OSDN Git Service

target/arm: Use correct mmu_idx for exception-return unstacking
authorPeter Maydell <peter.maydell@linaro.org>
Mon, 9 Jul 2018 12:45:35 +0000 (13:45 +0100)
committerPeter Maydell <peter.maydell@linaro.org>
Tue, 10 Jul 2018 09:54:40 +0000 (10:54 +0100)
For M-profile exception returns, the mmu index to use for exception
return unstacking is supposed to be that of wherever we are returning to:
 * if returning to handler mode, privileged
 * if returning to thread mode, privileged or unprivileged depending on
   CONTROL.nPRIV for the destination security state

We were passing the wrong thing as the 'priv' argument to
arm_v7m_mmu_idx_for_secstate_and_priv(). The effect was that guests
which programmed the MPU to behave differently for privileged and
unprivileged code could get spurious MemManage Unstack exceptions.

Reported-by: Adithya Baglody <adithya.nagaraj.baglody@intel.com>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180709124535.1116-1-peter.maydell@linaro.org

target/arm/helper.c

index a2ac960..0604a0e 100644 (file)
@@ -7171,9 +7171,11 @@ static void do_v7m_exception_exit(ARMCPU *cpu)
         uint32_t frameptr = *frame_sp_p;
         bool pop_ok = true;
         ARMMMUIdx mmu_idx;
+        bool return_to_priv = return_to_handler ||
+            !(env->v7m.control[return_to_secure] & R_V7M_CONTROL_NPRIV_MASK);
 
         mmu_idx = arm_v7m_mmu_idx_for_secstate_and_priv(env, return_to_secure,
-                                                        !return_to_handler);
+                                                        return_to_priv);
 
         if (!QEMU_IS_ALIGNED(frameptr, 8) &&
             arm_feature(env, ARM_FEATURE_V8)) {