[Qemu-devel] [PATCH 8/8] target-arm/a64: use HTM with stop-the-world fall-back path

["Emilio G. Cota" <cota@braap.org>]

TODO: convert paired atomics as well.

Signed-off-by: Emilio G. Cota <cota@braap.org>
---
 cpu-exec.c                 |  4 ++++
 target-arm/helper-a64.c    | 31 +++++++++++++++++++++++++++++++
 target-arm/helper-a64.h    |  4 ++++
 target-arm/op_helper.c     |  4 ++++
 target-arm/translate-a64.c | 16 ++++++++++------
 5 files changed, 53 insertions(+), 6 deletions(-)

diff --git a/cpu-exec.c b/cpu-exec.c
index 8f1adc4..6e2531f 100644
--- a/cpu-exec.c
+++ b/cpu-exec.c
@@ -26,6 +26,7 @@
 #include "sysemu/qtest.h"
 #include "qemu/timer.h"
 #include "exec/address-spaces.h"
+#include "qemu/htm.h"
 #include "qemu/rcu.h"
 #include "exec/tb-hash.h"
 #include "exec/log.h"
@@ -651,6 +652,9 @@ int cpu_exec(CPUState *cpu)
             cpu->can_do_io = 1;
             stop_the_world_reset();
             tb_lock_reset();
+            if (unlikely(htm_test())) {
+                htm_end();
+            }
         }
     } /* for(;;) */
 
diff --git a/target-arm/helper-a64.c b/target-arm/helper-a64.c
index 6f3fd17..741e6de 100644
--- a/target-arm/helper-a64.c
+++ b/target-arm/helper-a64.c
@@ -25,6 +25,7 @@
 #include "qemu/log.h"
 #include "sysemu/sysemu.h"
 #include "qemu/bitops.h"
+#include "qemu/htm.h"
 #include "internals.h"
 #include "qemu/crc32c.h"
 #include "exec/exec-all.h"
@@ -579,3 +580,33 @@ uint64_t HELPER(paired_cmpxchg64_be)(CPUARMState *env, uint64_t addr,
 
     return !success;
 }
+
+void HELPER(xbegin)(CPUARMState *env)
+{
+    int status;
+    int retries = 100;
+
+ retry:
+    status = htm_begin();
+    if (unlikely(status != HTM_OK)) {
+        if ((status & HTM_ABORT_RETRY) && retries) {
+            retries--;
+            goto retry;
+        }
+        stop_the_world_lock(ENV_GET_CPU(env));
+    }
+}
+
+void HELPER(xend)(void)
+{
+    if (likely(htm_test())) {
+        htm_end();
+    } else {
+        stop_the_world_unlock();
+    }
+}
+
+uint64_t HELPER(x_ok)(void)
+{
+    return likely(htm_test()) || stw_held;
+}
diff --git a/target-arm/helper-a64.h b/target-arm/helper-a64.h
index dd32000..e7ede43 100644
--- a/target-arm/helper-a64.h
+++ b/target-arm/helper-a64.h
@@ -48,3 +48,7 @@ DEF_HELPER_FLAGS_3(crc32_64, TCG_CALL_NO_RWG_SE, i64, i64, i64, i32)
 DEF_HELPER_FLAGS_3(crc32c_64, TCG_CALL_NO_RWG_SE, i64, i64, i64, i32)
 DEF_HELPER_FLAGS_4(paired_cmpxchg64_le, TCG_CALL_NO_WG, i64, env, i64, i64, i64)
 DEF_HELPER_FLAGS_4(paired_cmpxchg64_be, TCG_CALL_NO_WG, i64, env, i64, i64, i64)
+
+DEF_HELPER_1(xbegin, void, env)
+DEF_HELPER_0(x_ok, i64)
+DEF_HELPER_0(xend, void)
diff --git a/target-arm/op_helper.c b/target-arm/op_helper.c
index 73da759..91b1413 100644
--- a/target-arm/op_helper.c
+++ b/target-arm/op_helper.c
@@ -17,6 +17,7 @@
  * License along with this library; if not, see <http://www.gnu.org/licenses/>.
  */
 #include "qemu/osdep.h"
+#include "qemu/htm.h"
 #include "cpu.h"
 #include "exec/helper-proto.h"
 #include "internals.h"
@@ -31,6 +32,9 @@ static void raise_exception(CPUARMState *env, uint32_t excp,
 {
     CPUState *cs = CPU(arm_env_get_cpu(env));
 
+    if (unlikely(htm_test())) {
+        htm_end();
+    }
     assert(!excp_is_internal(excp));
     cs->exception_index = excp;
     env->exception.syndrome = syndrome;
diff --git a/target-arm/translate-a64.c b/target-arm/translate-a64.c
index 450c359..cc3baa0 100644
--- a/target-arm/translate-a64.c
+++ b/target-arm/translate-a64.c
@@ -1760,6 +1760,8 @@ static void gen_load_exclusive(DisasContext *s, int rt, int rt2,
     TCGv_i64 tmp = tcg_temp_new_i64();
     TCGMemOp be = s->be_data;
 
+    gen_helper_xbegin(cpu_env);
+
     g_assert(size <= 3);
     if (is_pair) {
         TCGv_i64 hitmp = tcg_temp_new_i64();
@@ -1825,6 +1827,10 @@ static void gen_store_exclusive(DisasContext *s, int rd, int rt, int rt2,
     tcg_gen_brcond_i64(TCG_COND_NE, addr, cpu_exclusive_addr, fail_label);
 
     tmp = tcg_temp_new_i64();
+    /* strex without a prior ldrex should just fail */
+    gen_helper_x_ok(tmp);
+    tcg_gen_brcondi_i64(TCG_COND_EQ, tmp, 0, fail_label);
+
     if (is_pair) {
         if (size == 2) {
             TCGv_i64 val = tcg_temp_new_i64();
@@ -1844,16 +1850,14 @@ static void gen_store_exclusive(DisasContext *s, int rd, int rt, int rt2,
         }
     } else {
         TCGv_i64 val = cpu_reg(s, rt);
-        tcg_gen_atomic_cmpxchg_i64(tmp, addr, cpu_exclusive_val, val,
-                                   get_mem_index(s),
-                                   size | MO_ALIGN | s->be_data);
-        tcg_gen_setcond_i64(TCG_COND_NE, tmp, tmp, cpu_exclusive_val);
+        tcg_gen_qemu_st_i64(val, addr, get_mem_index(s), s->be_data + size);
     }
 
     tcg_temp_free_i64(addr);
-
-    tcg_gen_mov_i64(cpu_reg(s, rd), tmp);
     tcg_temp_free_i64(tmp);
+
+    tcg_gen_movi_i64(cpu_reg(s, rd), 0);
+    gen_helper_xend();
     tcg_gen_br(done_label);
 
     gen_set_label(fail_label);
-- 
2.5.0