PA space into account, and not the PA space of the result of the stage 2 walk itself.

I believe I follow: because the walk uses walkstate.address.paddress.paspace, the ipa input parameter is unchanged, and it is ipa that is passed to AArch64.S2NextWalkStateLast() to form the output address.

r~

#9 lore raw

Re: [PATCH 3/3] target/arm: Determine final stage 2 output PA space based on original IPA

Quoted 5 lines · Richard Henderson

I believe I follow: because the walk uses walkstate.address.paddress.paspace, the ipa input parameter is unchanged, and it is ipa that is passed to AArch64.S2NextWalkStateLast() to form the output address.

Indeed, I initially found the issue when a test case passed on arm IP but raised a data abort in QEMU.
Since fixing this issue solved the inconsistency, I believe this is the intended behaviour and not a spec bug.

Quoted 2 lines

r~

Idan Horowitz

#10 lore raw

Re: [PATCH 3/3] target/arm: Determine final stage 2 output PA space based on original IPA

In reply to Richard Henderson
Quoted 7 lines · Richard Henderson
Quoted 5 lines · Idan Horowitz

As per the AArch64.S2Walk() psuedo-code in the ARMv8 ARM, the final decision as to the output address's PA space based on the SA/SW/NSA/NSA bits needs to take the input IPA's PA space into account, and not the PA space of the result of the stage 2 walk itself.

Quoted 5 lines
Quoted 1 line

I believe I follow: because the walk uses walkstate.address.paddress.paspace, the ipa input parameter is unchanged, and it is ipa that is passed to AArch64.S2NextWalkStateLast() to form the output address.

Textually, this is described on page D5-4802 of DDI 0487H.a; the security of the output address of the memory access isn't affected by the security of the output address of the translation table walk.

-- PMM

#11 lore raw
In reply to Idan Horowitz
Quoted 8 lines · Idan Horowitz

Idan Horowitz (3):
target/arm: Check VSTCR.SW when assigning the stage 2 output PA space target/arm: Take VSTCR.SW, VTCR.NSW into account in final stage 2 walk target/arm: Determine final stage 2 output PA space based on original
IPA

Applied to target-arm.next, thanks. (I fixed a couple of typos in the patch 3 commit message.)

-- PMM