@@ -956,6 +956,18 @@ DO_GEN_ST(8, MO_UB)
DO_GEN_ST(16, MO_TEUW)
DO_GEN_ST(32, MO_TEUL)
+/* Load/Store exclusive generators (always unsigned) */
+static inline void gen_aa32_ldex32(TCGv_i32 val, TCGv_i32 addr, int index)
+{
+ tcg_gen_qemu_ldlink_i32(val, addr, index, MO_TEUL);
+}
+
+static inline void gen_aa32_stex32(TCGv_i32 is_excl, TCGv_i32 val,
+ TCGv_i32 addr, int index)
+{
+ tcg_gen_qemu_stcond_i32(is_excl, val, addr, index, MO_TEUL);
+}
+
static inline void gen_set_pc_im(DisasContext *s, target_ulong val)
{
tcg_gen_movi_i32(cpu_R[15], val);
@@ -7420,6 +7432,26 @@ static void gen_load_exclusive(DisasContext *s, int rt, int rt2,
tcg_gen_extu_i32_i64(cpu_exclusive_addr, addr);
}
+static void gen_load_exclusive_multi(DisasContext *s, int rt, int rt2,
+ TCGv_i32 addr, int size)
+{
+ TCGv_i32 tmp = tcg_temp_new_i32();
+
+ switch (size) {
+ case 0:
+ case 1:
+ abort();
+ case 2:
+ gen_aa32_ldex32(tmp, addr, get_mem_index(s));
+ break;
+ case 3:
+ default:
+ abort();
+ }
+
+ store_reg(s, rt, tmp);
+}
+
static void gen_clrex(DisasContext *s)
{
tcg_gen_movi_i64(cpu_exclusive_addr, -1);
@@ -7518,6 +7550,63 @@ static void gen_store_exclusive(DisasContext *s, int rd, int rt, int rt2,
gen_set_label(done_label);
tcg_gen_movi_i64(cpu_exclusive_addr, -1);
}
+
+static void gen_store_exclusive_multi(DisasContext *s, int rd, int rt, int rt2,
+ TCGv_i32 addr, int size)
+{
+ TCGv_i32 tmp;
+ TCGv_i32 is_dirty;
+ TCGv_i32 success;
+ TCGLabel *done_label;
+ TCGLabel *fail_label;
+
+ fail_label = gen_new_label();
+ done_label = gen_new_label();
+
+ tmp = tcg_temp_new_i32();
+ is_dirty = tcg_temp_new_i32();
+ success = tcg_temp_new_i32();
+ switch (size) {
+ case 0:
+ case 1:
+ abort();
+ break;
+ case 2:
+ gen_aa32_stex32(is_dirty, tmp, addr, get_mem_index(s));
+ break;
+ case 3:
+ default:
+ abort();
+ }
+
+ tcg_temp_free_i32(tmp);
+
+ /* Check if the store conditional has to fail */
+ tcg_gen_movi_i32(success, 1);
+ tcg_gen_brcond_i32(TCG_COND_EQ, is_dirty, success, fail_label);
+ tcg_temp_free_i32(success);
+ tcg_temp_free_i32(is_dirty);
+
+ tmp = load_reg(s, rt);
+ switch (size) {
+ case 0:
+ case 1:
+ abort();
+ case 2:
+ gen_aa32_st32(tmp, addr, get_mem_index(s));
+ break;
+ case 3:
+ default:
+ abort();
+ }
+ tcg_temp_free_i32(tmp);
+
+ tcg_gen_movi_i32(cpu_R[rd], 0); // is_dirty = 0
+ tcg_gen_br(done_label);
+ gen_set_label(fail_label);
+ tcg_gen_movi_i32(cpu_R[rd], 1); // is_dirty = 1
+ gen_set_label(done_label);
+}
#endif
/* gen_srs:
@@ -8365,7 +8454,7 @@ static void disas_arm_insn(DisasContext *s, unsigned int insn)
} else if (insn & (1 << 20)) {
switch (op1) {
case 0: /* ldrex */
- gen_load_exclusive(s, rd, 15, addr, 2);
+ gen_load_exclusive_multi(s, rd, 15, addr, 2);
break;
case 1: /* ldrexd */
gen_load_exclusive(s, rd, rd + 1, addr, 3);
@@ -8383,7 +8472,8 @@ static void disas_arm_insn(DisasContext *s, unsigned int insn)
rm = insn & 0xf;
switch (op1) {
case 0: /* strex */
- gen_store_exclusive(s, rd, rm, 15, addr, 2);
+ gen_store_exclusive_multi(s, rd, rm, 15,
+ addr, 2);
break;
case 1: /* strexd */
gen_store_exclusive(s, rd, rm, rm + 1, addr, 3);
@@ -1069,6 +1069,11 @@ static void * const qemu_ld_helpers[16] = {
[MO_BESL] = helper_be_ldul_mmu,
};
+/* LoadLink helpers, only unsigned */
+static void * const qemu_ldex_helpers[16] = {
+ [MO_LEUL] = helper_le_ldlinkul_mmu,
+};
+
/* helper signature: helper_ret_st_mmu(CPUState *env, target_ulong addr,
* uintxx_t val, int mmu_idx, uintptr_t ra)
*/
@@ -1082,6 +1087,11 @@ static void * const qemu_st_helpers[16] = {
[MO_BEQ] = helper_be_stq_mmu,
};
+/* StoreConditional helpers, only unsigned */
+static void * const qemu_stex_helpers[16] = {
+ [MO_LEUL] = helper_le_stcondl_mmu,
+};
+
/* Helper routines for marshalling helper function arguments into
* the correct registers and stack.
* argreg is where we want to put this argument, arg is the argument itself.
@@ -1222,6 +1232,7 @@ static TCGReg tcg_out_tlb_read(TCGContext *s, TCGReg addrlo, TCGReg addrhi,
path for a load or store, so that we can later generate the correct
helper code. */
static void add_qemu_ldst_label(TCGContext *s, bool is_ld, TCGMemOp opc,
+ bool is_excl, TCGReg llsc_success,
TCGReg datalo, TCGReg datahi, TCGReg addrlo,
TCGReg addrhi, int mem_index,
tcg_insn_unit *raddr, tcg_insn_unit *label_ptr)
@@ -1229,6 +1240,8 @@ static void add_qemu_ldst_label(TCGContext *s, bool is_ld, TCGMemOp opc,
TCGLabelQemuLdst *label = new_ldst_label(s);
label->is_ld = is_ld;
+ label->is_llsc = is_excl;
+ label->llsc_success = llsc_success;
label->opc = opc;
label->datalo_reg = datalo;
label->datahi_reg = datahi;
@@ -1259,12 +1272,16 @@ static void tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
/* For armv6 we can use the canonical unsigned helpers and minimize
icache usage. For pre-armv6, use the signed helpers since we do
not have a single insn sign-extend. */
- if (use_armv6_instructions) {
- func = qemu_ld_helpers[opc & ~MO_SIGN];
+ if (lb->is_llsc) {
+ func = qemu_ldex_helpers[opc];
} else {
- func = qemu_ld_helpers[opc];
- if (opc & MO_SIGN) {
- opc = MO_UL;
+ if (use_armv6_instructions) {
+ func = qemu_ld_helpers[opc & ~MO_SIGN];
+ } else {
+ func = qemu_ld_helpers[opc];
+ if (opc & MO_SIGN) {
+ opc = MO_UL;
+ }
}
}
tcg_out_call(s, func);
@@ -1336,7 +1353,14 @@ static void tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
argreg = tcg_out_arg_reg32(s, argreg, TCG_REG_R14);
/* Tail-call to the helper, which will return to the fast path. */
- tcg_out_goto(s, COND_AL, qemu_st_helpers[opc]);
+ if (lb->is_llsc) {
+ tcg_out_call(s, qemu_stex_helpers[opc]);
+ /* Save the output of the StoreConditional */
+ tcg_out_mov_reg(s, COND_AL, lb->llsc_success, TCG_REG_R0);
+ tcg_out_goto(s, COND_AL, lb->raddr);
+ } else {
+ tcg_out_goto(s, COND_AL, qemu_st_helpers[opc]);
+ }
}
#endif /* SOFTMMU */
@@ -1460,7 +1484,8 @@ static inline void tcg_out_qemu_ld_direct(TCGContext *s, TCGMemOp opc,
}
}
-static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is64)
+static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is64,
+ bool isLoadLink)
{
TCGReg addrlo, datalo, datahi, addrhi __attribute__((unused));
TCGMemOp opc;
@@ -1478,17 +1503,23 @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is64)
#ifdef CONFIG_SOFTMMU
mem_index = *args;
- addend = tcg_out_tlb_read(s, addrlo, addrhi, opc & MO_SIZE, mem_index, 1);
/* This a conditional BL only to load a pointer within this opcode into LR
- for the slow path. We will not be using the value for a tail call. */
- label_ptr = s->code_ptr;
- tcg_out_bl_noaddr(s, COND_NE);
-
- tcg_out_qemu_ld_index(s, opc, datalo, datahi, addrlo, addend);
+ for the slow path. We will not be using the value for a tail call.
+ In the context of a LoadLink instruction, we don't check the TLB but we
+ always follow the slow path. */
+ if (isLoadLink) {
+ label_ptr = s->code_ptr;
+ tcg_out_bl_noaddr(s, COND_AL);
+ } else {
+ addend = tcg_out_tlb_read(s, addrlo, addrhi, opc & MO_SIZE, mem_index, 1);
+ label_ptr = s->code_ptr;
+ tcg_out_bl_noaddr(s, COND_NE);
+ tcg_out_qemu_ld_index(s, opc, datalo, datahi, addrlo, addend);
+ }
- add_qemu_ldst_label(s, true, opc, datalo, datahi, addrlo, addrhi,
- mem_index, s->code_ptr, label_ptr);
+ add_qemu_ldst_label(s, true, opc, isLoadLink, 0, datalo, datahi, addrlo,
+ addrhi, mem_index, s->code_ptr, label_ptr);
#else /* !CONFIG_SOFTMMU */
if (GUEST_BASE) {
tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_TMP, GUEST_BASE);
@@ -1589,9 +1620,11 @@ static inline void tcg_out_qemu_st_direct(TCGContext *s, TCGMemOp opc,
}
}
-static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is64)
+static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is64,
+ bool isStoreCond)
{
TCGReg addrlo, datalo, datahi, addrhi __attribute__((unused));
+ TCGReg llsc_success;
TCGMemOp opc;
#ifdef CONFIG_SOFTMMU
int mem_index;
@@ -1599,6 +1632,8 @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is64)
tcg_insn_unit *label_ptr;
#endif
+ llsc_success = (isStoreCond ? *args++ : 0);
+
datalo = *args++;
datahi = (is64 ? *args++ : 0);
addrlo = *args++;
@@ -1607,16 +1642,24 @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is64)
#ifdef CONFIG_SOFTMMU
mem_index = *args;
- addend = tcg_out_tlb_read(s, addrlo, addrhi, opc & MO_SIZE, mem_index, 0);
- tcg_out_qemu_st_index(s, COND_EQ, opc, datalo, datahi, addrlo, addend);
+ if (isStoreCond) {
+ /* Always follow the slow-path for an exclusive access */
+ label_ptr = s->code_ptr;
+ tcg_out_bl_noaddr(s, COND_AL);
+ } else {
+ addend = tcg_out_tlb_read(s, addrlo, addrhi, opc & MO_SIZE, mem_index, 0);
- /* The conditional call must come last, as we're going to return here. */
- label_ptr = s->code_ptr;
- tcg_out_bl_noaddr(s, COND_NE);
+ tcg_out_qemu_st_index(s, COND_EQ, opc, datalo, datahi, addrlo, addend);
- add_qemu_ldst_label(s, false, opc, datalo, datahi, addrlo, addrhi,
- mem_index, s->code_ptr, label_ptr);
+ /* The conditional call must come last, as we're going to return here. */
+ label_ptr = s->code_ptr;
+ tcg_out_bl_noaddr(s, COND_NE);
+ }
+
+ add_qemu_ldst_label(s, false, opc, isStoreCond, llsc_success, datalo,
+ datahi, addrlo, addrhi, mem_index, s->code_ptr,
+ label_ptr);
#else /* !CONFIG_SOFTMMU */
if (GUEST_BASE) {
tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_TMP, GUEST_BASE);
@@ -1859,16 +1902,22 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc,
break;
case INDEX_op_qemu_ld_i32:
- tcg_out_qemu_ld(s, args, 0);
+ tcg_out_qemu_ld(s, args, 0, 0);
+ break;
+ case INDEX_op_qemu_ldlink_i32:
+ tcg_out_qemu_ld(s, args, 0, 1); // LoadLink
break;
case INDEX_op_qemu_ld_i64:
- tcg_out_qemu_ld(s, args, 1);
+ tcg_out_qemu_ld(s, args, 1, 0);
break;
case INDEX_op_qemu_st_i32:
- tcg_out_qemu_st(s, args, 0);
+ tcg_out_qemu_st(s, args, 0, 0);
+ break;
+ case INDEX_op_qemu_stcond_i32:
+ tcg_out_qemu_st(s, args, 0, 1); // StoreConditional
break;
case INDEX_op_qemu_st_i64:
- tcg_out_qemu_st(s, args, 1);
+ tcg_out_qemu_st(s, args, 1, 0);
break;
case INDEX_op_bswap16_i32:
@@ -1952,8 +2001,10 @@ static const TCGTargetOpDef arm_op_defs[] = {
#if TARGET_LONG_BITS == 32
{ INDEX_op_qemu_ld_i32, { "r", "l" } },
+ { INDEX_op_qemu_ldlink_i32, { "r", "l" } },
{ INDEX_op_qemu_ld_i64, { "r", "r", "l" } },
{ INDEX_op_qemu_st_i32, { "s", "s" } },
+ { INDEX_op_qemu_stcond_i32, { "r", "s", "s" } },
{ INDEX_op_qemu_st_i64, { "s", "s", "s" } },
#else
{ INDEX_op_qemu_ld_i32, { "r", "l", "l" } },
Implement strex and ldrex instruction relying on TCG's qemu_ldlink and qemu_stcond. For the time being only the 32bit instructions are supported. Suggested-by: Jani Kokkonen <jani.kokkonen@huawei.com> Suggested-by: Claudio Fontana <claudio.fontana@huawei.com> Signed-off-by: Alvise Rigo <a.rigo@virtualopensystems.com> --- target-arm/translate.c | 94 ++++++++++++++++++++++++++++++++++++++++++- tcg/arm/tcg-target.c | 105 ++++++++++++++++++++++++++++++++++++------------- 2 files changed, 170 insertions(+), 29 deletions(-)