[Qemu-devel] [PATCH 0/9] v8M: BLXNS, SG, secure function return

[Qemu-devel] [PATCH 0/9] v8M: BLXNS, SG, secure function return

[Peter Maydell]

This patchset is mostly implementation of SG, BLXNS and secure
function return. Parts of it were in the previous patchset,
but had a bug which I've fixed in the SG implementation where
we read the first half of the insn twice by mistake.

Patch 1 is a new trivial bugfix.
Patches 2-4 were in the previous set; only 2 has changed.
Patches 5-8 perform some refactoring so that we can correctly
implement the behaviour of a handful of Thumb instructions
which must be unconditional even when they appear inside
an IT block: SG, HLT and BKPT.
Patch 9 implements the final edge cases of the SG insn.

The refactoring has the nice property that we don't have to
pass the CPU env pointer into the bulk of the thumb decode
functions any more. I suspect it also might fix bugs in
setting breakpoints on instructions inside IT blocks, though
we haven't had any complaints about that so perhaps gdb
works around it by setting breakpoints on both the cc pass
and cc fail next instructions in an IT block...

Peter Maydell (9):
  target/arm: Add M profile secure MMU index values to
    get_a32_user_mem_index()
  target/arm: Implement SG instruction
  target/arm: Implement BLXNS
  target/arm: Implement secure function return
  target-arm: Don't check for "Thumb2 or M profile" for not-Thumb1
  target/arm: Pull Thumb insn word loads up to top level
  target-arm: Simplify insn_crosses_page()
  target/arm: Support some Thumb insns being always unconditional
  target/arm: Implement SG instruction corner cases

 target/arm/helper.h    |   1 +
 target/arm/internals.h |   8 ++
 target/arm/helper.c    | 306 +++++++++++++++++++++++++++++++++++++++++++++---
 target/arm/translate.c | 310 +++++++++++++++++++++++++++++++++----------------
 4 files changed, 515 insertions(+), 110 deletions(-)

-- 
2.7.4

[Qemu-devel] [PATCH 1/9] target/arm: Add M profile secure MMU index values to get_a32_user_mem_index()

[Peter Maydell]

Add the M profile secure MMU index values to the switch in
get_a32_user_mem_index() so that LDRT/STRT work correctly
rather than asserting at translate time.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
---
 target/arm/translate.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/target/arm/translate.c b/target/arm/translate.c
index ab1a12a..e1b83b7 100644
--- a/target/arm/translate.c
+++ b/target/arm/translate.c
@@ -165,6 +165,10 @@ static inline int get_a32_user_mem_index(DisasContext *s)
     case ARMMMUIdx_MPriv:
     case ARMMMUIdx_MNegPri:
         return arm_to_core_mmu_idx(ARMMMUIdx_MUser);
+    case ARMMMUIdx_MSUser:
+    case ARMMMUIdx_MSPriv:
+    case ARMMMUIdx_MSNegPri:
+        return arm_to_core_mmu_idx(ARMMMUIdx_MSUser);
     case ARMMMUIdx_S2NS:
     default:
         g_assert_not_reached();
-- 
2.7.4

[Qemu-devel] [PATCH 2/9] target/arm: Implement SG instruction

[Peter Maydell]

Implement the SG instruction, which we emulate 'by hand' in the
exception handling code path.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
---
 target/arm/helper.c | 132 ++++++++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 127 insertions(+), 5 deletions(-)

diff --git a/target/arm/helper.c b/target/arm/helper.c
index 1d689f0..9cc881e 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -41,6 +41,10 @@ typedef struct V8M_SAttributes {
     bool irvalid;
 } V8M_SAttributes;
 
+static void v8m_security_lookup(CPUARMState *env, uint32_t address,
+                                MMUAccessType access_type, ARMMMUIdx mmu_idx,
+                                V8M_SAttributes *sattrs);
+
 /* Definitions for the PMCCNTR and PMCR registers */
 #define PMCRD   0x8
 #define PMCRC   0x4
@@ -6736,6 +6740,126 @@ static void arm_log_exception(int idx)
     }
 }
 
+static bool v7m_read_half_insn(ARMCPU *cpu, ARMMMUIdx mmu_idx,
+                               uint32_t addr, uint16_t *insn)
+{
+    /* Load a 16-bit portion of a v7M instruction, returning true on success,
+     * or false on failure (in which case we will have pended the appropriate
+     * exception).
+     * We need to do the instruction fetch's MPU and SAU checks
+     * like this because there is no MMU index that would allow
+     * doing the load with a single function call. Instead we must
+     * first check that the security attributes permit the load
+     * and that they don't mismatch on the two halves of the instruction,
+     * and then we do the load as a secure load (ie using the security
+     * attributes of the address, not the CPU, as architecturally required).
+     */
+    CPUState *cs = CPU(cpu);
+    CPUARMState *env = &cpu->env;
+    V8M_SAttributes sattrs = {};
+    MemTxAttrs attrs = {};
+    ARMMMUFaultInfo fi = {};
+    MemTxResult txres;
+    target_ulong page_size;
+    hwaddr physaddr;
+    int prot;
+    uint32_t fsr;
+
+    v8m_security_lookup(env, addr, MMU_INST_FETCH, mmu_idx, &sattrs);
+    if (!sattrs.nsc || sattrs.ns) {
+        /* This must be the second half of the insn, and it straddles a
+         * region boundary with the second half not being S&NSC.
+         */
+        env->v7m.sfsr |= R_V7M_SFSR_INVEP_MASK;
+        armv7m_nvic_set_pending(env->nvic, ARMV7M_EXCP_SECURE, false);
+        qemu_log_mask(CPU_LOG_INT,
+                      "...really SecureFault with SFSR.INVEP\n");
+        return false;
+    }
+    if (get_phys_addr(env, addr, MMU_INST_FETCH, mmu_idx,
+                      &physaddr, &attrs, &prot, &page_size, &fsr, &fi)) {
+        /* the MPU lookup failed */
+        env->v7m.cfsr[env->v7m.secure] |= R_V7M_CFSR_IACCVIOL_MASK;
+        armv7m_nvic_set_pending(env->nvic, ARMV7M_EXCP_MEM, env->v7m.secure);
+        qemu_log_mask(CPU_LOG_INT, "...really MemManage with CFSR.IACCVIOL\n");
+        return false;
+    }
+    *insn = address_space_lduw_le(arm_addressspace(cs, attrs), physaddr,
+                                 attrs, &txres);
+    if (txres != MEMTX_OK) {
+        env->v7m.cfsr[M_REG_NS] |= R_V7M_CFSR_IBUSERR_MASK;
+        armv7m_nvic_set_pending(env->nvic, ARMV7M_EXCP_BUS, false);
+        qemu_log_mask(CPU_LOG_INT, "...really BusFault with CFSR.IBUSERR\n");
+        return false;
+    }
+    return true;
+}
+
+static bool v7m_handle_execute_nsc(ARMCPU *cpu)
+{
+    /* Check whether this attempt to execute code in a Secure & NS-Callable
+     * memory region is for an SG instruction; if so, then emulate the
+     * effect of the SG instruction and return true. Otherwise pend
+     * the correct kind of exception and return false.
+     */
+    CPUARMState *env = &cpu->env;
+    ARMMMUIdx mmu_idx;
+    uint16_t insn;
+
+    /* We should never get here unless get_phys_addr_pmsav8() caused
+     * an exception for NS executing in S&NSC memory.
+     */
+    assert(!env->v7m.secure);
+    assert(arm_feature(env, ARM_FEATURE_M_SECURITY));
+
+    /* We want to do the MPU lookup as secure; work out what mmu_idx that is */
+    mmu_idx = arm_v7m_mmu_idx_for_secstate(env, true);
+
+    if (!v7m_read_half_insn(cpu, mmu_idx, env->regs[15], &insn)) {
+        return false;
+    }
+
+    if (!env->thumb) {
+        goto gen_invep;
+    }
+
+    if (insn != 0xe97f) {
+        /* Not an SG instruction first half (we choose the IMPDEF
+         * early-SG-check option).
+         */
+        goto gen_invep;
+    }
+
+    if (!v7m_read_half_insn(cpu, mmu_idx, env->regs[15] + 2, &insn)) {
+        return false;
+    }
+
+    if (insn != 0xe97f) {
+        /* Not an SG instruction second half (yes, both halves of the SG
+         * insn have the same hex value)
+         */
+        goto gen_invep;
+    }
+
+    /* OK, we have confirmed that we really have an SG instruction.
+     * We know we're NS in S memory so don't need to repeat those checks.
+     */
+    qemu_log_mask(CPU_LOG_INT, "...really an SG instruction at 0x%08" PRIx32
+                  ", executing it\n", env->regs[15]);
+    env->regs[14] &= ~1;
+    switch_v7m_security_state(env, true);
+    xpsr_write(env, 0, XPSR_IT);
+    env->regs[15] += 4;
+    return true;
+
+gen_invep:
+    env->v7m.sfsr |= R_V7M_SFSR_INVEP_MASK;
+    armv7m_nvic_set_pending(env->nvic, ARMV7M_EXCP_SECURE, false);
+    qemu_log_mask(CPU_LOG_INT,
+                  "...really SecureFault with SFSR.INVEP\n");
+    return false;
+}
+
 void arm_v7m_cpu_do_interrupt(CPUState *cs)
 {
     ARMCPU *cpu = ARM_CPU(cs);
@@ -6778,12 +6902,10 @@ void arm_v7m_cpu_do_interrupt(CPUState *cs)
              * the SG instruction have the same security attributes.)
              * Everything else must generate an INVEP SecureFault, so we
              * emulate the SG instruction here.
-             * TODO: actually emulate SG.
              */
-            env->v7m.sfsr |= R_V7M_SFSR_INVEP_MASK;
-            armv7m_nvic_set_pending(env->nvic, ARMV7M_EXCP_SECURE, false);
-            qemu_log_mask(CPU_LOG_INT,
-                          "...really SecureFault with SFSR.INVEP\n");
+            if (v7m_handle_execute_nsc(cpu)) {
+                return;
+            }
             break;
         case M_FAKE_FSR_SFAULT:
             /* Various flavours of SecureFault for attempts to execute or
-- 
2.7.4

[Qemu-devel] [PATCH 3/9] target/arm: Implement BLXNS

[Peter Maydell]

Implement the BLXNS instruction, which allows secure code to
call non-secure code.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 target/arm/helper.h    |  1 +
 target/arm/internals.h |  1 +
 target/arm/helper.c    | 59 ++++++++++++++++++++++++++++++++++++++++++++++++++
 target/arm/translate.c | 17 +++++++++++++--
 4 files changed, 76 insertions(+), 2 deletions(-)

diff --git a/target/arm/helper.h b/target/arm/helper.h
index 64afbac..2cf6f74 100644
--- a/target/arm/helper.h
+++ b/target/arm/helper.h
@@ -64,6 +64,7 @@ DEF_HELPER_3(v7m_msr, void, env, i32, i32)
 DEF_HELPER_2(v7m_mrs, i32, env, i32)
 
 DEF_HELPER_2(v7m_bxns, void, env, i32)
+DEF_HELPER_2(v7m_blxns, void, env, i32)
 
 DEF_HELPER_4(access_check_cp_reg, void, env, ptr, i32, i32)
 DEF_HELPER_3(set_cp_reg, void, env, ptr, i32)
diff --git a/target/arm/internals.h b/target/arm/internals.h
index fd9a7e8..1746737 100644
--- a/target/arm/internals.h
+++ b/target/arm/internals.h
@@ -60,6 +60,7 @@ static inline bool excp_is_internal(int excp)
 FIELD(V7M_CONTROL, NPRIV, 0, 1)
 FIELD(V7M_CONTROL, SPSEL, 1, 1)
 FIELD(V7M_CONTROL, FPCA, 2, 1)
+FIELD(V7M_CONTROL, SFPA, 3, 1)
 
 /* Bit definitions for v7M exception return payload */
 FIELD(V7M_EXCRET, ES, 0, 1)
diff --git a/target/arm/helper.c b/target/arm/helper.c
index 9cc881e..47c5767 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -5897,6 +5897,12 @@ void HELPER(v7m_bxns)(CPUARMState *env, uint32_t dest)
     g_assert_not_reached();
 }
 
+void HELPER(v7m_blxns)(CPUARMState *env, uint32_t dest)
+{
+    /* translate.c should never generate calls here in user-only mode */
+    g_assert_not_reached();
+}
+
 void switch_mode(CPUARMState *env, int mode)
 {
     ARMCPU *cpu = arm_env_get_cpu(env);
@@ -6189,6 +6195,59 @@ void HELPER(v7m_bxns)(CPUARMState *env, uint32_t dest)
     env->regs[15] = dest & ~1;
 }
 
+void HELPER(v7m_blxns)(CPUARMState *env, uint32_t dest)
+{
+    /* Handle v7M BLXNS:
+     *  - bit 0 of the destination address is the target security state
+     */
+
+    /* At this point regs[15] is the address just after the BLXNS */
+    uint32_t nextinst = env->regs[15] | 1;
+    uint32_t sp = env->regs[13] - 8;
+    uint32_t saved_psr;
+
+    /* translate.c will have made BLXNS UNDEF unless we're secure */
+    assert(env->v7m.secure);
+
+    if (dest & 1) {
+        /* target is Secure, so this is just a normal BLX,
+         * except that the low bit doesn't indicate Thumb/not.
+         */
+        env->regs[14] = nextinst;
+        env->thumb = 1;
+        env->regs[15] = dest & ~1;
+        return;
+    }
+
+    /* Target is non-secure: first push a stack frame */
+    if (!QEMU_IS_ALIGNED(sp, 8)) {
+        qemu_log_mask(LOG_GUEST_ERROR,
+                      "BLXNS with misaligned SP is UNPREDICTABLE\n");
+    }
+
+    saved_psr = env->v7m.exception;
+    if (env->v7m.control[M_REG_S] & R_V7M_CONTROL_SFPA_MASK) {
+        saved_psr |= XPSR_SFPA;
+    }
+
+    /* Note that these stores can throw exceptions on MPU faults */
+    cpu_stl_data(env, sp, nextinst);
+    cpu_stl_data(env, sp + 4, saved_psr);
+
+    env->regs[13] = sp;
+    env->regs[14] = 0xfeffffff;
+    if (arm_v7m_is_handler_mode(env)) {
+        /* Write a dummy value to IPSR, to avoid leaking the current secure
+         * exception number to non-secure code. This is guaranteed not
+         * to cause write_v7m_exception() to actually change stacks.
+         */
+        write_v7m_exception(env, 1);
+    }
+    switch_v7m_security_state(env, 0);
+    env->thumb = 1;
+    env->regs[15] = dest;
+}
+
 static uint32_t *get_v7m_sp_ptr(CPUARMState *env, bool secure, bool threadmode,
                                 bool spsel)
 {
diff --git a/target/arm/translate.c b/target/arm/translate.c
index e1b83b7..933a52f 100644
--- a/target/arm/translate.c
+++ b/target/arm/translate.c
@@ -1017,6 +1017,20 @@ static inline void gen_bxns(DisasContext *s, int rm)
     s->base.is_jmp = DISAS_EXIT;
 }
 
+static inline void gen_blxns(DisasContext *s, int rm)
+{
+    TCGv_i32 var = load_reg(s, rm);
+
+    /* We don't need to sync condexec state, for the same reason as bxns.
+     * We do however need to set the PC, because the blxns helper reads it.
+     * The blxns helper may throw an exception.
+     */
+    gen_set_pc_im(s, s->pc);
+    gen_helper_v7m_blxns(cpu_env, var);
+    tcg_temp_free_i32(var);
+    s->base.is_jmp = DISAS_EXIT;
+}
+
 /* Variant of store_reg which uses branch&exchange logic when storing
    to r15 in ARM architecture v7 and above. The source must be a temporary
    and will be marked as dead. */
@@ -11225,8 +11239,7 @@ static void disas_thumb_insn(CPUARMState *env, DisasContext *s)
                         goto undef;
                     }
                     if (link) {
-                        /* BLXNS: not yet implemented */
-                        goto undef;
+                        gen_blxns(s, rm);
                     } else {
                         gen_bxns(s, rm);
                     }
-- 
2.7.4

[Qemu-devel] [PATCH 4/9] target/arm: Implement secure function return

[Peter Maydell]

Secure function return happens when a non-secure function has been
called using BLXNS and so has a particular magic LR value (either
0xfefffffe or 0xfeffffff). The function return via BX behaves
specially when the new PC value is this magic value, in the same
way that exception returns are handled.

Adjust our BX excret guards so that they recognize the function
return magic number as well, and perform the function-return
unstacking in do_v7m_exception_exit().

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Acked-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 target/arm/internals.h |   7 +++
 target/arm/helper.c    | 115 +++++++++++++++++++++++++++++++++++++++++++++----
 target/arm/translate.c |  14 +++++-
 3 files changed, 126 insertions(+), 10 deletions(-)

diff --git a/target/arm/internals.h b/target/arm/internals.h
index 1746737..43106a2 100644
--- a/target/arm/internals.h
+++ b/target/arm/internals.h
@@ -72,6 +72,13 @@ FIELD(V7M_EXCRET, DCRS, 5, 1)
 FIELD(V7M_EXCRET, S, 6, 1)
 FIELD(V7M_EXCRET, RES1, 7, 25) /* including the must-be-1 prefix */
 
+/* Minimum value which is a magic number for exception return */
+#define EXC_RETURN_MIN_MAGIC 0xff000000
+/* Minimum number which is a magic number for function or exception return
+ * when using v8M security extension
+ */
+#define FNC_RETURN_MIN_MAGIC 0xfefffffe
+
 /* We use a few fake FSR values for internal purposes in M profile.
  * M profile cores don't have A/R format FSRs, but currently our
  * get_phys_addr() code assumes A/R profile and reports failures via
diff --git a/target/arm/helper.c b/target/arm/helper.c
index 47c5767..96113fe 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -6174,7 +6174,17 @@ void HELPER(v7m_bxns)(CPUARMState *env, uint32_t dest)
      *  - if the return value is a magic value, do exception return (like BX)
      *  - otherwise bit 0 of the return value is the target security state
      */
-    if (dest >= 0xff000000) {
+    uint32_t min_magic;
+
+    if (arm_feature(env, ARM_FEATURE_M_SECURITY)) {
+        /* Covers FNC_RETURN and EXC_RETURN magic */
+        min_magic = FNC_RETURN_MIN_MAGIC;
+    } else {
+        /* EXC_RETURN magic only */
+        min_magic = EXC_RETURN_MIN_MAGIC;
+    }
+
+    if (dest >= min_magic) {
         /* This is an exception return magic value; put it where
          * do_v7m_exception_exit() expects and raise EXCEPTION_EXIT.
          * Note that if we ever add gen_ss_advance() singlestep support to
@@ -6470,12 +6480,19 @@ static void do_v7m_exception_exit(ARMCPU *cpu)
     bool exc_secure = false;
     bool return_to_secure;
 
-    /* We can only get here from an EXCP_EXCEPTION_EXIT, and
-     * gen_bx_excret() enforces the architectural rule
-     * that jumps to magic addresses don't have magic behaviour unless
-     * we're in Handler mode (compare pseudocode BXWritePC()).
+    /* If we're not in Handler mode then jumps to magic exception-exit
+     * addresses don't have magic behaviour. However for the v8M
+     * security extensions the magic secure-function-return has to
+     * work in thread mode too, so to avoid doing an extra check in
+     * the generated code we allow exception-exit magic to also cause the
+     * internal exception and bring us here in thread mode. Correct code
+     * will never try to do this (the following insn fetch will always
+     * fault) so we the overhead of having taken an unnecessary exception
+     * doesn't matter.
      */
-    assert(arm_v7m_is_handler_mode(env));
+    if (!arm_v7m_is_handler_mode(env)) {
+        return;
+    }
 
     /* In the spec pseudocode ExceptionReturn() is called directly
      * from BXWritePC() and gets the full target PC value including
@@ -6765,6 +6782,78 @@ static void do_v7m_exception_exit(ARMCPU *cpu)
     qemu_log_mask(CPU_LOG_INT, "...successful exception return\n");
 }
 
+static bool do_v7m_function_return(ARMCPU *cpu)
+{
+    /* v8M security extensions magic function return.
+     * We may either:
+     *  (1) throw an exception (longjump)
+     *  (2) return true if we successfully handled the function return
+     *  (3) return false if we failed a consistency check and have
+     *      pended a UsageFault that needs to be taken now
+     *
+     * At this point the magic return value is split between env->regs[15]
+     * and env->thumb. We don't bother to reconstitute it because we don't
+     * need it (all values are handled the same way).
+     */
+    CPUARMState *env = &cpu->env;
+    uint32_t newpc, newpsr, newpsr_exc;
+
+    qemu_log_mask(CPU_LOG_INT, "...really v7M secure function return\n");
+
+    {
+        bool threadmode, spsel;
+        TCGMemOpIdx oi;
+        ARMMMUIdx mmu_idx;
+        uint32_t *frame_sp_p;
+        uint32_t frameptr;
+
+        /* Pull the return address and IPSR from the Secure stack */
+        threadmode = !arm_v7m_is_handler_mode(env);
+        spsel = env->v7m.control[M_REG_S] & R_V7M_CONTROL_SPSEL_MASK;
+
+        frame_sp_p = get_v7m_sp_ptr(env, true, threadmode, spsel);
+        frameptr = *frame_sp_p;
+
+        /* These loads may throw an exception (for MPU faults). We want to
+         * do them as secure, so work out what MMU index that is.
+         */
+        mmu_idx = arm_v7m_mmu_idx_for_secstate(env, true);
+        oi = make_memop_idx(MO_LE, arm_to_core_mmu_idx(mmu_idx));
+        newpc = helper_le_ldul_mmu(env, frameptr, oi, 0);
+        newpsr = helper_le_ldul_mmu(env, frameptr + 4, oi, 0);
+
+        /* Consistency checks on new IPSR */
+        newpsr_exc = newpsr & XPSR_EXCP;
+        if (!((env->v7m.exception == 0 && newpsr_exc == 0) ||
+              (env->v7m.exception == 1 && newpsr_exc != 0))) {
+            /* Pend the fault and tell our caller to take it */
+            env->v7m.cfsr[env->v7m.secure] |= R_V7M_CFSR_INVPC_MASK;
+            armv7m_nvic_set_pending(env->nvic, ARMV7M_EXCP_USAGE,
+                                    env->v7m.secure);
+            qemu_log_mask(CPU_LOG_INT,
+                          "...taking INVPC UsageFault: "
+                          "IPSR consistency check failed\n");
+            return false;
+        }
+
+        *frame_sp_p = frameptr + 8;
+    }
+
+    /* This invalidates frame_sp_p */
+    switch_v7m_security_state(env, true);
+    env->v7m.exception = newpsr_exc;
+    env->v7m.control[M_REG_S] &= ~R_V7M_CONTROL_SFPA_MASK;
+    if (newpsr & XPSR_SFPA) {
+        env->v7m.control[M_REG_S] |= R_V7M_CONTROL_SFPA_MASK;
+    }
+    xpsr_write(env, 0, XPSR_IT);
+    env->thumb = newpc & 1;
+    env->regs[15] = newpc & ~1;
+
+    qemu_log_mask(CPU_LOG_INT, "...function return successful\n");
+    return true;
+}
+
 static void arm_log_exception(int idx)
 {
     if (qemu_loglevel_mask(CPU_LOG_INT)) {
@@ -7049,8 +7138,18 @@ void arm_v7m_cpu_do_interrupt(CPUState *cs)
     case EXCP_IRQ:
         break;
     case EXCP_EXCEPTION_EXIT:
-        do_v7m_exception_exit(cpu);
-        return;
+        if (env->regs[15] < EXC_RETURN_MIN_MAGIC) {
+            /* Must be v8M security extension function return */
+            assert(env->regs[15] >= FNC_RETURN_MIN_MAGIC);
+            assert(arm_feature(env, ARM_FEATURE_M_SECURITY));
+            if (do_v7m_function_return(cpu)) {
+                return;
+            }
+        } else {
+            do_v7m_exception_exit(cpu);
+            return;
+        }
+        break;
     default:
         cpu_abort(cs, "Unhandled exception 0x%x\n", cs->exception_index);
         return; /* Never happens.  Keep compiler happy.  */
diff --git a/target/arm/translate.c b/target/arm/translate.c
index 933a52f..58d706c 100644
--- a/target/arm/translate.c
+++ b/target/arm/translate.c
@@ -964,7 +964,8 @@ static inline void gen_bx_excret(DisasContext *s, TCGv_i32 var)
      * s->base.is_jmp that we need to do the rest of the work later.
      */
     gen_bx(s, var);
-    if (s->v7m_handler_mode && arm_dc_feature(s, ARM_FEATURE_M)) {
+    if (arm_dc_feature(s, ARM_FEATURE_M_SECURITY) ||
+        (s->v7m_handler_mode && arm_dc_feature(s, ARM_FEATURE_M))) {
         s->base.is_jmp = DISAS_BX_EXCRET;
     }
 }
@@ -973,9 +974,18 @@ static inline void gen_bx_excret_final_code(DisasContext *s)
 {
     /* Generate the code to finish possible exception return and end the TB */
     TCGLabel *excret_label = gen_new_label();
+    uint32_t min_magic;
+
+    if (arm_dc_feature(s, ARM_FEATURE_M_SECURITY)) {
+        /* Covers FNC_RETURN and EXC_RETURN magic */
+        min_magic = FNC_RETURN_MIN_MAGIC;
+    } else {
+        /* EXC_RETURN magic only */
+        min_magic = EXC_RETURN_MIN_MAGIC;
+    }
 
     /* Is the new PC value in the magic range indicating exception return? */
-    tcg_gen_brcondi_i32(TCG_COND_GEU, cpu_R[15], 0xff000000, excret_label);
+    tcg_gen_brcondi_i32(TCG_COND_GEU, cpu_R[15], min_magic, excret_label);
     /* No: end the TB as we would for a DISAS_JMP */
     if (is_singlestepping(s)) {
         gen_singlestep_exception(s);
-- 
2.7.4

[Qemu-devel] [PATCH 5/9] target-arm: Don't check for "Thumb2 or M profile" for not-Thumb1

[Peter Maydell]

The code which implements the Thumb1 split BL/BLX instructions
is guarded by a check on "not M or THUMB2". All we really need
to check here is "not THUMB2" (and we assume that elsewhere too,
eg in the ARCH(6T2) test that UNDEFs the Thumb2 insns).

This doesn't change behaviour because all M profile cores
have Thumb2 and so ARM_FEATURE_M implies ARM_FEATURE_THUMB2.
(v6M implements a very restricted subset of Thumb2, but we
can cross that bridge when we get to it with appropriate
feature bits.)

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
---
 target/arm/translate.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/target/arm/translate.c b/target/arm/translate.c
index 58d706c..f5ca87f 100644
--- a/target/arm/translate.c
+++ b/target/arm/translate.c
@@ -9722,8 +9722,7 @@ static int disas_thumb2_insn(CPUARMState *env, DisasContext *s, uint16_t insn_hw
     int conds;
     int logic_cc;
 
-    if (!(arm_dc_feature(s, ARM_FEATURE_THUMB2)
-          || arm_dc_feature(s, ARM_FEATURE_M))) {
+    if (!arm_dc_feature(s, ARM_FEATURE_THUMB2)) {
         /* Thumb-1 cores may need to treat bl and blx as a pair of
            16-bit instructions to get correct prefetch abort behavior.  */
         insn = insn_hw1;
-- 
2.7.4

[Qemu-devel] [PATCH 6/9] target/arm: Pull Thumb insn word loads up to top level

[Peter Maydell]

Refactor the Thumb decode to do the loads of the instruction words at
the top level rather than only loading the second half of a 32-bit
Thumb insn in the middle of the decode.

This is simple apart from the awkward case of Thumb1, where the
BL/BLX prefix and suffix instructions live in what in Thumb2 is the
32-bit insn space.  To handle these we decode enough to identify
whether we're looking at a prefix/suffix that we handle as a 16 bit
insn, or a prefix that we're going to merge with the following suffix
to consider as a 32 bit insn.  The translation of the 16 bit cases
then moves from disas_thumb2_insn() to disas_thumb_insn().

The refactoring has the benefit that we don't need to pass the
CPUARMState* down into the decoder code any more, but the major
reason for doing this is that some Thumb instructions must be always
unconditional regardless of the IT state bits, so we need to know the
whole insn before we emit the "skip this insn if the IT bits and cond
state tell us to" code.  (The always unconditional insns are BKPT,
HLT and SG; the last of these is 32 bits.)

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
---
 target/arm/translate.c | 178 ++++++++++++++++++++++++++++++-------------------
 1 file changed, 108 insertions(+), 70 deletions(-)

diff --git a/target/arm/translate.c b/target/arm/translate.c
index f5ca87f..8d3203e 100644
--- a/target/arm/translate.c
+++ b/target/arm/translate.c
@@ -9623,6 +9623,44 @@ static void disas_arm_insn(DisasContext *s, unsigned int insn)
     }
 }
 
+static bool thumb_insn_is_16bit(DisasContext *s, uint32_t insn)
+{
+    /* Return true if this is a 16 bit instruction. We must be precise
+     * about this (matching the decode).  We assume that s->pc still
+     * points to the first 16 bits of the insn.
+     */
+    if ((insn >> 11) < 0x1d) {
+        /* Definitely a 16-bit instruction */
+        return true;
+    }
+
+    /* Top five bits 0b11101 / 0b11110 / 0b11111 : this is the
+     * first half of a 32-bit Thumb insn. Thumb-1 cores might
+     * end up actually treating this as two 16-bit insns, though,
+     * if it's half of a bl/blx pair that might span a page boundary.
+     */
+    if (arm_dc_feature(s, ARM_FEATURE_THUMB2)) {
+        /* Thumb2 cores (including all M profile ones) always treat
+         * 32-bit insns as 32-bit.
+         */
+        return false;
+    }
+
+    if ((insn >> 11) == 0x1e && (s->pc < s->next_page_start - 3)) {
+        /* 0b1111_0xxx_xxxx_xxxx : BL/BLX prefix, and the suffix
+         * is not on the next page; we merge this into a 32-bit
+         * insn.
+         */
+        return false;
+    }
+    /* 0b1110_1xxx_xxxx_xxxx : BLX suffix (or UNDEF);
+     * 0b1111_1xxx_xxxx_xxxx : BL suffix;
+     * 0b1111_0xxx_xxxx_xxxx : BL/BLX prefix on the end of a page
+     *  -- handle as single 16 bit insn
+     */
+    return true;
+}
+
 /* Return true if this is a Thumb-2 logical op.  */
 static int
 thumb2_logic_op(int op)
@@ -9708,9 +9746,9 @@ gen_thumb2_data_op(DisasContext *s, int op, int conds, uint32_t shifter_out,
 
 /* Translate a 32-bit thumb instruction.  Returns nonzero if the instruction
    is not legal.  */
-static int disas_thumb2_insn(CPUARMState *env, DisasContext *s, uint16_t insn_hw1)
+static int disas_thumb2_insn(DisasContext *s, uint32_t insn)
 {
-    uint32_t insn, imm, shift, offset;
+    uint32_t imm, shift, offset;
     uint32_t rd, rn, rm, rs;
     TCGv_i32 tmp;
     TCGv_i32 tmp2;
@@ -9722,51 +9760,9 @@ static int disas_thumb2_insn(CPUARMState *env, DisasContext *s, uint16_t insn_hw
     int conds;
     int logic_cc;
 
-    if (!arm_dc_feature(s, ARM_FEATURE_THUMB2)) {
-        /* Thumb-1 cores may need to treat bl and blx as a pair of
-           16-bit instructions to get correct prefetch abort behavior.  */
-        insn = insn_hw1;
-        if ((insn & (1 << 12)) == 0) {
-            ARCH(5);
-            /* Second half of blx.  */
-            offset = ((insn & 0x7ff) << 1);
-            tmp = load_reg(s, 14);
-            tcg_gen_addi_i32(tmp, tmp, offset);
-            tcg_gen_andi_i32(tmp, tmp, 0xfffffffc);
-
-            tmp2 = tcg_temp_new_i32();
-            tcg_gen_movi_i32(tmp2, s->pc | 1);
-            store_reg(s, 14, tmp2);
-            gen_bx(s, tmp);
-            return 0;
-        }
-        if (insn & (1 << 11)) {
-            /* Second half of bl.  */
-            offset = ((insn & 0x7ff) << 1) | 1;
-            tmp = load_reg(s, 14);
-            tcg_gen_addi_i32(tmp, tmp, offset);
-
-            tmp2 = tcg_temp_new_i32();
-            tcg_gen_movi_i32(tmp2, s->pc | 1);
-            store_reg(s, 14, tmp2);
-            gen_bx(s, tmp);
-            return 0;
-        }
-        if ((s->pc & ~TARGET_PAGE_MASK) == 0) {
-            /* Instruction spans a page boundary.  Implement it as two
-               16-bit instructions in case the second half causes an
-               prefetch abort.  */
-            offset = ((int32_t)insn << 21) >> 9;
-            tcg_gen_movi_i32(cpu_R[14], s->pc + 2 + offset);
-            return 0;
-        }
-        /* Fall through to 32-bit decode.  */
-    }
-
-    insn = arm_lduw_code(env, s->pc, s->sctlr_b);
-    s->pc += 2;
-    insn |= (uint32_t)insn_hw1 << 16;
-
+    /* The only 32 bit insn that's allowed for Thumb1 is the combined
+     * BL/BLX prefix and suffix.
+     */
     if ((insn & 0xf800e800) != 0xf000e800) {
         ARCH(6T2);
     }
@@ -11081,27 +11077,15 @@ illegal_op:
     return 1;
 }
 
-static void disas_thumb_insn(CPUARMState *env, DisasContext *s)
+static void disas_thumb_insn(DisasContext *s, uint32_t insn)
 {
-    uint32_t val, insn, op, rm, rn, rd, shift, cond;
+    uint32_t val, op, rm, rn, rd, shift, cond;
     int32_t offset;
     int i;
     TCGv_i32 tmp;
     TCGv_i32 tmp2;
     TCGv_i32 addr;
 
-    if (s->condexec_mask) {
-        cond = s->condexec_cond;
-        if (cond != 0x0e) {     /* Skip conditional when condition is AL. */
-          s->condlabel = gen_new_label();
-          arm_gen_test_cc(cond ^ 1, s->condlabel);
-          s->condjmp = 1;
-        }
-    }
-
-    insn = arm_lduw_code(env, s->pc, s->sctlr_b);
-    s->pc += 2;
-
     switch (insn >> 12) {
     case 0: case 1:
 
@@ -11832,8 +11816,21 @@ static void disas_thumb_insn(CPUARMState *env, DisasContext *s)
 
     case 14:
         if (insn & (1 << 11)) {
-            if (disas_thumb2_insn(env, s, insn))
-              goto undef32;
+            /* thumb_insn_is_16bit() ensures we can't get here for
+             * a Thumb2 CPU, so this must be a thumb1 split BL/BLX:
+             * 0b1110_1xxx_xxxx_xxxx : BLX suffix (or UNDEF)
+             */
+            assert(!arm_dc_feature(s, ARM_FEATURE_THUMB2));
+            ARCH(5);
+            offset = ((insn & 0x7ff) << 1);
+            tmp = load_reg(s, 14);
+            tcg_gen_addi_i32(tmp, tmp, offset);
+            tcg_gen_andi_i32(tmp, tmp, 0xfffffffc);
+
+            tmp2 = tcg_temp_new_i32();
+            tcg_gen_movi_i32(tmp2, s->pc | 1);
+            store_reg(s, 14, tmp2);
+            gen_bx(s, tmp);
             break;
         }
         /* unconditional branch */
@@ -11844,15 +11841,30 @@ static void disas_thumb_insn(CPUARMState *env, DisasContext *s)
         break;
 
     case 15:
-        if (disas_thumb2_insn(env, s, insn))
-            goto undef32;
+        /* thumb_insn_is_16bit() ensures we can't get here for
+         * a Thumb2 CPU, so this must be a thumb1 split BL/BLX.
+         */
+        assert(!arm_dc_feature(s, ARM_FEATURE_THUMB2));
+
+        if (insn & (1 << 11)) {
+            /* 0b1111_1xxx_xxxx_xxxx : BL suffix */
+            offset = ((insn & 0x7ff) << 1) | 1;
+            tmp = load_reg(s, 14);
+            tcg_gen_addi_i32(tmp, tmp, offset);
+
+            tmp2 = tcg_temp_new_i32();
+            tcg_gen_movi_i32(tmp2, s->pc | 1);
+            store_reg(s, 14, tmp2);
+            gen_bx(s, tmp);
+        } else {
+            /* 0b1111_0xxx_xxxx_xxxx : BL/BLX prefix */
+            uint32_t uoffset = ((int32_t)insn << 21) >> 9;
+
+            tcg_gen_movi_i32(cpu_R[14], s->pc + 2 + uoffset);
+        }
         break;
     }
     return;
-undef32:
-    gen_exception_insn(s, 4, EXCP_UDEF, syn_uncategorized(),
-                       default_exception_el(s));
-    return;
 illegal_op:
 undef:
     gen_exception_insn(s, 2, EXCP_UDEF, syn_uncategorized(),
@@ -12122,12 +12134,38 @@ static void thumb_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu)
 {
     DisasContext *dc = container_of(dcbase, DisasContext, base);
     CPUARMState *env = cpu->env_ptr;
+    uint32_t insn;
+    bool is_16bit;
 
     if (arm_pre_translate_insn(dc)) {
         return;
     }
 
-    disas_thumb_insn(env, dc);
+    insn = arm_lduw_code(env, dc->pc, dc->sctlr_b);
+    is_16bit = thumb_insn_is_16bit(dc, insn);
+    dc->pc += 2;
+    if (!is_16bit) {
+        uint32_t insn2 = arm_lduw_code(env, dc->pc, dc->sctlr_b);
+
+        insn = insn << 16 | insn2;
+        dc->pc += 2;
+    }
+
+    if (dc->condexec_mask) {
+        uint32_t cond = dc->condexec_cond;
+
+        if (cond != 0x0e) {     /* Skip conditional when condition is AL. */
+            dc->condlabel = gen_new_label();
+            arm_gen_test_cc(cond ^ 1, dc->condlabel);
+            dc->condjmp = 1;
+        }
+    }
+
+    if (is_16bit) {
+        disas_thumb_insn(dc, insn);
+    } else {
+        disas_thumb2_insn(dc, insn);
+    }
 
     /* Advance the Thumb condexec condition.  */
     if (dc->condexec_mask) {
-- 
2.7.4

[Qemu-devel] [PATCH 7/9] target-arm: Simplify insn_crosses_page()

[Peter Maydell]

Recent changes have left insn_crosses_page() more complicated
than it needed to be:
 * it's only called from thumb_tr_translate_insn() so we know
   for certain that we're looking at a Thumb insn
 * the caller's check for dc->pc >= dc->next_page_start - 3
   means that dc->pc can't possibly be 4 aligned, so there's
   no need to check that (the check was partly there to ensure
   that we didn't treat an ARM insn as Thumb, I think)
 * we now have thumb_insn_is_16bit() which lets us do a precise
   check of the length of the next insn, rather than opencoding
   an inaccurate check

Simplify it down to just loading the first half of the insn
and calling thumb_insn_is_16bit() on it.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
---
 target/arm/translate.c | 27 ++++++---------------------
 1 file changed, 6 insertions(+), 21 deletions(-)

diff --git a/target/arm/translate.c b/target/arm/translate.c
index 8d3203e..5838e67 100644
--- a/target/arm/translate.c
+++ b/target/arm/translate.c
@@ -11875,29 +11875,14 @@ static bool insn_crosses_page(CPUARMState *env, DisasContext *s)
 {
     /* Return true if the insn at dc->pc might cross a page boundary.
      * (False positives are OK, false negatives are not.)
+     * We know this is a Thumb insn, and our caller ensures we are
+     * only called if dc->pc is less than 4 bytes from the page
+     * boundary, so we cross the page if the first 16 bits indicate
+     * that this is a 32 bit insn.
      */
-    uint16_t insn;
+    uint16_t insn = arm_lduw_code(env, s->pc, s->sctlr_b);
 
-    if ((s->pc & 3) == 0) {
-        /* At a 4-aligned address we can't be crossing a page */
-        return false;
-    }
-
-    /* This must be a Thumb insn */
-    insn = arm_lduw_code(env, s->pc, s->sctlr_b);
-
-    if ((insn >> 11) >= 0x1d) {
-        /* Top five bits 0b11101 / 0b11110 / 0b11111 : this is the
-         * First half of a 32-bit Thumb insn. Thumb-1 cores might
-         * end up actually treating this as two 16-bit insns (see the
-         * code at the start of disas_thumb2_insn()) but we don't bother
-         * to check for that as it is unlikely, and false positives here
-         * are harmless.
-         */
-        return true;
-    }
-    /* Definitely a 16-bit insn, can't be crossing a page. */
-    return false;
+    return !thumb_insn_is_16bit(s, insn);
 }
 
 static int arm_tr_init_disas_context(DisasContextBase *dcbase,
-- 
2.7.4

[Qemu-devel] [PATCH 8/9] target/arm: Support some Thumb insns being always unconditional

[Peter Maydell]

A few Thumb instructions are always unconditional even inside an
IT block (as opposed to being UNPREDICTABLE if used inside an
IT block): BKPT, the v8M SG instruction, and the A profile
HLT (debug halt) instruction.

This means we need to suppress the jump-over-instruction-on-condfail
code generation (though the IT state still advances as usual and
subsequent insns in the IT block may be conditional).

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
---
 target/arm/translate.c | 48 +++++++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 47 insertions(+), 1 deletion(-)

diff --git a/target/arm/translate.c b/target/arm/translate.c
index 5838e67..9d16760 100644
--- a/target/arm/translate.c
+++ b/target/arm/translate.c
@@ -12115,6 +12115,52 @@ static void arm_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu)
        in init_disas_context by adjusting max_insns.  */
 }
 
+static bool thumb_insn_is_unconditional(DisasContext *s, uint32_t insn)
+{
+    /* Return true if this Thumb insn is always unconditional,
+     * even inside an IT block. This is true of only a very few
+     * instructions: BKPT, HLT, and SG.
+     *
+     * A larger class of instructions are UNPREDICTABLE if used
+     * inside an IT block; we do not need to detect those here, because
+     * what we do by default (perform the cc check and update the IT
+     * bits state machine) is a permitted CONSTRAINED UNPREDICTABLE
+     * choice for those situations.
+     *
+     * insn is either a 16-bit or a 32-bit instruction; the two are
+     * distinguishable because for the 16-bit case the top 16 bits
+     * are zeroes, and that isn't a valid 32-bit encoding.
+     */
+    if ((insn & 0xffffff00) == 0xbe00) {
+        /* BKPT */
+        return true;
+    }
+
+    if ((insn & 0xffffffc0) == 0xba80 && arm_dc_feature(s, ARM_FEATURE_V8) &&
+        !arm_dc_feature(s, ARM_FEATURE_M)) {
+        /* HLT: v8A only. This is unconditional even when it is going to
+         * UNDEF; see the v8A ARM ARM DDI0487B.a H3.3.
+         * For v7 cores this was a plain old undefined encoding and so
+         * honours its cc check. (We might be using the encoding as
+         * a semihosting trap, but we don't change the cc check behaviour
+         * on that account, because a debugger connected to a real v7A
+         * core and emulating semihosting traps by catching the UNDEF
+         * exception would also only see cases where the cc check passed.
+         * No guest code should be trying to do a HLT semihosting trap
+         * in an IT block anyway.
+         */
+        return true;
+    }
+
+    if (insn == 0xe97fe97f && arm_dc_feature(s, ARM_FEATURE_V8) &&
+        arm_dc_feature(s, ARM_FEATURE_M)) {
+        /* SG: v8M only */
+        return true;
+    }
+
+    return false;
+}
+
 static void thumb_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu)
 {
     DisasContext *dc = container_of(dcbase, DisasContext, base);
@@ -12136,7 +12182,7 @@ static void thumb_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu)
         dc->pc += 2;
     }
 
-    if (dc->condexec_mask) {
+    if (dc->condexec_mask && !thumb_insn_is_unconditional(dc, insn)) {
         uint32_t cond = dc->condexec_cond;
 
         if (cond != 0x0e) {     /* Skip conditional when condition is AL. */
-- 
2.7.4

[Qemu-devel] [PATCH 9/9] target/arm: Implement SG instruction corner cases

[Peter Maydell]

The common situation of the SG instruction is that it is
executed from S&NSC memory by a CPU in NS state. That case
is handled by v7m_handle_execute_nsc(). However the instruction
also has defined behaviour in a couple of other cases:
 * SG instruction in NS memory (behaves as a NOP)
 * SG in S memory but CPU already secure (clears IT bits and
   does nothing else)
 * SG instruction in v8M without Security Extension (NOP)

These can be implemented in translate.c.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
---
 target/arm/translate.c | 23 ++++++++++++++++++++++-
 1 file changed, 22 insertions(+), 1 deletion(-)

diff --git a/target/arm/translate.c b/target/arm/translate.c
index 9d16760..3db6d73 100644
--- a/target/arm/translate.c
+++ b/target/arm/translate.c
@@ -9781,7 +9781,28 @@ static int disas_thumb2_insn(DisasContext *s, uint32_t insn)
              * - load/store doubleword, load/store exclusive, ldacq/strel,
              *   table branch.
              */
-            if (insn & 0x01200000) {
+            if (insn == 0xe97fe97f && arm_dc_feature(s, ARM_FEATURE_M) &&
+                arm_dc_feature(s, ARM_FEATURE_V8)) {
+                /* 0b1110_1001_0111_1111_1110_1001_0111_111
+                 *  - SG (v8M only)
+                 * The bulk of the behaviour for this instruction is implemented
+                 * in v7m_handle_execute_nsc(), which deals with the insn when
+                 * it is executed by a CPU in non-secure state from memory
+                 * which is Secure & NonSecure-Callable.
+                 * Here we only need to handle the remaining cases:
+                 *  * in NS memory (including the "security extension not
+                 *    implemented" case) : NOP
+                 *  * in S memory but CPU already secure (clear IT bits)
+                 * We know that the attribute for the memory this insn is
+                 * in must match the current CPU state, because otherwise
+                 * get_phys_addr_pmsav8 would have generated an exception.
+                 */
+                if (s->v8m_secure) {
+                    /* Like the IT insn, we don't need to generate any code */
+                    s->condexec_cond = 0;
+                    s->condexec_mask = 0;
+                }
+            } else if (insn & 0x01200000) {
                 /* 0b1110_1000_x11x_xxxx_xxxx_xxxx_xxxx_xxxx
                  *  - load/store dual (post-indexed)
                  * 0b1111_1001_x10x_xxxx_xxxx_xxxx_xxxx_xxxx
-- 
2.7.4

Re: [Qemu-devel] [PATCH 1/9] target/arm: Add M profile secure MMU index values to get_a32_user_mem_index()

[Richard Henderson]

On 10/09/2017 06:48 AM, Peter Maydell wrote:
> Add the M profile secure MMU index values to the switch in
> get_a32_user_mem_index() so that LDRT/STRT work correctly
> rather than asserting at translate time.
> 
> Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
> ---
>  target/arm/translate.c | 4 ++++
>  1 file changed, 4 insertions(+)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>


r~

Re: [Qemu-devel] [PATCH 2/9] target/arm: Implement SG instruction

[Richard Henderson]

On 10/09/2017 06:48 AM, Peter Maydell wrote:
> Implement the SG instruction, which we emulate 'by hand' in the
> exception handling code path.
> 
> Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
> ---
>  target/arm/helper.c | 132 ++++++++++++++++++++++++++++++++++++++++++++++++++--
>  1 file changed, 127 insertions(+), 5 deletions(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>


r~

Re: [Qemu-devel] [PATCH 5/9] target-arm: Don't check for "Thumb2 or M profile" for not-Thumb1

[Richard Henderson]

On 10/09/2017 06:48 AM, Peter Maydell wrote:
> The code which implements the Thumb1 split BL/BLX instructions
> is guarded by a check on "not M or THUMB2". All we really need
> to check here is "not THUMB2" (and we assume that elsewhere too,
> eg in the ARCH(6T2) test that UNDEFs the Thumb2 insns).
> 
> This doesn't change behaviour because all M profile cores
> have Thumb2 and so ARM_FEATURE_M implies ARM_FEATURE_THUMB2.
> (v6M implements a very restricted subset of Thumb2, but we
> can cross that bridge when we get to it with appropriate
> feature bits.)
> 
> Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
> ---
>  target/arm/translate.c | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>


r~

Re: [Qemu-devel] [PATCH 6/9] target/arm: Pull Thumb insn word loads up to top level

[Richard Henderson]

On 10/09/2017 06:48 AM, Peter Maydell wrote:
> +    if ((insn >> 11) == 0x1e && (s->pc < s->next_page_start - 3)) {
> +        /* 0b1111_0xxx_xxxx_xxxx : BL/BLX prefix, and the suffix
> +         * is not on the next page; we merge this into a 32-bit
> +         * insn.
> +         */
> +        return false;
> +    }
> +    /* 0b1110_1xxx_xxxx_xxxx : BLX suffix (or UNDEF);
> +     * 0b1111_1xxx_xxxx_xxxx : BL suffix;
> +     * 0b1111_0xxx_xxxx_xxxx : BL/BLX prefix on the end of a page
> +     *  -- handle as single 16 bit insn
> +     */

I really had to search to find docs for this.  I suspect it's quite clear in v5
manuals, but I didn't have one handy and arm.com seems to have dropped
everything before v6M.  However, I did see a footnote in v7M that explained.

But, everything appears to be in order.  It's a nice cleanup.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>


r~

Re: [Qemu-devel] [PATCH 7/9] target-arm: Simplify insn_crosses_page()

[Richard Henderson]

On 10/09/2017 06:48 AM, Peter Maydell wrote:
> Recent changes have left insn_crosses_page() more complicated
> than it needed to be:
>  * it's only called from thumb_tr_translate_insn() so we know
>    for certain that we're looking at a Thumb insn
>  * the caller's check for dc->pc >= dc->next_page_start - 3
>    means that dc->pc can't possibly be 4 aligned, so there's
>    no need to check that (the check was partly there to ensure
>    that we didn't treat an ARM insn as Thumb, I think)
>  * we now have thumb_insn_is_16bit() which lets us do a precise
>    check of the length of the next insn, rather than opencoding
>    an inaccurate check
> 
> Simplify it down to just loading the first half of the insn
> and calling thumb_insn_is_16bit() on it.
> 
> Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
> ---
>  target/arm/translate.c | 27 ++++++---------------------
>  1 file changed, 6 insertions(+), 21 deletions(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>


r~

Re: [Qemu-devel] [PATCH 8/9] target/arm: Support some Thumb insns being always unconditional

[Richard Henderson]

On 10/09/2017 06:48 AM, Peter Maydell wrote:
> -    if (dc->condexec_mask) {
> +    if (dc->condexec_mask && !thumb_insn_is_unconditional(dc, insn)) {
>          uint32_t cond = dc->condexec_cond;
>  
>          if (cond != 0x0e) {     /* Skip conditional when condition is AL. */

Don't you still need to advance the condexec_mask?

For HLT it doesn't matter, clearly.
I'm not sure what happens to an IT block for debug breakpoints.

But SG behaves as NOP when in non-secure memory.
That would seem to require that

	itft	eq
	add	r1, r0, r0
	sg
	add	r2, r0, r0
	add	r3, r0, r0

should modify both r1 and r2 for T and r3 should be unconditional.  If you
don't advance condexec_mask, it would seem that r2 gets F and r3 gets T.


r~

Re: [Qemu-devel] [PATCH 9/9] target/arm: Implement SG instruction corner cases

[Richard Henderson]

On 10/09/2017 06:48 AM, Peter Maydell wrote:
> The common situation of the SG instruction is that it is
> executed from S&NSC memory by a CPU in NS state. That case
> is handled by v7m_handle_execute_nsc(). However the instruction
> also has defined behaviour in a couple of other cases:
>  * SG instruction in NS memory (behaves as a NOP)
>  * SG in S memory but CPU already secure (clears IT bits and
>    does nothing else)
>  * SG instruction in v8M without Security Extension (NOP)
> 
> These can be implemented in translate.c.
> 
> Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
> ---
>  target/arm/translate.c | 23 ++++++++++++++++++++++-
>  1 file changed, 22 insertions(+), 1 deletion(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>


r~

Re: [Qemu-devel] [PATCH 6/9] target/arm: Pull Thumb insn word loads up to top level

[Peter Maydell]

On 11 October 2017 at 03:18, Richard Henderson
<richard.henderson@linaro.org> wrote:
> On 10/09/2017 06:48 AM, Peter Maydell wrote:
>> +    if ((insn >> 11) == 0x1e && (s->pc < s->next_page_start - 3)) {
>> +        /* 0b1111_0xxx_xxxx_xxxx : BL/BLX prefix, and the suffix
>> +         * is not on the next page; we merge this into a 32-bit
>> +         * insn.
>> +         */
>> +        return false;
>> +    }
>> +    /* 0b1110_1xxx_xxxx_xxxx : BLX suffix (or UNDEF);
>> +     * 0b1111_1xxx_xxxx_xxxx : BL suffix;
>> +     * 0b1111_0xxx_xxxx_xxxx : BL/BLX prefix on the end of a page
>> +     *  -- handle as single 16 bit insn
>> +     */
>
> I really had to search to find docs for this.  I suspect it's quite clear in v5
> manuals, but I didn't have one handy and arm.com seems to have dropped
> everything before v6M.  However, I did see a footnote in v7M that explained.

Yes, it's in the ARMv5 and v6 ARM, which you can get from here:
https://developer.arm.com/docs/ddi0100/latest/armv5-architecture-reference-manual

thanks
-- PMM

Re: [Qemu-devel] [PATCH 8/9] target/arm: Support some Thumb insns being always unconditional

[Peter Maydell]

On 11 October 2017 at 03:52, Richard Henderson
<richard.henderson@linaro.org> wrote:
> On 10/09/2017 06:48 AM, Peter Maydell wrote:
>> -    if (dc->condexec_mask) {
>> +    if (dc->condexec_mask && !thumb_insn_is_unconditional(dc, insn)) {
>>          uint32_t cond = dc->condexec_cond;
>>
>>          if (cond != 0x0e) {     /* Skip conditional when condition is AL. */
>
> Don't you still need to advance the condexec_mask?

Yes -- that happens after we've called disas_thumb{,2}_insn()
in thumb_tr_translate_insn().

This patch just makes always-unconditional insns behave
as if the condition is AL.

thanks
-- PMM

Re: [Qemu-devel] [PATCH 8/9] target/arm: Support some Thumb insns being always unconditional

[Richard Henderson]

On 10/11/2017 02:57 AM, Peter Maydell wrote:
> On 11 October 2017 at 03:52, Richard Henderson
> <richard.henderson@linaro.org> wrote:
>> On 10/09/2017 06:48 AM, Peter Maydell wrote:
>>> -    if (dc->condexec_mask) {
>>> +    if (dc->condexec_mask && !thumb_insn_is_unconditional(dc, insn)) {
>>>          uint32_t cond = dc->condexec_cond;
>>>
>>>          if (cond != 0x0e) {     /* Skip conditional when condition is AL. */
>>
>> Don't you still need to advance the condexec_mask?
> 
> Yes -- that happens after we've called disas_thumb{,2}_insn()
> in thumb_tr_translate_insn().

Ah, mis-read the context of this hunk.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>


r~