[qemu-s390x] [PATCH v3 for-2.12 07/14] s390x/tcg: implement Interlocked-Access Facility 2

[David Hildenbrand]

With this facility, OI/OIY, NI/NIY and XI/XIY are atomic. All operate on
one byte (MO_UB). Emulate old behavior.

Signed-off-by: David Hildenbrand <address@hidden>
---
 target/s390x/cpu_models.c  |  1 +
 target/s390x/insn-data.def | 12 ++++-----
 target/s390x/translate.c   | 63 ++++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 70 insertions(+), 6 deletions(-)

diff --git a/target/s390x/cpu_models.c b/target/s390x/cpu_models.c
index c4c37b3b15..94d24e423d 100644
--- a/target/s390x/cpu_models.c
+++ b/target/s390x/cpu_models.c
@@ -842,6 +842,7 @@ static void add_qemu_cpu_model_features(S390FeatBitmap fbm)
         S390_FEAT_STFLE_45,
         S390_FEAT_STFLE_49,
         S390_FEAT_LOCAL_TLB_CLEARING,
+        S390_FEAT_INTERLOCKED_ACCESS_2,
         S390_FEAT_STFLE_53,
         S390_FEAT_MSA_EXT_5,
         S390_FEAT_MSA_EXT_3,
diff --git a/target/s390x/insn-data.def b/target/s390x/insn-data.def
index 166ee7c80b..6cbd604814 100644
--- a/target/s390x/insn-data.def
+++ b/target/s390x/insn-data.def
@@ -99,8 +99,8 @@
     D(0xa505, NIHL,    RI_a,  Z,   r1_o, i2_16u, r1, 0, andi, 0, 0x1020)
     D(0xa506, NILH,    RI_a,  Z,   r1_o, i2_16u, r1, 0, andi, 0, 0x1010)
     D(0xa507, NILL,    RI_a,  Z,   r1_o, i2_16u, r1, 0, andi, 0, 0x1000)
-    C(0x9400, NI,      SI,    Z,   m1_8u, i2_8u, new, m1_8, and, nz64)
-    C(0xeb54, NIY,     SIY,   LD,  m1_8u, i2_8u, new, m1_8, and, nz64)
+    D(0x9400, NI,      SI,    Z,   la1, i2_8u, new, 0, ni, nz64, MO_UB)
+    D(0xeb54, NIY,     SIY,   LD,  la1, i2_8u, new, 0, ni, nz64, MO_UB)
 
 /* BRANCH AND SAVE */
     C(0x0d00, BASR,    RR_a,  Z,   0, r2_nz, r1, 0, bas, 0)
@@ -357,8 +357,8 @@
 /* EXCLUSIVE OR IMMEDIATE */
     D(0xc006, XIHF,    RIL_a, EI,  r1_o, i2_32u, r1, 0, xori, 0, 0x2020)
     D(0xc007, XILF,    RIL_a, EI,  r1_o, i2_32u, r1, 0, xori, 0, 0x2000)
-    C(0x9700, XI,      SI,    Z,   m1_8u, i2_8u, new, m1_8, xor, nz64)
-    C(0xeb57, XIY,     SIY,   LD,  m1_8u, i2_8u, new, m1_8, xor, nz64)
+    D(0x9700, XI,      SI,    Z,   la1, i2_8u, new, 0, xi, nz64, MO_UB)
+    D(0xeb57, XIY,     SIY,   LD,  la1, i2_8u, new, 0, xi, nz64, MO_UB)
 
 /* EXECUTE */
     C(0x4400, EX,      RX_a,  Z,   0, a2, 0, 0, ex, 0)
@@ -698,8 +698,8 @@
     D(0xa509, OIHL,    RI_a,  Z,   r1_o, i2_16u, r1, 0, ori, 0, 0x1020)
     D(0xa50a, OILH,    RI_a,  Z,   r1_o, i2_16u, r1, 0, ori, 0, 0x1010)
     D(0xa50b, OILL,    RI_a,  Z,   r1_o, i2_16u, r1, 0, ori, 0, 0x1000)
-    C(0x9600, OI,      SI,    Z,   m1_8u, i2_8u, new, m1_8, or, nz64)
-    C(0xeb56, OIY,     SIY,   LD,  m1_8u, i2_8u, new, m1_8, or, nz64)
+    D(0x9600, OI,      SI,    Z,   la1, i2_8u, new, 0, oi, nz64, MO_UB)
+    D(0xeb56, OIY,     SIY,   LD,  la1, i2_8u, new, 0, oi, nz64, MO_UB)
 
 /* PACK */
     /* Really format SS_b, but we pack both lengths into one argument
diff --git a/target/s390x/translate.c b/target/s390x/translate.c
index 7ab8e853ab..de00b9471a 100644
--- a/target/s390x/translate.c
+++ b/target/s390x/translate.c
@@ -1427,6 +1427,27 @@ static ExitStatus op_andi(DisasContext *s, DisasOps *o)
     return NO_EXIT;
 }
 
+static ExitStatus op_ni(DisasContext *s, DisasOps *o)
+{
+    o->in1 = tcg_temp_new_i64();
+
+    if (!s390_has_feat(S390_FEAT_INTERLOCKED_ACCESS_2)) {
+        tcg_gen_qemu_ld_tl(o->in1, o->addr1, get_mem_index(s), s->insn->data);
+    } else {
+        /* Perform the atomic operation in memory. */
+        tcg_gen_atomic_fetch_and_i64(o->in1, o->addr1, o->in2, 
get_mem_index(s),
+                                     s->insn->data);
+    }
+
+    /* Recompute also for atomic case: needed for setting CC. */
+    tcg_gen_and_i64(o->out, o->in1, o->in2);
+
+    if (!s390_has_feat(S390_FEAT_INTERLOCKED_ACCESS_2)) {
+        tcg_gen_qemu_st_tl(o->out, o->addr1, get_mem_index(s), s->insn->data);
+    }
+    return NO_EXIT;
+}
+
 static ExitStatus op_bas(DisasContext *s, DisasOps *o)
 {
     tcg_gen_movi_i64(o->out, pc_to_link_info(s, s->next_pc));
@@ -3378,6 +3399,27 @@ static ExitStatus op_ori(DisasContext *s, DisasOps *o)
     return NO_EXIT;
 }
 
+static ExitStatus op_oi(DisasContext *s, DisasOps *o)
+{
+    o->in1 = tcg_temp_new_i64();
+
+    if (!s390_has_feat(S390_FEAT_INTERLOCKED_ACCESS_2)) {
+        tcg_gen_qemu_ld_tl(o->in1, o->addr1, get_mem_index(s), s->insn->data);
+    } else {
+        /* Perform the atomic operation in memory. */
+        tcg_gen_atomic_fetch_or_i64(o->in1, o->addr1, o->in2, get_mem_index(s),
+                                    s->insn->data);
+    }
+
+    /* Recompute also for atomic case: needed for setting CC. */
+    tcg_gen_or_i64(o->out, o->in1, o->in2);
+
+    if (!s390_has_feat(S390_FEAT_INTERLOCKED_ACCESS_2)) {
+        tcg_gen_qemu_st_tl(o->out, o->addr1, get_mem_index(s), s->insn->data);
+    }
+    return NO_EXIT;
+}
+
 static ExitStatus op_pack(DisasContext *s, DisasOps *o)
 {
     TCGv_i32 l = tcg_const_i32(get_field(s->fields, l1));
@@ -4643,6 +4685,27 @@ static ExitStatus op_xori(DisasContext *s, DisasOps *o)
     return NO_EXIT;
 }
 
+static ExitStatus op_xi(DisasContext *s, DisasOps *o)
+{
+    o->in1 = tcg_temp_new_i64();
+
+    if (!s390_has_feat(S390_FEAT_INTERLOCKED_ACCESS_2)) {
+        tcg_gen_qemu_ld_tl(o->in1, o->addr1, get_mem_index(s), s->insn->data);
+    } else {
+        /* Perform the atomic operation in memory. */
+        tcg_gen_atomic_fetch_xor_i64(o->in1, o->addr1, o->in2, 
get_mem_index(s),
+                                     s->insn->data);
+    }
+
+    /* Recompute also for atomic case: needed for setting CC. */
+    tcg_gen_xor_i64(o->out, o->in1, o->in2);
+
+    if (!s390_has_feat(S390_FEAT_INTERLOCKED_ACCESS_2)) {
+        tcg_gen_qemu_st_tl(o->out, o->addr1, get_mem_index(s), s->insn->data);
+    }
+    return NO_EXIT;
+}
+
 static ExitStatus op_zero(DisasContext *s, DisasOps *o)
 {
     o->out = tcg_const_i64(0);
-- 
2.14.3