Merge remote-tracking branch 'remotes/xtensa/tags/20170124-xtensa' into staging

target/xtensa updates:

- refactor CCOUNT/CCOMPARE (use QEMU timers instead of instruction counting);
- support icount; run target/xtensa TCG tests with icount;
- implement SMP prerequisites: static vector selection, RUNSTALL and RER/WER.

# gpg: Signature made Wed 25 Jan 2017 00:27:51 GMT
# gpg:                using RSA key 0x51F9CC91F83FA044
# gpg: Good signature from "Max Filippov <max.filippov@cogentembedded.com>"
# gpg:                 aka "Max Filippov <jcmvbkbc@gmail.com>"
# Primary key fingerprint: 2B67 854B 98E5 327D CDEB  17D8 51F9 CC91 F83F A044

* remotes/xtensa/tags/20170124-xtensa:
  target-xtensa: implement RER/WER instructions
  target/xtensa: tests: clean up interrupt tests
  target/xtensa: tests: add memctl test
  target/xtensa: implement MEMCTL SR
  target/xtensa: fix ICACHE/DCACHE options detection
  target/xtensa: tests: add ccount write tests
  target/xtensa: tests: replace hardcoded interrupt masks
  target/xtensa: tests: fix timer tests
  target/xtensa: tests: run tests with icount
  target/xtensa: don't continue translation after exception
  target/xtensa: support icount
  target/xtensa: refactor CCOUNT/CCOMPARE
  target/xtensa: implement RUNSTALL
  target/xtensa: add static vectors selection

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

12 files changed, 456 insertions, 203 deletions

diff --git a/hw/xtensa/pic_cpu.c b/hw/xtensa/pic_cpu.c
index 2bed64f15b..0e812d7f06 100644
--- a/hw/xtensa/pic_cpu.c
+++ b/hw/xtensa/pic_cpu.c
@@ -31,22 +31,6 @@
 #include "qemu/log.h"
 #include "qemu/timer.h"
 
-void xtensa_advance_ccount(CPUXtensaState *env, uint32_t d)
-{
-    uint32_t old_ccount = env->sregs[CCOUNT] + 1;
-
-    env->sregs[CCOUNT] += d;
-
-    if (xtensa_option_enabled(env->config, XTENSA_OPTION_TIMER_INTERRUPT)) {
-        int i;
-        for (i = 0; i < env->config->nccompare; ++i) {
-            if (env->sregs[CCOMPARE + i] - old_ccount < d) {
-                xtensa_timer_irq(env, i, 1);
-            }
-        }
-    }
-}
-
 void check_interrupts(CPUXtensaState *env)
 {
     CPUState *cs = CPU(xtensa_env_get_cpu(env));
@@ -54,17 +38,6 @@ void check_interrupts(CPUXtensaState *env)
     uint32_t int_set_enabled = env->sregs[INTSET] & env->sregs[INTENABLE];
     int level;
 
-    /* If the CPU is halted advance CCOUNT according to the QEMU_CLOCK_VIRTUAL time
-     * elapsed since the moment when it was advanced last time.
-     */
-    if (cs->halted) {
-        int64_t now = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
-
-        xtensa_advance_ccount(env,
-                muldiv64(now - env->halt_clock,
-                    env->config->clock_freq_khz, 1000000));
-        env->halt_clock = now;
-    }
     for (level = env->config->nlevel; level > minlevel; --level) {
         if (env->config->level_mask[level] & int_set_enabled) {
             env->pending_irq_level = level;
@@ -109,49 +82,29 @@ void xtensa_timer_irq(CPUXtensaState *env, uint32_t id, uint32_t active)
     qemu_set_irq(env->irq_inputs[env->config->timerint[id]], active);
 }
 
-void xtensa_rearm_ccompare_timer(CPUXtensaState *env)
-{
-    int i;
-    uint32_t wake_ccount = env->sregs[CCOUNT] - 1;
-
-    for (i = 0; i < env->config->nccompare; ++i) {
-        if (env->sregs[CCOMPARE + i] - env->sregs[CCOUNT] <
-                wake_ccount - env->sregs[CCOUNT]) {
-            wake_ccount = env->sregs[CCOMPARE + i];
-        }
-    }
-    env->wake_ccount = wake_ccount;
-    timer_mod(env->ccompare_timer, env->halt_clock +
-            (uint64_t)(wake_ccount - env->sregs[CCOUNT]) *
-            1000000 / env->config->clock_freq_khz);
-}
-
 static void xtensa_ccompare_cb(void *opaque)
 {
-    XtensaCPU *cpu = opaque;
-    CPUXtensaState *env = &cpu->env;
-    CPUState *cs = CPU(cpu);
+    XtensaCcompareTimer *ccompare = opaque;
+    CPUXtensaState *env = ccompare->env;
+    unsigned i = ccompare - env->ccompare;
 
-    if (cs->halted) {
-        env->halt_clock = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
-        xtensa_advance_ccount(env, env->wake_ccount - env->sregs[CCOUNT]);
-        if (!cpu_has_work(cs)) {
-            env->sregs[CCOUNT] = env->wake_ccount + 1;
-            xtensa_rearm_ccompare_timer(env);
-        }
-    }
+    xtensa_timer_irq(env, i, 1);
 }
 
 void xtensa_irq_init(CPUXtensaState *env)
 {
-    XtensaCPU *cpu = xtensa_env_get_cpu(env);
-
     env->irq_inputs = (void **)qemu_allocate_irqs(
             xtensa_set_irq, env, env->config->ninterrupt);
-    if (xtensa_option_enabled(env->config, XTENSA_OPTION_TIMER_INTERRUPT) &&
-            env->config->nccompare > 0) {
-        env->ccompare_timer =
-            timer_new_ns(QEMU_CLOCK_VIRTUAL, &xtensa_ccompare_cb, cpu);
+    if (xtensa_option_enabled(env->config, XTENSA_OPTION_TIMER_INTERRUPT)) {
+        unsigned i;
+
+        env->time_base = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
+        env->ccount_base = env->sregs[CCOUNT];
+        for (i = 0; i < env->config->nccompare; ++i) {
+            env->ccompare[i].env = env;
+            env->ccompare[i].timer = timer_new_ns(QEMU_CLOCK_VIRTUAL,
+                    xtensa_ccompare_cb, env->ccompare + i);
+        }
     }
 }
 
diff --git a/target/xtensa/cpu.c b/target/xtensa/cpu.c
index e8e9f9175b..cd7f95823f 100644
--- a/target/xtensa/cpu.c
+++ b/target/xtensa/cpu.c
@@ -47,7 +47,7 @@ static bool xtensa_cpu_has_work(CPUState *cs)
 {
     XtensaCPU *cpu = XTENSA_CPU(cs);
 
-    return cpu->env.pending_irq_level;
+    return !cpu->env.runstall && cpu->env.pending_irq_level;
 }
 
 /* CPUClass::reset() */
@@ -60,12 +60,13 @@ static void xtensa_cpu_reset(CPUState *s)
     xcc->parent_reset(s);
 
     env->exception_taken = 0;
-    env->pc = env->config->exception_vector[EXC_RESET];
+    env->pc = env->config->exception_vector[EXC_RESET0 + env->static_vectors];
     env->sregs[LITBASE] &= ~1;
     env->sregs[PS] = xtensa_option_enabled(env->config,
             XTENSA_OPTION_INTERRUPT) ? 0x1f : 0x10;
     env->sregs[VECBASE] = env->config->vecbase;
     env->sregs[IBREAKENABLE] = 0;
+    env->sregs[MEMCTL] = MEMCTL_IL0EN & env->config->memctl_mask;
     env->sregs[CACHEATTR] = 0x22222222;
     env->sregs[ATOMCTL] = xtensa_option_enabled(env->config,
             XTENSA_OPTION_ATOMCTL) ? 0x28 : 0x15;
@@ -74,6 +75,7 @@ static void xtensa_cpu_reset(CPUState *s)
 
     env->pending_irq_level = 0;
     reset_mmu(env);
+    s->halted = env->runstall;
 }
 
 static ObjectClass *xtensa_cpu_class_by_name(const char *cpu_model)
@@ -125,6 +127,12 @@ static void xtensa_cpu_initfn(Object *obj)
     cs->env_ptr = env;
     env->config = xcc->config;
 
+    env->address_space_er = g_malloc(sizeof(*env->address_space_er));
+    env->system_er = g_malloc(sizeof(*env->system_er));
+    memory_region_init_io(env->system_er, NULL, NULL, env, "er",
+                          UINT64_C(0x100000000));
+    address_space_init(env->address_space_er, env->system_er, "ER");
+
     if (tcg_enabled() && !tcg_inited) {
         tcg_inited = true;
         xtensa_translate_init();
diff --git a/target/xtensa/cpu.h b/target/xtensa/cpu.h
index 7fe82a37af..7e7131a596 100644
--- a/target/xtensa/cpu.h
+++ b/target/xtensa/cpu.h
@@ -103,6 +103,7 @@ enum {
     XTENSA_OPTION_PROCESSOR_ID,
     XTENSA_OPTION_DEBUG,
     XTENSA_OPTION_TRACE_PORT,
+    XTENSA_OPTION_EXTERN_REGS,
 };
 
 enum {
@@ -129,6 +130,7 @@ enum {
     ITLBCFG = 91,
     DTLBCFG = 92,
     IBREAKENABLE = 96,
+    MEMCTL = 97,
     CACHEATTR = 98,
     ATOMCTL = 99,
     IBREAKA = 128,
@@ -189,6 +191,20 @@ enum {
 #define DBREAKC_SB_LB (DBREAKC_SB | DBREAKC_LB)
 #define DBREAKC_MASK 0x3f
 
+#define MEMCTL_INIT 0x00800000
+#define MEMCTL_IUSEWAYS_SHIFT 18
+#define MEMCTL_IUSEWAYS_LEN 5
+#define MEMCTL_IUSEWAYS_MASK 0x007c0000
+#define MEMCTL_DALLOCWAYS_SHIFT 13
+#define MEMCTL_DALLOCWAYS_LEN 5
+#define MEMCTL_DALLOCWAYS_MASK 0x0003e000
+#define MEMCTL_DUSEWAYS_SHIFT 8
+#define MEMCTL_DUSEWAYS_LEN 5
+#define MEMCTL_DUSEWAYS_MASK 0x00001f00
+#define MEMCTL_ISNP 0x4
+#define MEMCTL_DSNP 0x2
+#define MEMCTL_IL0EN 0x1
+
 #define MAX_NAREG 64
 #define MAX_NINTERRUPT 32
 #define MAX_NLEVEL 6
@@ -209,7 +225,8 @@ enum {
 
 enum {
     /* Static vectors */
-    EXC_RESET,
+    EXC_RESET0,
+    EXC_RESET1,
     EXC_MEMORY_ERROR,
 
     /* Dynamic vectors */
@@ -268,6 +285,8 @@ typedef enum {
     INTTYPE_MAX
 } interrupt_type;
 
+struct CPUXtensaState;
+
 typedef struct xtensa_tlb_entry {
     uint32_t vaddr;
     uint32_t paddr;
@@ -297,6 +316,11 @@ typedef struct XtensaGdbRegmap {
     XtensaGdbReg reg[1 + 16 + 64 + 256 + 256];
 } XtensaGdbRegmap;
 
+typedef struct XtensaCcompareTimer {
+    struct CPUXtensaState *env;
+    QEMUTimer *timer;
+} XtensaCcompareTimer;
+
 struct XtensaConfig {
     const char *name;
     uint64_t options;
@@ -324,6 +348,10 @@ struct XtensaConfig {
     unsigned nibreak;
     unsigned ndbreak;
 
+    unsigned icache_ways;
+    unsigned dcache_ways;
+    uint32_t memctl_mask;
+
     uint32_t configid[2];
 
     uint32_t clock_freq_khz;
@@ -365,14 +393,19 @@ typedef struct CPUXtensaState {
     xtensa_tlb_entry itlb[7][MAX_TLB_WAY_SIZE];
     xtensa_tlb_entry dtlb[10][MAX_TLB_WAY_SIZE];
     unsigned autorefill_idx;
-
+    bool runstall;
+    AddressSpace *address_space_er;
+    MemoryRegion *system_er;
     int pending_irq_level; /* level of last raised IRQ */
     void **irq_inputs;
-    QEMUTimer *ccompare_timer;
-    uint32_t wake_ccount;
-    int64_t halt_clock;
+    XtensaCcompareTimer ccompare[MAX_NCCOMPARE];
+    uint64_t time_base;
+    uint64_t ccount_time;
+    uint32_t ccount_base;
 
     int exception_taken;
+    int yield_needed;
+    unsigned static_vectors;
 
     /* Watchpoints for DBREAK registers */
     struct CPUWatchpoint *cpu_watchpoint[MAX_NDBREAK];
@@ -437,9 +470,7 @@ void xtensa_register_core(XtensaConfigList *node);
 void check_interrupts(CPUXtensaState *s);
 void xtensa_irq_init(CPUXtensaState *env);
 void *xtensa_get_extint(CPUXtensaState *env, unsigned extint);
-void xtensa_advance_ccount(CPUXtensaState *env, uint32_t d);
 void xtensa_timer_irq(CPUXtensaState *env, uint32_t id, uint32_t active);
-void xtensa_rearm_ccompare_timer(CPUXtensaState *env);
 int cpu_xtensa_signal_handler(int host_signum, void *pinfo, void *puc);
 void xtensa_cpu_list(FILE *f, fprintf_function cpu_fprintf);
 void xtensa_sync_window_from_phys(CPUXtensaState *env);
@@ -460,7 +491,18 @@ int xtensa_get_physical_addr(CPUXtensaState *env, bool update_tlb,
 void reset_mmu(CPUXtensaState *env);
 void dump_mmu(FILE *f, fprintf_function cpu_fprintf, CPUXtensaState *env);
 void debug_exception_env(CPUXtensaState *new_env, uint32_t cause);
+static inline MemoryRegion *xtensa_get_er_region(CPUXtensaState *env)
+{
+    return env->system_er;
+}
 
+static inline void xtensa_select_static_vectors(CPUXtensaState *env,
+                                                unsigned n)
+{
+    assert(n < 2);
+    env->static_vectors = n;
+}
+void xtensa_runstall(CPUXtensaState *env, bool runstall);
 
 #define XTENSA_OPTION_BIT(opt) (((uint64_t)1) << (opt))
 #define XTENSA_OPTION_ALL (~(uint64_t)0)
@@ -539,6 +581,7 @@ static inline int cpu_mmu_index(CPUXtensaState *env, bool ifetch)
 #define XTENSA_TBFLAG_EXCEPTION 0x4000
 #define XTENSA_TBFLAG_WINDOW_MASK 0x18000
 #define XTENSA_TBFLAG_WINDOW_SHIFT 15
+#define XTENSA_TBFLAG_YIELD 0x20000
 
 static inline void cpu_get_tb_cpu_state(CPUXtensaState *env, target_ulong *pc,
         target_ulong *cs_base, uint32_t *flags)
@@ -580,6 +623,9 @@ static inline void cpu_get_tb_cpu_state(CPUXtensaState *env, target_ulong *pc,
     } else {
         *flags |= 3 << XTENSA_TBFLAG_WINDOW_SHIFT;
     }
+    if (env->yield_needed) {
+        *flags |= XTENSA_TBFLAG_YIELD;
+    }
 }
 
 #include "exec/cpu-all.h"
diff --git a/target/xtensa/helper.c b/target/xtensa/helper.c
index 768b32c417..c67d715c4b 100644
--- a/target/xtensa/helper.c
+++ b/target/xtensa/helper.c
@@ -728,3 +728,16 @@ void dump_mmu(FILE *f, fprintf_function cpu_fprintf, CPUXtensaState *env)
         cpu_fprintf(f, "No TLB for this CPU core\n");
     }
 }
+
+void xtensa_runstall(CPUXtensaState *env, bool runstall)
+{
+    CPUState *cpu = CPU(xtensa_env_get_cpu(env));
+
+    env->runstall = runstall;
+    cpu->halted = runstall;
+    if (runstall) {
+        cpu_interrupt(cpu, CPU_INTERRUPT_HALT);
+    } else {
+        cpu_reset_interrupt(cpu, CPU_INTERRUPT_HALT);
+    }
+}
diff --git a/target/xtensa/helper.h b/target/xtensa/helper.h
index 0c8adae9d4..cc751c98fb 100644
--- a/target/xtensa/helper.h
+++ b/target/xtensa/helper.h
@@ -16,10 +16,12 @@ DEF_HELPER_1(simcall, void, env)
 DEF_HELPER_1(dump_state, void, env)
 
 DEF_HELPER_3(waiti, void, env, i32, i32)
-DEF_HELPER_3(timer_irq, void, env, i32, i32)
-DEF_HELPER_2(advance_ccount, void, env, i32)
+DEF_HELPER_1(update_ccount, void, env)
+DEF_HELPER_2(wsr_ccount, void, env, i32)
+DEF_HELPER_2(update_ccompare, void, env, i32)
 DEF_HELPER_1(check_interrupts, void, env)
 DEF_HELPER_3(check_atomctl, void, env, i32, i32)
+DEF_HELPER_2(wsr_memctl, void, env, i32)
 
 DEF_HELPER_2(itlb_hit_test, void, env, i32)
 DEF_HELPER_2(wsr_rasid, void, env, i32)
@@ -54,3 +56,6 @@ DEF_HELPER_4(olt_s, void, env, i32, f32, f32)
 DEF_HELPER_4(ult_s, void, env, i32, f32, f32)
 DEF_HELPER_4(ole_s, void, env, i32, f32, f32)
 DEF_HELPER_4(ule_s, void, env, i32, f32, f32)
+
+DEF_HELPER_2(rer, i32, env, i32)
+DEF_HELPER_3(wer, void, env, i32, i32)
diff --git a/target/xtensa/op_helper.c b/target/xtensa/op_helper.c
index dc0dd351bb..af2723445d 100644
--- a/target/xtensa/op_helper.c
+++ b/target/xtensa/op_helper.c
@@ -105,6 +105,9 @@ void HELPER(exception)(CPUXtensaState *env, uint32_t excp)
     CPUState *cs = CPU(xtensa_env_get_cpu(env));
 
     cs->exception_index = excp;
+    if (excp == EXCP_YIELD) {
+        env->yield_needed = 0;
+    }
     if (excp == EXCP_DEBUG) {
         env->exception_taken = 0;
     }
@@ -385,22 +388,40 @@ void HELPER(waiti)(CPUXtensaState *env, uint32_t pc, uint32_t intlevel)
     }
 
     cpu = CPU(xtensa_env_get_cpu(env));
-    env->halt_clock = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
     cpu->halted = 1;
-    if (xtensa_option_enabled(env->config, XTENSA_OPTION_TIMER_INTERRUPT)) {
-        xtensa_rearm_ccompare_timer(env);
-    }
     HELPER(exception)(env, EXCP_HLT);
 }
 
-void HELPER(timer_irq)(CPUXtensaState *env, uint32_t id, uint32_t active)
+void HELPER(update_ccount)(CPUXtensaState *env)
+{
+    uint64_t now = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
+
+    env->ccount_time = now;
+    env->sregs[CCOUNT] = env->ccount_base +
+        (uint32_t)((now - env->time_base) *
+                   env->config->clock_freq_khz / 1000000);
+}
+
+void HELPER(wsr_ccount)(CPUXtensaState *env, uint32_t v)
 {
-    xtensa_timer_irq(env, id, active);
+    int i;
+
+    HELPER(update_ccount)(env);
+    env->ccount_base += v - env->sregs[CCOUNT];
+    for (i = 0; i < env->config->nccompare; ++i) {
+        HELPER(update_ccompare)(env, i);
+    }
 }
 
-void HELPER(advance_ccount)(CPUXtensaState *env, uint32_t d)
+void HELPER(update_ccompare)(CPUXtensaState *env, uint32_t i)
 {
-    xtensa_advance_ccount(env, d);
+    uint64_t dcc;
+
+    HELPER(update_ccount)(env);
+    dcc = (uint64_t)(env->sregs[CCOMPARE + i] - env->sregs[CCOUNT] - 1) + 1;
+    timer_mod(env->ccompare[i].timer,
+              env->ccount_time + (dcc * 1000000) / env->config->clock_freq_khz);
+    env->yield_needed = 1;
 }
 
 void HELPER(check_interrupts)(CPUXtensaState *env)
@@ -472,6 +493,30 @@ void HELPER(check_atomctl)(CPUXtensaState *env, uint32_t pc, uint32_t vaddr)
     }
 }
 
+void HELPER(wsr_memctl)(CPUXtensaState *env, uint32_t v)
+{
+    if (xtensa_option_enabled(env->config, XTENSA_OPTION_ICACHE)) {
+        if (extract32(v, MEMCTL_IUSEWAYS_SHIFT, MEMCTL_IUSEWAYS_LEN) >
+            env->config->icache_ways) {
+            deposit32(v, MEMCTL_IUSEWAYS_SHIFT, MEMCTL_IUSEWAYS_LEN,
+                      env->config->icache_ways);
+        }
+    }
+    if (xtensa_option_enabled(env->config, XTENSA_OPTION_DCACHE)) {
+        if (extract32(v, MEMCTL_DUSEWAYS_SHIFT, MEMCTL_DUSEWAYS_LEN) >
+            env->config->dcache_ways) {
+            deposit32(v, MEMCTL_DUSEWAYS_SHIFT, MEMCTL_DUSEWAYS_LEN,
+                      env->config->dcache_ways);
+        }
+        if (extract32(v, MEMCTL_DALLOCWAYS_SHIFT, MEMCTL_DALLOCWAYS_LEN) >
+            env->config->dcache_ways) {
+            deposit32(v, MEMCTL_DALLOCWAYS_SHIFT, MEMCTL_DALLOCWAYS_LEN,
+                      env->config->dcache_ways);
+        }
+    }
+    env->sregs[MEMCTL] = v & env->config->memctl_mask;
+}
+
 void HELPER(wsr_rasid)(CPUXtensaState *env, uint32_t v)
 {
     XtensaCPU *cpu = xtensa_env_get_cpu(env);
@@ -969,3 +1014,15 @@ void HELPER(ule_s)(CPUXtensaState *env, uint32_t br, float32 a, float32 b)
     int v = float32_compare_quiet(a, b, &env->fp_status);
     set_br(env, v != float_relation_greater, br);
 }
+
+uint32_t HELPER(rer)(CPUXtensaState *env, uint32_t addr)
+{
+    return address_space_ldl(env->address_space_er, addr,
+                             (MemTxAttrs){0}, NULL);
+}
+
+void HELPER(wer)(CPUXtensaState *env, uint32_t data, uint32_t addr)
+{
+    address_space_stl(env->address_space_er, addr, data,
+                      (MemTxAttrs){0}, NULL);
+}
diff --git a/target/xtensa/overlay_tool.h b/target/xtensa/overlay_tool.h
index e8a7fda3d8..38e9be9ff5 100644
--- a/target/xtensa/overlay_tool.h
+++ b/target/xtensa/overlay_tool.h
@@ -47,10 +47,26 @@
 #define XCHAL_VECBASE_RESET_VADDR 0
 #endif
 
+#ifndef XCHAL_RESET_VECTOR0_VADDR
+#define XCHAL_RESET_VECTOR0_VADDR XCHAL_RESET_VECTOR_VADDR
+#endif
+
+#ifndef XCHAL_RESET_VECTOR1_VADDR
+#define XCHAL_RESET_VECTOR1_VADDR XCHAL_RESET_VECTOR_VADDR
+#endif
+
 #ifndef XCHAL_HW_MIN_VERSION
 #define XCHAL_HW_MIN_VERSION 0
 #endif
 
+#ifndef XCHAL_LOOP_BUFFER_SIZE
+#define XCHAL_LOOP_BUFFER_SIZE 0
+#endif
+
+#ifndef XCHAL_HAVE_EXTERN_REGS
+#define XCHAL_HAVE_EXTERN_REGS 0
+#endif
+
 #define XCHAL_OPTION(xchal, qemu) ((xchal) ? XTENSA_OPTION_BIT(qemu) : 0)
 
 #define XTENSA_OPTIONS ( \
@@ -84,10 +100,10 @@
         XTENSA_OPTION_HIGH_PRIORITY_INTERRUPT) | \
     XCHAL_OPTION(XCHAL_HAVE_CCOUNT, XTENSA_OPTION_TIMER_INTERRUPT) | \
     /* Local memory, TODO */ \
-    XCHAL_OPTION(XCHAL_ICACHE_WAYS, XTENSA_OPTION_ICACHE) | \
+    XCHAL_OPTION(XCHAL_ICACHE_SIZE, XTENSA_OPTION_ICACHE) | \
     XCHAL_OPTION(XCHAL_ICACHE_LINE_LOCKABLE, \
             XTENSA_OPTION_ICACHE_INDEX_LOCK) | \
-    XCHAL_OPTION(XCHAL_DCACHE_WAYS, XTENSA_OPTION_DCACHE) | \
+    XCHAL_OPTION(XCHAL_DCACHE_SIZE, XTENSA_OPTION_DCACHE) | \
     XCHAL_OPTION(XCHAL_DCACHE_LINE_LOCKABLE, \
             XTENSA_OPTION_DCACHE_INDEX_LOCK) | \
     XCHAL_OPTION(XCHAL_UNALIGNED_LOAD_HW, XTENSA_OPTION_HW_ALIGNMENT) | \
@@ -103,7 +119,8 @@
     XCHAL_OPTION(XCHAL_HAVE_DEBUG, XTENSA_OPTION_DEBUG) |\
     XCHAL_OPTION(XCHAL_NUM_MISC_REGS > 0, XTENSA_OPTION_MISC_SR) | \
     XCHAL_OPTION(XCHAL_HAVE_THREADPTR, XTENSA_OPTION_THREAD_POINTER) | \
-    XCHAL_OPTION(XCHAL_HAVE_PRID, XTENSA_OPTION_PROCESSOR_ID))
+    XCHAL_OPTION(XCHAL_HAVE_PRID, XTENSA_OPTION_PROCESSOR_ID) | \
+    XCHAL_OPTION(XCHAL_HAVE_EXTERN_REGS, XTENSA_OPTION_EXTERN_REGS))
 
 #ifndef XCHAL_WINDOW_OF4_VECOFS
 #define XCHAL_WINDOW_OF4_VECOFS         0x00000000
@@ -133,7 +150,8 @@
 #endif
 
 #define EXCEPTION_VECTORS { \
-        [EXC_RESET] = XCHAL_RESET_VECTOR_VADDR, \
+        [EXC_RESET0] = XCHAL_RESET_VECTOR0_VADDR, \
+        [EXC_RESET1] = XCHAL_RESET_VECTOR1_VADDR, \
         WINDOW_VECTORS \
         [EXC_KERNEL] = XCHAL_KERNEL_VECTOR_VADDR, \
         [EXC_USER] = XCHAL_USER_VECTOR_VADDR, \
@@ -334,6 +352,16 @@
     .nibreak = XCHAL_NUM_IBREAK, \
     .ndbreak = XCHAL_NUM_DBREAK
 
+#define CACHE_SECTION \
+    .icache_ways = XCHAL_ICACHE_WAYS, \
+    .dcache_ways = XCHAL_DCACHE_WAYS, \
+    .memctl_mask = \
+        (XCHAL_ICACHE_SIZE ? MEMCTL_IUSEWAYS_MASK : 0) | \
+        (XCHAL_DCACHE_SIZE ? \
+         MEMCTL_DALLOCWAYS_MASK | MEMCTL_DUSEWAYS_MASK : 0) | \
+        MEMCTL_ISNP | MEMCTL_DSNP | \
+        (XCHAL_HAVE_LOOPS && XCHAL_LOOP_BUFFER_SIZE ? MEMCTL_IL0EN : 0)
+
 #define CONFIG_SECTION \
     .configid = { \
         XCHAL_HW_CONFIGID0, \
@@ -348,6 +376,7 @@
     INTERRUPTS_SECTION, \
     TLB_SECTION, \
     DEBUG_SECTION, \
+    CACHE_SECTION, \
     CONFIG_SECTION
 
 
diff --git a/target/xtensa/translate.c b/target/xtensa/translate.c
index 5a93705fac..263002486c 100644
--- a/target/xtensa/translate.c
+++ b/target/xtensa/translate.c
@@ -64,7 +64,6 @@ typedef struct DisasContext {
     bool sar_m32_allocated;
     TCGv_i32 sar_m32;
 
-    uint32_t ccount_delta;
     unsigned window;
 
     bool debug;
@@ -134,6 +133,7 @@ static const XtensaReg sregnames[256] = {
     [ITLBCFG] = XTENSA_REG("ITLBCFG", XTENSA_OPTION_MMU),
     [DTLBCFG] = XTENSA_REG("DTLBCFG", XTENSA_OPTION_MMU),
     [IBREAKENABLE] = XTENSA_REG("IBREAKENABLE", XTENSA_OPTION_DEBUG),
+    [MEMCTL] = XTENSA_REG_BITS("MEMCTL", XTENSA_OPTION_ALL),
     [CACHEATTR] = XTENSA_REG("CACHEATTR", XTENSA_OPTION_CACHEATTR),
     [ATOMCTL] = XTENSA_REG("ATOMCTL", XTENSA_OPTION_ATOMCTL),
     [IBREAKA] = XTENSA_REG("IBREAKA0", XTENSA_OPTION_DEBUG),
@@ -314,20 +314,9 @@ static void gen_left_shift_sar(DisasContext *dc, TCGv_i32 sa)
     tcg_temp_free(tmp);
 }
 
-static void gen_advance_ccount(DisasContext *dc)
-{
-    if (dc->ccount_delta > 0) {
-        TCGv_i32 tmp = tcg_const_i32(dc->ccount_delta);
-        gen_helper_advance_ccount(cpu_env, tmp);
-        tcg_temp_free(tmp);
-    }
-    dc->ccount_delta = 0;
-}
-
 static void gen_exception(DisasContext *dc, int excp)
 {
     TCGv_i32 tmp = tcg_const_i32(excp);
-    gen_advance_ccount(dc);
     gen_helper_exception(cpu_env, tmp);
     tcg_temp_free(tmp);
 }
@@ -336,7 +325,6 @@ static void gen_exception_cause(DisasContext *dc, uint32_t cause)
 {
     TCGv_i32 tpc = tcg_const_i32(dc->pc);
     TCGv_i32 tcause = tcg_const_i32(cause);
-    gen_advance_ccount(dc);
     gen_helper_exception_cause(cpu_env, tpc, tcause);
     tcg_temp_free(tpc);
     tcg_temp_free(tcause);
@@ -351,7 +339,6 @@ static void gen_exception_cause_vaddr(DisasContext *dc, uint32_t cause,
 {
     TCGv_i32 tpc = tcg_const_i32(dc->pc);
     TCGv_i32 tcause = tcg_const_i32(cause);
-    gen_advance_ccount(dc);
     gen_helper_exception_cause_vaddr(cpu_env, tpc, tcause, vaddr);
     tcg_temp_free(tpc);
     tcg_temp_free(tcause);
@@ -361,7 +348,6 @@ static void gen_debug_exception(DisasContext *dc, uint32_t cause)
 {
     TCGv_i32 tpc = tcg_const_i32(dc->pc);
     TCGv_i32 tcause = tcg_const_i32(cause);
-    gen_advance_ccount(dc);
     gen_helper_debug_exception(cpu_env, tpc, tcause);
     tcg_temp_free(tpc);
     tcg_temp_free(tcause);
@@ -394,7 +380,6 @@ static bool gen_check_cpenable(DisasContext *dc, unsigned cp)
 static void gen_jump_slot(DisasContext *dc, TCGv dest, int slot)
 {
     tcg_gen_mov_i32(cpu_pc, dest);
-    gen_advance_ccount(dc);
     if (dc->icount) {
         tcg_gen_mov_i32(cpu_SR[ICOUNT], dc->next_icount);
     }
@@ -465,7 +450,6 @@ static bool gen_check_loop_end(DisasContext *dc, int slot)
             dc->next_pc == dc->lend) {
         TCGLabel *label = gen_new_label();
 
-        gen_advance_ccount(dc);
         tcg_gen_brcondi_i32(TCG_COND_EQ, cpu_SR[LCOUNT], 0, label);
         tcg_gen_subi_i32(cpu_SR[LCOUNT], cpu_SR[LCOUNT], 1);
         gen_jumpi(dc, dc->lbeg, slot);
@@ -488,7 +472,6 @@ static void gen_brcond(DisasContext *dc, TCGCond cond,
 {
     TCGLabel *label = gen_new_label();
 
-    gen_advance_ccount(dc);
     tcg_gen_brcond_i32(cond, t0, t1, label);
     gen_jumpi_check_loop_end(dc, 0);
     gen_set_label(label);
@@ -528,47 +511,60 @@ static bool gen_check_sr(DisasContext *dc, uint32_t sr, unsigned access)
     return true;
 }
 
-static void gen_rsr_ccount(DisasContext *dc, TCGv_i32 d, uint32_t sr)
+static bool gen_rsr_ccount(DisasContext *dc, TCGv_i32 d, uint32_t sr)
 {
-    gen_advance_ccount(dc);
+    if (dc->tb->cflags & CF_USE_ICOUNT) {
+        gen_io_start();
+    }
+    gen_helper_update_ccount(cpu_env);
     tcg_gen_mov_i32(d, cpu_SR[sr]);
+    if (dc->tb->cflags & CF_USE_ICOUNT) {
+        gen_io_end();
+        return true;
+    }
+    return false;
 }
 
-static void gen_rsr_ptevaddr(DisasContext *dc, TCGv_i32 d, uint32_t sr)
+static bool gen_rsr_ptevaddr(DisasContext *dc, TCGv_i32 d, uint32_t sr)
 {
     tcg_gen_shri_i32(d, cpu_SR[EXCVADDR], 10);
     tcg_gen_or_i32(d, d, cpu_SR[sr]);
     tcg_gen_andi_i32(d, d, 0xfffffffc);
+    return false;
 }
 
-static void gen_rsr(DisasContext *dc, TCGv_i32 d, uint32_t sr)
+static bool gen_rsr(DisasContext *dc, TCGv_i32 d, uint32_t sr)
 {
-    static void (* const rsr_handler[256])(DisasContext *dc,
+    static bool (* const rsr_handler[256])(DisasContext *dc,
             TCGv_i32 d, uint32_t sr) = {
         [CCOUNT] = gen_rsr_ccount,
+        [INTSET] = gen_rsr_ccount,
         [PTEVADDR] = gen_rsr_ptevaddr,
     };
 
     if (rsr_handler[sr]) {
-        rsr_handler[sr](dc, d, sr);
+        return rsr_handler[sr](dc, d, sr);
     } else {
         tcg_gen_mov_i32(d, cpu_SR[sr]);
+        return false;
     }
 }
 
-static void gen_wsr_lbeg(DisasContext *dc, uint32_t sr, TCGv_i32 s)
+static bool gen_wsr_lbeg(DisasContext *dc, uint32_t sr, TCGv_i32 s)
 {
     gen_helper_wsr_lbeg(cpu_env, s);
     gen_jumpi_check_loop_end(dc, 0);
+    return false;
 }
 
-static void gen_wsr_lend(DisasContext *dc, uint32_t sr, TCGv_i32 s)
+static bool gen_wsr_lend(DisasContext *dc, uint32_t sr, TCGv_i32 s)
 {
     gen_helper_wsr_lend(cpu_env, s);
     gen_jumpi_check_loop_end(dc, 0);
+    return false;
 }
 
-static void gen_wsr_sar(DisasContext *dc, uint32_t sr, TCGv_i32 s)
+static bool gen_wsr_sar(DisasContext *dc, uint32_t sr, TCGv_i32 s)
 {
     tcg_gen_andi_i32(cpu_SR[sr], s, 0x3f);
     if (dc->sar_m32_5bit) {
@@ -576,68 +572,85 @@ static void gen_wsr_sar(DisasContext *dc, uint32_t sr, TCGv_i32 s)
     }
     dc->sar_5bit = false;
     dc->sar_m32_5bit = false;
+    return false;
 }
 
-static void gen_wsr_br(DisasContext *dc, uint32_t sr, TCGv_i32 s)
+static bool gen_wsr_br(DisasContext *dc, uint32_t sr, TCGv_i32 s)
 {
     tcg_gen_andi_i32(cpu_SR[sr], s, 0xffff);
+    return false;
 }
 
-static void gen_wsr_litbase(DisasContext *dc, uint32_t sr, TCGv_i32 s)
+static bool gen_wsr_litbase(DisasContext *dc, uint32_t sr, TCGv_i32 s)
 {
     tcg_gen_andi_i32(cpu_SR[sr], s, 0xfffff001);
     /* This can change tb->flags, so exit tb */
     gen_jumpi_check_loop_end(dc, -1);
+    return true;
 }
 
-static void gen_wsr_acchi(DisasContext *dc, uint32_t sr, TCGv_i32 s)
+static bool gen_wsr_acchi(DisasContext *dc, uint32_t sr, TCGv_i32 s)
 {
     tcg_gen_ext8s_i32(cpu_SR[sr], s);
+    return false;
 }
 
-static void gen_wsr_windowbase(DisasContext *dc, uint32_t sr, TCGv_i32 v)
+static bool gen_wsr_windowbase(DisasContext *dc, uint32_t sr, TCGv_i32 v)
 {
     gen_helper_wsr_windowbase(cpu_env, v);
     /* This can change tb->flags, so exit tb */
     gen_jumpi_check_loop_end(dc, -1);
+    return true;
 }
 
-static void gen_wsr_windowstart(DisasContext *dc, uint32_t sr, TCGv_i32 v)
+static bool gen_wsr_windowstart(DisasContext *dc, uint32_t sr, TCGv_i32 v)
 {
     tcg_gen_andi_i32(cpu_SR[sr], v, (1 << dc->config->nareg / 4) - 1);
     /* This can change tb->flags, so exit tb */
     gen_jumpi_check_loop_end(dc, -1);
+    return true;
 }
 
-static void gen_wsr_ptevaddr(DisasContext *dc, uint32_t sr, TCGv_i32 v)
+static bool gen_wsr_ptevaddr(DisasContext *dc, uint32_t sr, TCGv_i32 v)
 {
     tcg_gen_andi_i32(cpu_SR[sr], v, 0xffc00000);
+    return false;
 }
 
-static void gen_wsr_rasid(DisasContext *dc, uint32_t sr, TCGv_i32 v)
+static bool gen_wsr_rasid(DisasContext *dc, uint32_t sr, TCGv_i32 v)
 {
     gen_helper_wsr_rasid(cpu_env, v);
     /* This can change tb->flags, so exit tb */
     gen_jumpi_check_loop_end(dc, -1);
+    return true;
 }
 
-static void gen_wsr_tlbcfg(DisasContext *dc, uint32_t sr, TCGv_i32 v)
+static bool gen_wsr_tlbcfg(DisasContext *dc, uint32_t sr, TCGv_i32 v)
 {
     tcg_gen_andi_i32(cpu_SR[sr], v, 0x01130000);
+    return false;
 }
 
-static void gen_wsr_ibreakenable(DisasContext *dc, uint32_t sr, TCGv_i32 v)
+static bool gen_wsr_ibreakenable(DisasContext *dc, uint32_t sr, TCGv_i32 v)
 {
     gen_helper_wsr_ibreakenable(cpu_env, v);
     gen_jumpi_check_loop_end(dc, 0);
+    return true;
+}
+
+static bool gen_wsr_memctl(DisasContext *dc, uint32_t sr, TCGv_i32 v)
+{
+    gen_helper_wsr_memctl(cpu_env, v);
+    return false;
 }
 
-static void gen_wsr_atomctl(DisasContext *dc, uint32_t sr, TCGv_i32 v)
+static bool gen_wsr_atomctl(DisasContext *dc, uint32_t sr, TCGv_i32 v)
 {
     tcg_gen_andi_i32(cpu_SR[sr], v, 0x3f);
+    return false;
 }
 
-static void gen_wsr_ibreaka(DisasContext *dc, uint32_t sr, TCGv_i32 v)
+static bool gen_wsr_ibreaka(DisasContext *dc, uint32_t sr, TCGv_i32 v)
 {
     unsigned id = sr - IBREAKA;
 
@@ -646,10 +659,12 @@ static void gen_wsr_ibreaka(DisasContext *dc, uint32_t sr, TCGv_i32 v)
         gen_helper_wsr_ibreaka(cpu_env, tmp, v);
         tcg_temp_free(tmp);
         gen_jumpi_check_loop_end(dc, 0);
+        return true;
     }
+    return false;
 }
 
-static void gen_wsr_dbreaka(DisasContext *dc, uint32_t sr, TCGv_i32 v)
+static bool gen_wsr_dbreaka(DisasContext *dc, uint32_t sr, TCGv_i32 v)
 {
     unsigned id = sr - DBREAKA;
 
@@ -658,9 +673,10 @@ static void gen_wsr_dbreaka(DisasContext *dc, uint32_t sr, TCGv_i32 v)
         gen_helper_wsr_dbreaka(cpu_env, tmp, v);
         tcg_temp_free(tmp);
     }
+    return false;
 }
 
-static void gen_wsr_dbreakc(DisasContext *dc, uint32_t sr, TCGv_i32 v)
+static bool gen_wsr_dbreakc(DisasContext *dc, uint32_t sr, TCGv_i32 v)
 {
     unsigned id = sr - DBREAKC;
 
@@ -669,24 +685,38 @@ static void gen_wsr_dbreakc(DisasContext *dc, uint32_t sr, TCGv_i32 v)
         gen_helper_wsr_dbreakc(cpu_env, tmp, v);
         tcg_temp_free(tmp);
     }
+    return false;
 }
 
-static void gen_wsr_cpenable(DisasContext *dc, uint32_t sr, TCGv_i32 v)
+static bool gen_wsr_cpenable(DisasContext *dc, uint32_t sr, TCGv_i32 v)
 {
     tcg_gen_andi_i32(cpu_SR[sr], v, 0xff);
     /* This can change tb->flags, so exit tb */
     gen_jumpi_check_loop_end(dc, -1);
+    return true;
 }
 
-static void gen_wsr_intset(DisasContext *dc, uint32_t sr, TCGv_i32 v)
+static void gen_check_interrupts(DisasContext *dc)
+{
+    if (dc->tb->cflags & CF_USE_ICOUNT) {
+        gen_io_start();
+    }
+    gen_helper_check_interrupts(cpu_env);
+    if (dc->tb->cflags & CF_USE_ICOUNT) {
+        gen_io_end();
+    }
+}
+
+static bool gen_wsr_intset(DisasContext *dc, uint32_t sr, TCGv_i32 v)
 {
     tcg_gen_andi_i32(cpu_SR[sr], v,
             dc->config->inttype_mask[INTTYPE_SOFTWARE]);
-    gen_helper_check_interrupts(cpu_env);
+    gen_check_interrupts(dc);
     gen_jumpi_check_loop_end(dc, 0);
+    return true;
 }
 
-static void gen_wsr_intclear(DisasContext *dc, uint32_t sr, TCGv_i32 v)
+static bool gen_wsr_intclear(DisasContext *dc, uint32_t sr, TCGv_i32 v)
 {
     TCGv_i32 tmp = tcg_temp_new_i32();
 
@@ -696,17 +726,20 @@ static void gen_wsr_intclear(DisasContext *dc, uint32_t sr, TCGv_i32 v)
             dc->config->inttype_mask[INTTYPE_SOFTWARE]);
     tcg_gen_andc_i32(cpu_SR[INTSET], cpu_SR[INTSET], tmp);
     tcg_temp_free(tmp);
-    gen_helper_check_interrupts(cpu_env);
+    gen_check_interrupts(dc);
+    gen_jumpi_check_loop_end(dc, 0);
+    return true;
 }
 
-static void gen_wsr_intenable(DisasContext *dc, uint32_t sr, TCGv_i32 v)
+static bool gen_wsr_intenable(DisasContext *dc, uint32_t sr, TCGv_i32 v)
 {
     tcg_gen_mov_i32(cpu_SR[sr], v);
-    gen_helper_check_interrupts(cpu_env);
+    gen_check_interrupts(dc);
     gen_jumpi_check_loop_end(dc, 0);
+    return true;
 }
 
-static void gen_wsr_ps(DisasContext *dc, uint32_t sr, TCGv_i32 v)
+static bool gen_wsr_ps(DisasContext *dc, uint32_t sr, TCGv_i32 v)
 {
     uint32_t mask = PS_WOE | PS_CALLINC | PS_OWB |
         PS_UM | PS_EXCM | PS_INTLEVEL;
@@ -715,42 +748,72 @@ static void gen_wsr_ps(DisasContext *dc, uint32_t sr, TCGv_i32 v)
         mask |= PS_RING;
     }
     tcg_gen_andi_i32(cpu_SR[sr], v, mask);
-    gen_helper_check_interrupts(cpu_env);
+    gen_check_interrupts(dc);
     /* This can change mmu index and tb->flags, so exit tb */
     gen_jumpi_check_loop_end(dc, -1);
+    return true;
+}
+
+static bool gen_wsr_ccount(DisasContext *dc, uint32_t sr, TCGv_i32 v)
+{
+    if (dc->tb->cflags & CF_USE_ICOUNT) {
+        gen_io_start();
+    }
+    gen_helper_wsr_ccount(cpu_env, v);
+    if (dc->tb->cflags & CF_USE_ICOUNT) {
+        gen_io_end();
+        gen_jumpi_check_loop_end(dc, 0);
+        return true;
+    }
+    return false;
 }
 
-static void gen_wsr_icount(DisasContext *dc, uint32_t sr, TCGv_i32 v)
+static bool gen_wsr_icount(DisasContext *dc, uint32_t sr, TCGv_i32 v)
 {
     if (dc->icount) {
         tcg_gen_mov_i32(dc->next_icount, v);
     } else {
         tcg_gen_mov_i32(cpu_SR[sr], v);
     }
+    return false;
 }
 
-static void gen_wsr_icountlevel(DisasContext *dc, uint32_t sr, TCGv_i32 v)
+static bool gen_wsr_icountlevel(DisasContext *dc, uint32_t sr, TCGv_i32 v)
 {
     tcg_gen_andi_i32(cpu_SR[sr], v, 0xf);
     /* This can change tb->flags, so exit tb */
     gen_jumpi_check_loop_end(dc, -1);
+    return true;
 }
 
-static void gen_wsr_ccompare(DisasContext *dc, uint32_t sr, TCGv_i32 v)
+static bool gen_wsr_ccompare(DisasContext *dc, uint32_t sr, TCGv_i32 v)
 {
     uint32_t id = sr - CCOMPARE;
+    bool ret = false;
+
     if (id < dc->config->nccompare) {
         uint32_t int_bit = 1 << dc->config->timerint[id];
-        gen_advance_ccount(dc);
+        TCGv_i32 tmp = tcg_const_i32(id);
+
         tcg_gen_mov_i32(cpu_SR[sr], v);
         tcg_gen_andi_i32(cpu_SR[INTSET], cpu_SR[INTSET], ~int_bit);
-        gen_helper_check_interrupts(cpu_env);
+        if (dc->tb->cflags & CF_USE_ICOUNT) {
+            gen_io_start();
+        }
+        gen_helper_update_ccompare(cpu_env, tmp);
+        if (dc->tb->cflags & CF_USE_ICOUNT) {
+            gen_io_end();
+            gen_jumpi_check_loop_end(dc, 0);
+            ret = true;
+        }
+        tcg_temp_free(tmp);
     }
+    return ret;
 }
 
-static void gen_wsr(DisasContext *dc, uint32_t sr, TCGv_i32 s)
+static bool gen_wsr(DisasContext *dc, uint32_t sr, TCGv_i32 s)
 {
-    static void (* const wsr_handler[256])(DisasContext *dc,
+    static bool (* const wsr_handler[256])(DisasContext *dc,
             uint32_t sr, TCGv_i32 v) = {
         [LBEG] = gen_wsr_lbeg,
         [LEND] = gen_wsr_lend,
@@ -765,6 +828,7 @@ static void gen_wsr(DisasContext *dc, uint32_t sr, TCGv_i32 s)
         [ITLBCFG] = gen_wsr_tlbcfg,
         [DTLBCFG] = gen_wsr_tlbcfg,
         [IBREAKENABLE] = gen_wsr_ibreakenable,
+        [MEMCTL] = gen_wsr_memctl,
         [ATOMCTL] = gen_wsr_atomctl,
         [IBREAKA] = gen_wsr_ibreaka,
         [IBREAKA + 1] = gen_wsr_ibreaka,
@@ -777,6 +841,7 @@ static void gen_wsr(DisasContext *dc, uint32_t sr, TCGv_i32 s)
         [INTCLEAR] = gen_wsr_intclear,
         [INTENABLE] = gen_wsr_intenable,
         [PS] = gen_wsr_ps,
+        [CCOUNT] = gen_wsr_ccount,
         [ICOUNT] = gen_wsr_icount,
         [ICOUNTLEVEL] = gen_wsr_icountlevel,
         [CCOMPARE] = gen_wsr_ccompare,
@@ -785,9 +850,10 @@ static void gen_wsr(DisasContext *dc, uint32_t sr, TCGv_i32 s)
     };
 
     if (wsr_handler[sr]) {
-        wsr_handler[sr](dc, sr, s);
+        return wsr_handler[sr](dc, sr, s);
     } else {
         tcg_gen_mov_i32(cpu_SR[sr], s);
+        return false;
     }
 }
 
@@ -829,10 +895,17 @@ static void gen_waiti(DisasContext *dc, uint32_t imm4)
 {
     TCGv_i32 pc = tcg_const_i32(dc->next_pc);
     TCGv_i32 intlevel = tcg_const_i32(imm4);
-    gen_advance_ccount(dc);
+
+    if (dc->tb->cflags & CF_USE_ICOUNT) {
+        gen_io_start();
+    }
     gen_helper_waiti(cpu_env, pc, intlevel);
+    if (dc->tb->cflags & CF_USE_ICOUNT) {
+        gen_io_end();
+    }
     tcg_temp_free(pc);
     tcg_temp_free(intlevel);
+    gen_jumpi_check_loop_end(dc, 0);
 }
 
 static bool gen_window_check1(DisasContext *dc, unsigned r1)
@@ -841,7 +914,6 @@ static bool gen_window_check1(DisasContext *dc, unsigned r1)
         TCGv_i32 pc = tcg_const_i32(dc->pc);
         TCGv_i32 w = tcg_const_i32(r1 / 4);
 
-        gen_advance_ccount(dc);
         gen_helper_window_check(cpu_env, pc, w);
         dc->is_jmp = DISAS_UPDATE;
         return false;
@@ -1037,7 +1109,6 @@ static void disas_xtensa_insn(CPUXtensaState *env, DisasContext *dc)
                             HAS_OPTION(XTENSA_OPTION_WINDOWED_REGISTER);
                             {
                                 TCGv_i32 tmp = tcg_const_i32(dc->pc);
-                                gen_advance_ccount(dc);
                                 gen_helper_retw(tmp, cpu_env, tmp);
                                 gen_jump(dc, tmp);
                                 tcg_temp_free(tmp);
@@ -1086,7 +1157,6 @@ static void disas_xtensa_insn(CPUXtensaState *env, DisasContext *dc)
                     HAS_OPTION(XTENSA_OPTION_WINDOWED_REGISTER);
                     if (gen_window_check2(dc, RRR_T, RRR_S)) {
                         TCGv_i32 pc = tcg_const_i32(dc->pc);
-                        gen_advance_ccount(dc);
                         gen_helper_movsp(cpu_env, pc);
                         tcg_gen_mov_i32(cpu_R[RRR_T], cpu_R[RRR_S]);
                         tcg_temp_free(pc);
@@ -1134,7 +1204,7 @@ static void disas_xtensa_insn(CPUXtensaState *env, DisasContext *dc)
                         case 0: /*RFEx*/
                             if (gen_check_privilege(dc)) {
                                 tcg_gen_andi_i32(cpu_SR[PS], cpu_SR[PS], ~PS_EXCM);
-                                gen_helper_check_interrupts(cpu_env);
+                                gen_check_interrupts(dc);
                                 gen_jump(dc, cpu_SR[EPC1]);
                             }
                             break;
@@ -1169,7 +1239,7 @@ static void disas_xtensa_insn(CPUXtensaState *env, DisasContext *dc)
                                 }
 
                                 gen_helper_restore_owb(cpu_env);
-                                gen_helper_check_interrupts(cpu_env);
+                                gen_check_interrupts(dc);
                                 gen_jump(dc, cpu_SR[EPC1]);
 
                                 tcg_temp_free(tmp);
@@ -1188,7 +1258,7 @@ static void disas_xtensa_insn(CPUXtensaState *env, DisasContext *dc)
                             if (gen_check_privilege(dc)) {
                                 tcg_gen_mov_i32(cpu_SR[PS],
                                                 cpu_SR[EPS2 + RRR_S - 2]);
-                                gen_helper_check_interrupts(cpu_env);
+                                gen_check_interrupts(dc);
                                 gen_jump(dc, cpu_SR[EPC1 + RRR_S - 1]);
                             }
                         } else {
@@ -1246,7 +1316,7 @@ static void disas_xtensa_insn(CPUXtensaState *env, DisasContext *dc)
                         tcg_gen_mov_i32(cpu_R[RRR_T], cpu_SR[PS]);
                         tcg_gen_andi_i32(cpu_SR[PS], cpu_SR[PS], ~PS_INTLEVEL);
                         tcg_gen_ori_i32(cpu_SR[PS], cpu_SR[PS], RRR_S);
-                        gen_helper_check_interrupts(cpu_env);
+                        gen_check_interrupts(dc);
                         gen_jumpi_check_loop_end(dc, 0);
                     }
                     break;
@@ -1350,11 +1420,19 @@ static void disas_xtensa_insn(CPUXtensaState *env, DisasContext *dc)
                     break;
 
                 case 6: /*RER*/
-                    TBD();
+                    HAS_OPTION(XTENSA_OPTION_EXTERN_REGS);
+                    if (gen_check_privilege(dc) &&
+                        gen_window_check2(dc, RRR_S, RRR_T)) {
+                        gen_helper_rer(cpu_R[RRR_T], cpu_env, cpu_R[RRR_S]);
+                    }
                     break;
 
                 case 7: /*WER*/
-                    TBD();
+                    HAS_OPTION(XTENSA_OPTION_EXTERN_REGS);
+                    if (gen_check_privilege(dc) &&
+                        gen_window_check2(dc, RRR_S, RRR_T)) {
+                        gen_helper_wer(cpu_env, cpu_R[RRR_T], cpu_R[RRR_S]);
+                    }
                     break;
 
                 case 8: /*ROTWw*/
@@ -1534,11 +1612,15 @@ static void disas_xtensa_insn(CPUXtensaState *env, DisasContext *dc)
                     (RSR_SR < 64 || gen_check_privilege(dc)) &&
                     gen_window_check1(dc, RRR_T)) {
                     TCGv_i32 tmp = tcg_temp_new_i32();
+                    bool rsr_end, wsr_end;
 
                     tcg_gen_mov_i32(tmp, cpu_R[RRR_T]);
-                    gen_rsr(dc, cpu_R[RRR_T], RSR_SR);
-                    gen_wsr(dc, RSR_SR, tmp);
+                    rsr_end = gen_rsr(dc, cpu_R[RRR_T], RSR_SR);
+                    wsr_end = gen_wsr(dc, RSR_SR, tmp);
                     tcg_temp_free(tmp);
+                    if (rsr_end && !wsr_end) {
+                        gen_jumpi_check_loop_end(dc, 0);
+                    }
                 }
                 break;
 
@@ -1759,7 +1841,9 @@ static void disas_xtensa_insn(CPUXtensaState *env, DisasContext *dc)
                 if (gen_check_sr(dc, RSR_SR, SR_R) &&
                     (RSR_SR < 64 || gen_check_privilege(dc)) &&
                     gen_window_check1(dc, RRR_T)) {
-                    gen_rsr(dc, cpu_R[RRR_T], RSR_SR);
+                    if (gen_rsr(dc, cpu_R[RRR_T], RSR_SR)) {
+                        gen_jumpi_check_loop_end(dc, 0);
+                    }
                 }
                 break;
 
@@ -2517,7 +2601,6 @@ static void disas_xtensa_insn(CPUXtensaState *env, DisasContext *dc)
                 tcg_gen_addi_i32(addr, cpu_R[RRI8_S], RRI8_IMM8 << 2);
                 gen_load_store_alignment(dc, 2, addr, true);
 
-                gen_advance_ccount(dc);
                 tpc = tcg_const_i32(dc->pc);
                 gen_helper_check_atomctl(cpu_env, tpc, addr);
                 tcg_gen_qemu_ld32u(cpu_R[RRI8_T], addr, dc->cring);
@@ -2747,7 +2830,6 @@ static void disas_xtensa_insn(CPUXtensaState *env, DisasContext *dc)
                     TCGv_i32 pc = tcg_const_i32(dc->pc);
                     TCGv_i32 s = tcg_const_i32(BRI12_S);
                     TCGv_i32 imm = tcg_const_i32(BRI12_IMM12);
-                    gen_advance_ccount(dc);
                     gen_helper_entry(cpu_env, pc, s, imm);
                     tcg_temp_free(imm);
                     tcg_temp_free(s);
@@ -2966,7 +3048,6 @@ static void disas_xtensa_insn(CPUXtensaState *env, DisasContext *dc)
                 HAS_OPTION(XTENSA_OPTION_WINDOWED_REGISTER);
                 {
                     TCGv_i32 tmp = tcg_const_i32(dc->pc);
-                    gen_advance_ccount(dc);
                     gen_helper_retw(tmp, cpu_env, tmp);
                     gen_jump(dc, tmp);
                     tcg_temp_free(tmp);
@@ -3063,7 +3144,6 @@ void gen_intermediate_code(CPUXtensaState *env, TranslationBlock *tb)
     dc.lbeg = env->sregs[LBEG];
     dc.lend = env->sregs[LEND];
     dc.is_jmp = DISAS_NEXT;
-    dc.ccount_delta = 0;
     dc.debug = tb->flags & XTENSA_TBFLAG_DEBUG;
     dc.icount = tb->flags & XTENSA_TBFLAG_ICOUNT;
     dc.cpenable = (tb->flags & XTENSA_TBFLAG_CPENABLE_MASK) >>
@@ -3079,17 +3159,26 @@ void gen_intermediate_code(CPUXtensaState *env, TranslationBlock *tb)
 
     gen_tb_start(tb);
 
+    if ((tb->cflags & CF_USE_ICOUNT) &&
+        (tb->flags & XTENSA_TBFLAG_YIELD)) {
+        tcg_gen_insn_start(dc.pc);
+        ++insn_count;
+        gen_exception(&dc, EXCP_YIELD);
+        dc.is_jmp = DISAS_UPDATE;
+        goto done;
+    }
     if (tb->flags & XTENSA_TBFLAG_EXCEPTION) {
-        tcg_gen_movi_i32(cpu_pc, dc.pc);
+        tcg_gen_insn_start(dc.pc);
+        ++insn_count;
         gen_exception(&dc, EXCP_DEBUG);
+        dc.is_jmp = DISAS_UPDATE;
+        goto done;
     }
 
     do {
         tcg_gen_insn_start(dc.pc);
         ++insn_count;
 
-        ++dc.ccount_delta;
-
         if (unlikely(cpu_breakpoint_test(cs, dc.pc, BP_ANY))) {
             tcg_gen_movi_i32(cpu_pc, dc.pc);
             gen_exception(&dc, EXCP_DEBUG);
@@ -3136,7 +3225,7 @@ void gen_intermediate_code(CPUXtensaState *env, TranslationBlock *tb)
             dc.pc < next_page_start &&
             dc.pc + xtensa_insn_len(env, &dc) <= next_page_start &&
             !tcg_op_buf_full());
-
+done:
     reset_litbase(&dc);
     reset_sar_tracker(&dc);
     if (dc.icount) {
diff --git a/tests/tcg/xtensa/Makefile b/tests/tcg/xtensa/Makefile
index 7f9f2d96c3..2882c431e4 100644
--- a/tests/tcg/xtensa/Makefile
+++ b/tests/tcg/xtensa/Makefile
@@ -5,7 +5,7 @@ CROSS=xtensa-$(CORE)-elf-
 
 ifndef XT
 SIM = ../../../xtensa-softmmu/qemu-system-xtensa
-SIMFLAGS = -M sim -cpu $(CORE) -nographic -semihosting $(EXTFLAGS) -kernel
+SIMFLAGS = -M sim -cpu $(CORE) -nographic -semihosting -icount 7 $(EXTFLAGS) -kernel
 SIMDEBUG = -s -S
 else
 SIM = xt-run
diff --git a/tests/tcg/xtensa/test_interrupt.S b/tests/tcg/xtensa/test_interrupt.S
index 334ddab287..876683518e 100644
--- a/tests/tcg/xtensa/test_interrupt.S
+++ b/tests/tcg/xtensa/test_interrupt.S
@@ -1,5 +1,7 @@
 #include "macros.inc"
 
+#define LSBIT(v) ((v) ^ ((v) & ((v) - 1)))
+
 test_suite interrupt
 
 .macro clear_interrupts
@@ -46,14 +48,17 @@ test soft_disabled
     set_vector kernel, 1f
     clear_interrupts
 
-    movi    a2, 0x80
+    movi    a2, LSBIT(XCHAL_INTTYPE_MASK_SOFTWARE)
     wsr     a2, intset
     esync
     rsr     a3, interrupt
+    movi    a4, ~XCHAL_INTTYPE_MASK_TIMER
+    and     a3, a3, a4
     assert  eq, a2, a3
     wsr     a2, intclear
     esync
     rsr     a3, interrupt
+    and     a3, a3, a4
     assert  eqi, a3, 0
     j       2f
 1:
@@ -65,10 +70,12 @@ test soft_intenable
     set_vector kernel, 1f
     clear_interrupts
 
-    movi    a2, 0x80
+    movi    a2, LSBIT(XCHAL_INTTYPE_MASK_SOFTWARE)
     wsr     a2, intset
     esync
     rsr     a3, interrupt
+    movi    a4, ~XCHAL_INTTYPE_MASK_TIMER
+    and     a3, a3, a4
     assert  eq, a2, a3
     rsil    a3, 0
     wsr     a2, intenable
@@ -82,10 +89,12 @@ test soft_rsil
     set_vector kernel, 1f
     clear_interrupts
 
-    movi    a2, 0x80
+    movi    a2, LSBIT(XCHAL_INTTYPE_MASK_SOFTWARE)
     wsr     a2, intset
     esync
     rsr     a3, interrupt
+    movi    a4, ~XCHAL_INTTYPE_MASK_TIMER
+    and     a3, a3, a4
     assert  eq, a2, a3
     wsr     a2, intenable
     rsil    a3, 0
@@ -99,10 +108,12 @@ test soft_waiti
     set_vector kernel, 1f
     clear_interrupts
 
-    movi    a2, 0x80
+    movi    a2, LSBIT(XCHAL_INTTYPE_MASK_SOFTWARE)
     wsr     a2, intset
     esync
     rsr     a3, interrupt
+    movi    a4, ~XCHAL_INTTYPE_MASK_TIMER
+    and     a3, a3, a4
     assert  eq, a2, a3
     wsr     a2, intenable
     waiti   0
@@ -116,10 +127,12 @@ test soft_user
     set_vector user, 2f
     clear_interrupts
 
-    movi    a2, 0x80
+    movi    a2, LSBIT(XCHAL_INTTYPE_MASK_SOFTWARE)
     wsr     a2, intset
     esync
     rsr     a3, interrupt
+    movi    a4, ~XCHAL_INTTYPE_MASK_TIMER
+    and     a3, a3, a4
     assert  eq, a2, a3
     wsr     a2, intenable
 
@@ -139,7 +152,7 @@ test soft_priority
     set_vector level3, 2f
     clear_interrupts
 
-    movi    a2, 0x880
+    movi    a2, XCHAL_INTTYPE_MASK_SOFTWARE
     wsr     a2, intenable
     rsil    a3, 0
     esync
@@ -161,7 +174,7 @@ test eps_epc_rfi
     clear_interrupts
     reset_ps
 
-    movi    a2, 0x880
+    movi    a2, XCHAL_INTTYPE_MASK_SOFTWARE
     wsr     a2, intenable
     rsil    a3, 0
     rsr     a3, ps
diff --git a/tests/tcg/xtensa/test_sr.S b/tests/tcg/xtensa/test_sr.S
index 4fac46e80f..42e3e5e386 100644
--- a/tests/tcg/xtensa/test_sr.S
+++ b/tests/tcg/xtensa/test_sr.S
@@ -44,6 +44,7 @@ test_end
 
 test_sr acchi, 1
 test_sr acclo, 1
+test_sr /*memctl*/97, 0
 test_sr_mask /*atomctl*/99, 0, 0
 test_sr_mask /*br*/4, 0, 0
 test_sr_mask /*cacheattr*/98, 0, 0
diff --git a/tests/tcg/xtensa/test_timer.S b/tests/tcg/xtensa/test_timer.S
index f8c6f7423a..6cda71adbb 100644
--- a/tests/tcg/xtensa/test_timer.S
+++ b/tests/tcg/xtensa/test_timer.S
@@ -1,12 +1,56 @@
 #include "macros.inc"
 
+#define CCOUNT_SHIFT 4
+#define WAIT_LOOPS 20
+
+.macro      make_ccount_delta target, delta
+    rsr     \delta, ccount
+    rsr     \target, ccount
+    sub     \delta, \target, \delta
+    slli    \delta, \delta, CCOUNT_SHIFT
+    add     \target, \target, \delta
+.endm
+
 test_suite timer
 
 test ccount
     rsr     a3, ccount
     rsr     a4, ccount
-    sub     a3, a4, a3
-    assert  eqi, a3, 1
+    assert  ne, a3, a4
+test_end
+
+test ccount_write
+    rsr     a3, ccount
+    rsr     a4, ccount
+    sub     a4, a4, a3
+    movi    a2, 0x12345678
+    wsr     a2, ccount
+    esync
+    rsr     a3, ccount
+    sub     a3, a3, a2
+    slli    a4, a4, 2
+    assert  ltu, a3, a4
+test_end
+
+test ccount_update_deadline
+    movi    a2, 0
+    wsr     a2, intenable
+    rsr     a2, interrupt
+    wsr     a2, intclear
+    movi    a2, 0
+    wsr     a2, ccompare1
+    wsr     a2, ccompare2
+    movi    a2, 0x12345678
+    wsr     a2, ccompare0
+    rsr     a3, interrupt
+    assert  eqi, a3, 0
+    movi    a2, 0x12345677
+    wsr     a2, ccount
+    esync
+    nop
+    rsr     a2, interrupt
+    movi    a3, 1 << XCHAL_TIMER0_INTERRUPT
+    assert  eq, a2, a3
 test_end
 
 test ccompare
@@ -18,18 +62,18 @@ test ccompare
     wsr     a2, ccompare1
     wsr     a2, ccompare2
 
-    movi    a3, 20
-    rsr     a2, ccount
-    addi    a2, a2, 20
+    make_ccount_delta a2, a15
     wsr     a2, ccompare0
-    rsr     a2, interrupt
-    assert  eqi, a2, 0
-    loop    a3, 1f
-    rsr     a3, interrupt
-    bnez    a3, 2f
 1:
-    test_fail
+    rsr     a3, interrupt
+    rsr     a4, ccount
+    rsr     a5, interrupt
+    sub     a4, a4, a2
+    bgez    a4, 2f
+    assert  eqi, a3, 0
+    j       1b
 2:
+    assert  nei, a5, 0
 test_end
 
 test ccompare0_interrupt
@@ -42,15 +86,14 @@ test ccompare0_interrupt
     wsr     a2, ccompare1
     wsr     a2, ccompare2
 
-    movi    a3, 20
-    rsr     a2, ccount
-    addi    a2, a2, 20
+    movi    a3, WAIT_LOOPS
+    make_ccount_delta a2, a15
     wsr     a2, ccompare0
     rsync
     rsr     a2, interrupt
     assert  eqi, a2, 0
 
-    movi    a2, 0x40
+    movi    a2, 1 << XCHAL_TIMER0_INTERRUPT
     wsr     a2, intenable
     rsil    a2, 0
     loop    a3, 1f
@@ -72,14 +115,13 @@ test ccompare1_interrupt
     wsr     a2, ccompare0
     wsr     a2, ccompare2
 
-    movi    a3, 20
-    rsr     a2, ccount
-    addi    a2, a2, 20
+    movi    a3, WAIT_LOOPS
+    make_ccount_delta a2, a15
     wsr     a2, ccompare1
     rsync
     rsr     a2, interrupt
     assert  eqi, a2, 0
-    movi    a2, 0x400
+    movi    a2, 1 << XCHAL_TIMER1_INTERRUPT
     wsr     a2, intenable
     rsil    a2, 2
     loop    a3, 1f
@@ -99,14 +141,13 @@ test ccompare2_interrupt
     wsr     a2, ccompare0
     wsr     a2, ccompare1
 
-    movi    a3, 20
-    rsr     a2, ccount
-    addi    a2, a2, 20
+    movi    a3, WAIT_LOOPS
+    make_ccount_delta a2, a15
     wsr     a2, ccompare2
     rsync
     rsr     a2, interrupt
     assert  eqi, a2, 0
-    movi    a2, 0x2000
+    movi    a2, 1 << XCHAL_TIMER2_INTERRUPT
     wsr     a2, intenable
     rsil    a2, 4
     loop    a3, 1f
@@ -125,17 +166,16 @@ test ccompare_interrupt_masked
     movi    a2, 0
     wsr     a2, ccompare2
 
-    movi    a3, 40
-    rsr     a2, ccount
-    addi    a2, a2, 20
+    movi    a3, 2 * WAIT_LOOPS
+    make_ccount_delta a2, a15
     wsr     a2, ccompare1
-    addi    a2, a2, 20
+    add     a2, a2, a15
     wsr     a2, ccompare0
     rsync
     rsr     a2, interrupt
     assert  eqi, a2, 0
 
-    movi    a2, 0x40
+    movi    a2, 1 << XCHAL_TIMER0_INTERRUPT
     wsr     a2, intenable
     rsil    a2, 0
     loop    a3, 1f
@@ -156,17 +196,16 @@ test ccompare_interrupt_masked_waiti
     movi    a2, 0
     wsr     a2, ccompare2
 
-    movi    a3, 40
-    rsr     a2, ccount
-    addi    a2, a2, 20
+    movi    a3, 2 * WAIT_LOOPS
+    make_ccount_delta a2, a15
     wsr     a2, ccompare1
-    addi    a2, a2, 20
+    add     a2, a2, a15
     wsr     a2, ccompare0
     rsync
     rsr     a2, interrupt
     assert  eqi, a2, 0
 
-    movi    a2, 0x40
+    movi    a2, 1 << XCHAL_TIMER0_INTERRUPT
     wsr     a2, intenable
     waiti   0
     test_fail