* [PATCH v8 0/9] target/arm: pc-relative translation blocks
@ 2022-10-20 3:06 Richard Henderson
2022-10-20 3:06 ` [PATCH v8 1/9] target/arm: Introduce curr_insn_len Richard Henderson
` (9 more replies)
0 siblings, 10 replies; 11+ messages in thread
From: Richard Henderson @ 2022-10-20 3:06 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm
This is the Arm specific changes required to reduce the
amount of translation for address space randomization.
Changes for v8:
* Add a comment to both gen_goto_tb concerning the ordering
of the pc update vs the goto_tb opcode.
r~
Richard Henderson (9):
target/arm: Introduce curr_insn_len
target/arm: Change gen_goto_tb to work on displacements
target/arm: Change gen_*set_pc_im to gen_*update_pc
target/arm: Change gen_exception_insn* to work on displacements
target/arm: Remove gen_exception_internal_insn pc argument
target/arm: Change gen_jmp* to work on displacements
target/arm: Introduce gen_pc_plus_diff for aarch64
target/arm: Introduce gen_pc_plus_diff for aarch32
target/arm: Enable TARGET_TB_PCREL
target/arm/cpu-param.h | 1 +
target/arm/translate-a32.h | 2 +-
target/arm/translate.h | 66 +++++++-
target/arm/cpu.c | 23 +--
target/arm/translate-a64.c | 199 ++++++++++++++----------
target/arm/translate-m-nocp.c | 8 +-
target/arm/translate-mve.c | 2 +-
target/arm/translate-vfp.c | 10 +-
target/arm/translate.c | 284 ++++++++++++++++++++--------------
9 files changed, 367 insertions(+), 228 deletions(-)
--
2.34.1
^ permalink raw reply [flat|nested] 11+ messages in thread
* [PATCH v8 1/9] target/arm: Introduce curr_insn_len
2022-10-20 3:06 [PATCH v8 0/9] target/arm: pc-relative translation blocks Richard Henderson
@ 2022-10-20 3:06 ` Richard Henderson
2022-10-20 3:06 ` [PATCH v8 2/9] target/arm: Change gen_goto_tb to work on displacements Richard Henderson
` (8 subsequent siblings)
9 siblings, 0 replies; 11+ messages in thread
From: Richard Henderson @ 2022-10-20 3:06 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Philippe Mathieu-Daudé
A simple helper to retrieve the length of the current insn.
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/translate.h | 5 +++++
target/arm/translate-vfp.c | 2 +-
target/arm/translate.c | 5 ++---
3 files changed, 8 insertions(+), 4 deletions(-)
diff --git a/target/arm/translate.h b/target/arm/translate.h
index af5d4a7086..90bf7c57fc 100644
--- a/target/arm/translate.h
+++ b/target/arm/translate.h
@@ -226,6 +226,11 @@ static inline void disas_set_insn_syndrome(DisasContext *s, uint32_t syn)
s->insn_start = NULL;
}
+static inline int curr_insn_len(DisasContext *s)
+{
+ return s->base.pc_next - s->pc_curr;
+}
+
/* is_jmp field values */
#define DISAS_JUMP DISAS_TARGET_0 /* only pc was modified dynamically */
/* CPU state was modified dynamically; exit to main loop for interrupts. */
diff --git a/target/arm/translate-vfp.c b/target/arm/translate-vfp.c
index bd5ae27d09..94cc1e4b77 100644
--- a/target/arm/translate-vfp.c
+++ b/target/arm/translate-vfp.c
@@ -242,7 +242,7 @@ static bool vfp_access_check_a(DisasContext *s, bool ignore_vfp_enabled)
if (s->sme_trap_nonstreaming) {
gen_exception_insn(s, s->pc_curr, EXCP_UDEF,
syn_smetrap(SME_ET_Streaming,
- s->base.pc_next - s->pc_curr == 2));
+ curr_insn_len(s) == 2));
return false;
}
diff --git a/target/arm/translate.c b/target/arm/translate.c
index 2f72afe019..5752b7af5c 100644
--- a/target/arm/translate.c
+++ b/target/arm/translate.c
@@ -6650,7 +6650,7 @@ static ISSInfo make_issinfo(DisasContext *s, int rd, bool p, bool w)
/* ISS not valid if writeback */
if (p && !w) {
ret = rd;
- if (s->base.pc_next - s->pc_curr == 2) {
+ if (curr_insn_len(s) == 2) {
ret |= ISSIs16Bit;
}
} else {
@@ -9812,8 +9812,7 @@ static void arm_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu)
/* nothing more to generate */
break;
case DISAS_WFI:
- gen_helper_wfi(cpu_env,
- tcg_constant_i32(dc->base.pc_next - dc->pc_curr));
+ gen_helper_wfi(cpu_env, tcg_constant_i32(curr_insn_len(dc)));
/*
* The helper doesn't necessarily throw an exception, but we
* must go back to the main loop to check for interrupts anyway.
--
2.34.1
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PATCH v8 2/9] target/arm: Change gen_goto_tb to work on displacements
2022-10-20 3:06 [PATCH v8 0/9] target/arm: pc-relative translation blocks Richard Henderson
2022-10-20 3:06 ` [PATCH v8 1/9] target/arm: Introduce curr_insn_len Richard Henderson
@ 2022-10-20 3:06 ` Richard Henderson
2022-10-20 3:06 ` [PATCH v8 3/9] target/arm: Change gen_*set_pc_im to gen_*update_pc Richard Henderson
` (7 subsequent siblings)
9 siblings, 0 replies; 11+ messages in thread
From: Richard Henderson @ 2022-10-20 3:06 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Peter Maydell
In preparation for TARGET_TB_PCREL, reduce reliance on absolute values.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/translate-a64.c | 40 ++++++++++++++++++++------------------
target/arm/translate.c | 10 ++++++----
2 files changed, 27 insertions(+), 23 deletions(-)
diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c
index 5b67375f4e..6a372ed184 100644
--- a/target/arm/translate-a64.c
+++ b/target/arm/translate-a64.c
@@ -370,8 +370,10 @@ static inline bool use_goto_tb(DisasContext *s, uint64_t dest)
return translator_use_goto_tb(&s->base, dest);
}
-static inline void gen_goto_tb(DisasContext *s, int n, uint64_t dest)
+static void gen_goto_tb(DisasContext *s, int n, int64_t diff)
{
+ uint64_t dest = s->pc_curr + diff;
+
if (use_goto_tb(s, dest)) {
tcg_gen_goto_tb(n);
gen_a64_set_pc_im(dest);
@@ -1354,7 +1356,7 @@ static inline AArch64DecodeFn *lookup_disas_fn(const AArch64DecodeTable *table,
*/
static void disas_uncond_b_imm(DisasContext *s, uint32_t insn)
{
- uint64_t addr = s->pc_curr + sextract32(insn, 0, 26) * 4;
+ int64_t diff = sextract32(insn, 0, 26) * 4;
if (insn & (1U << 31)) {
/* BL Branch with link */
@@ -1363,7 +1365,7 @@ static void disas_uncond_b_imm(DisasContext *s, uint32_t insn)
/* B Branch / BL Branch with link */
reset_btype(s);
- gen_goto_tb(s, 0, addr);
+ gen_goto_tb(s, 0, diff);
}
/* Compare and branch (immediate)
@@ -1375,14 +1377,14 @@ static void disas_uncond_b_imm(DisasContext *s, uint32_t insn)
static void disas_comp_b_imm(DisasContext *s, uint32_t insn)
{
unsigned int sf, op, rt;
- uint64_t addr;
+ int64_t diff;
TCGLabel *label_match;
TCGv_i64 tcg_cmp;
sf = extract32(insn, 31, 1);
op = extract32(insn, 24, 1); /* 0: CBZ; 1: CBNZ */
rt = extract32(insn, 0, 5);
- addr = s->pc_curr + sextract32(insn, 5, 19) * 4;
+ diff = sextract32(insn, 5, 19) * 4;
tcg_cmp = read_cpu_reg(s, rt, sf);
label_match = gen_new_label();
@@ -1391,9 +1393,9 @@ static void disas_comp_b_imm(DisasContext *s, uint32_t insn)
tcg_gen_brcondi_i64(op ? TCG_COND_NE : TCG_COND_EQ,
tcg_cmp, 0, label_match);
- gen_goto_tb(s, 0, s->base.pc_next);
+ gen_goto_tb(s, 0, 4);
gen_set_label(label_match);
- gen_goto_tb(s, 1, addr);
+ gen_goto_tb(s, 1, diff);
}
/* Test and branch (immediate)
@@ -1405,13 +1407,13 @@ static void disas_comp_b_imm(DisasContext *s, uint32_t insn)
static void disas_test_b_imm(DisasContext *s, uint32_t insn)
{
unsigned int bit_pos, op, rt;
- uint64_t addr;
+ int64_t diff;
TCGLabel *label_match;
TCGv_i64 tcg_cmp;
bit_pos = (extract32(insn, 31, 1) << 5) | extract32(insn, 19, 5);
op = extract32(insn, 24, 1); /* 0: TBZ; 1: TBNZ */
- addr = s->pc_curr + sextract32(insn, 5, 14) * 4;
+ diff = sextract32(insn, 5, 14) * 4;
rt = extract32(insn, 0, 5);
tcg_cmp = tcg_temp_new_i64();
@@ -1422,9 +1424,9 @@ static void disas_test_b_imm(DisasContext *s, uint32_t insn)
tcg_gen_brcondi_i64(op ? TCG_COND_NE : TCG_COND_EQ,
tcg_cmp, 0, label_match);
tcg_temp_free_i64(tcg_cmp);
- gen_goto_tb(s, 0, s->base.pc_next);
+ gen_goto_tb(s, 0, 4);
gen_set_label(label_match);
- gen_goto_tb(s, 1, addr);
+ gen_goto_tb(s, 1, diff);
}
/* Conditional branch (immediate)
@@ -1436,13 +1438,13 @@ static void disas_test_b_imm(DisasContext *s, uint32_t insn)
static void disas_cond_b_imm(DisasContext *s, uint32_t insn)
{
unsigned int cond;
- uint64_t addr;
+ int64_t diff;
if ((insn & (1 << 4)) || (insn & (1 << 24))) {
unallocated_encoding(s);
return;
}
- addr = s->pc_curr + sextract32(insn, 5, 19) * 4;
+ diff = sextract32(insn, 5, 19) * 4;
cond = extract32(insn, 0, 4);
reset_btype(s);
@@ -1450,12 +1452,12 @@ static void disas_cond_b_imm(DisasContext *s, uint32_t insn)
/* genuinely conditional branches */
TCGLabel *label_match = gen_new_label();
arm_gen_test_cc(cond, label_match);
- gen_goto_tb(s, 0, s->base.pc_next);
+ gen_goto_tb(s, 0, 4);
gen_set_label(label_match);
- gen_goto_tb(s, 1, addr);
+ gen_goto_tb(s, 1, diff);
} else {
/* 0xe and 0xf are both "always" conditions */
- gen_goto_tb(s, 0, addr);
+ gen_goto_tb(s, 0, diff);
}
}
@@ -1629,7 +1631,7 @@ static void handle_sync(DisasContext *s, uint32_t insn,
* any pending interrupts immediately.
*/
reset_btype(s);
- gen_goto_tb(s, 0, s->base.pc_next);
+ gen_goto_tb(s, 0, 4);
return;
case 7: /* SB */
@@ -1641,7 +1643,7 @@ static void handle_sync(DisasContext *s, uint32_t insn,
* MB and end the TB instead.
*/
tcg_gen_mb(TCG_MO_ALL | TCG_BAR_SC);
- gen_goto_tb(s, 0, s->base.pc_next);
+ gen_goto_tb(s, 0, 4);
return;
default:
@@ -14947,7 +14949,7 @@ static void aarch64_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu)
switch (dc->base.is_jmp) {
case DISAS_NEXT:
case DISAS_TOO_MANY:
- gen_goto_tb(dc, 1, dc->base.pc_next);
+ gen_goto_tb(dc, 1, 4);
break;
default:
case DISAS_UPDATE_EXIT:
diff --git a/target/arm/translate.c b/target/arm/translate.c
index 5752b7af5c..ae30c26ca4 100644
--- a/target/arm/translate.c
+++ b/target/arm/translate.c
@@ -2590,8 +2590,10 @@ static void gen_goto_ptr(void)
* cpu_loop_exec. Any live exit_requests will be processed as we
* enter the next TB.
*/
-static void gen_goto_tb(DisasContext *s, int n, target_ulong dest)
+static void gen_goto_tb(DisasContext *s, int n, int diff)
{
+ target_ulong dest = s->pc_curr + diff;
+
if (translator_use_goto_tb(&s->base, dest)) {
tcg_gen_goto_tb(n);
gen_set_pc_im(s, dest);
@@ -2625,7 +2627,7 @@ static inline void gen_jmp_tb(DisasContext *s, uint32_t dest, int tbno)
* gen_jmp();
* on the second call to gen_jmp().
*/
- gen_goto_tb(s, tbno, dest);
+ gen_goto_tb(s, tbno, dest - s->pc_curr);
break;
case DISAS_UPDATE_NOCHAIN:
case DISAS_UPDATE_EXIT:
@@ -9793,7 +9795,7 @@ static void arm_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu)
switch (dc->base.is_jmp) {
case DISAS_NEXT:
case DISAS_TOO_MANY:
- gen_goto_tb(dc, 1, dc->base.pc_next);
+ gen_goto_tb(dc, 1, curr_insn_len(dc));
break;
case DISAS_UPDATE_NOCHAIN:
gen_set_pc_im(dc, dc->base.pc_next);
@@ -9845,7 +9847,7 @@ static void arm_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu)
gen_set_pc_im(dc, dc->base.pc_next);
gen_singlestep_exception(dc);
} else {
- gen_goto_tb(dc, 1, dc->base.pc_next);
+ gen_goto_tb(dc, 1, curr_insn_len(dc));
}
}
}
--
2.34.1
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PATCH v8 3/9] target/arm: Change gen_*set_pc_im to gen_*update_pc
2022-10-20 3:06 [PATCH v8 0/9] target/arm: pc-relative translation blocks Richard Henderson
2022-10-20 3:06 ` [PATCH v8 1/9] target/arm: Introduce curr_insn_len Richard Henderson
2022-10-20 3:06 ` [PATCH v8 2/9] target/arm: Change gen_goto_tb to work on displacements Richard Henderson
@ 2022-10-20 3:06 ` Richard Henderson
2022-10-20 3:06 ` [PATCH v8 4/9] target/arm: Change gen_exception_insn* to work on displacements Richard Henderson
` (6 subsequent siblings)
9 siblings, 0 replies; 11+ messages in thread
From: Richard Henderson @ 2022-10-20 3:06 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Philippe Mathieu-Daudé
In preparation for TARGET_TB_PCREL, reduce reliance on
absolute values by passing in pc difference.
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/translate-a32.h | 2 +-
target/arm/translate.h | 6 ++--
target/arm/translate-a64.c | 32 +++++++++---------
target/arm/translate-vfp.c | 2 +-
target/arm/translate.c | 68 ++++++++++++++++++++------------------
5 files changed, 56 insertions(+), 54 deletions(-)
diff --git a/target/arm/translate-a32.h b/target/arm/translate-a32.h
index 78a84c1414..5339c22f1e 100644
--- a/target/arm/translate-a32.h
+++ b/target/arm/translate-a32.h
@@ -40,7 +40,7 @@ void write_neon_element64(TCGv_i64 src, int reg, int ele, MemOp memop);
TCGv_i32 add_reg_for_lit(DisasContext *s, int reg, int ofs);
void gen_set_cpsr(TCGv_i32 var, uint32_t mask);
void gen_set_condexec(DisasContext *s);
-void gen_set_pc_im(DisasContext *s, target_ulong val);
+void gen_update_pc(DisasContext *s, target_long diff);
void gen_lookup_tb(DisasContext *s);
long vfp_reg_offset(bool dp, unsigned reg);
long neon_full_reg_offset(unsigned reg);
diff --git a/target/arm/translate.h b/target/arm/translate.h
index 90bf7c57fc..d651044855 100644
--- a/target/arm/translate.h
+++ b/target/arm/translate.h
@@ -254,7 +254,7 @@ static inline int curr_insn_len(DisasContext *s)
* For instructions which want an immediate exit to the main loop, as opposed
* to attempting to use lookup_and_goto_ptr. Unlike DISAS_UPDATE_EXIT, this
* doesn't write the PC on exiting the translation loop so you need to ensure
- * something (gen_a64_set_pc_im or runtime helper) has done so before we reach
+ * something (gen_a64_update_pc or runtime helper) has done so before we reach
* return from cpu_tb_exec.
*/
#define DISAS_EXIT DISAS_TARGET_9
@@ -263,14 +263,14 @@ static inline int curr_insn_len(DisasContext *s)
#ifdef TARGET_AARCH64
void a64_translate_init(void);
-void gen_a64_set_pc_im(uint64_t val);
+void gen_a64_update_pc(DisasContext *s, target_long diff);
extern const TranslatorOps aarch64_translator_ops;
#else
static inline void a64_translate_init(void)
{
}
-static inline void gen_a64_set_pc_im(uint64_t val)
+static inline void gen_a64_update_pc(DisasContext *s, target_long diff)
{
}
#endif
diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c
index 6a372ed184..585d42d5b2 100644
--- a/target/arm/translate-a64.c
+++ b/target/arm/translate-a64.c
@@ -140,9 +140,9 @@ static void reset_btype(DisasContext *s)
}
}
-void gen_a64_set_pc_im(uint64_t val)
+void gen_a64_update_pc(DisasContext *s, target_long diff)
{
- tcg_gen_movi_i64(cpu_pc, val);
+ tcg_gen_movi_i64(cpu_pc, s->pc_curr + diff);
}
/*
@@ -334,14 +334,14 @@ static void gen_exception_internal(int excp)
static void gen_exception_internal_insn(DisasContext *s, uint64_t pc, int excp)
{
- gen_a64_set_pc_im(pc);
+ gen_a64_update_pc(s, pc - s->pc_curr);
gen_exception_internal(excp);
s->base.is_jmp = DISAS_NORETURN;
}
static void gen_exception_bkpt_insn(DisasContext *s, uint32_t syndrome)
{
- gen_a64_set_pc_im(s->pc_curr);
+ gen_a64_update_pc(s, 0);
gen_helper_exception_bkpt_insn(cpu_env, tcg_constant_i32(syndrome));
s->base.is_jmp = DISAS_NORETURN;
}
@@ -376,11 +376,11 @@ static void gen_goto_tb(DisasContext *s, int n, int64_t diff)
if (use_goto_tb(s, dest)) {
tcg_gen_goto_tb(n);
- gen_a64_set_pc_im(dest);
+ gen_a64_update_pc(s, diff);
tcg_gen_exit_tb(s->base.tb, n);
s->base.is_jmp = DISAS_NORETURN;
} else {
- gen_a64_set_pc_im(dest);
+ gen_a64_update_pc(s, diff);
if (s->ss_active) {
gen_step_complete_exception(s);
} else {
@@ -1952,7 +1952,7 @@ static void handle_sys(DisasContext *s, uint32_t insn, bool isread,
uint32_t syndrome;
syndrome = syn_aa64_sysregtrap(op0, op1, op2, crn, crm, rt, isread);
- gen_a64_set_pc_im(s->pc_curr);
+ gen_a64_update_pc(s, 0);
gen_helper_access_check_cp_reg(cpu_env,
tcg_constant_ptr(ri),
tcg_constant_i32(syndrome),
@@ -1962,7 +1962,7 @@ static void handle_sys(DisasContext *s, uint32_t insn, bool isread,
* The readfn or writefn might raise an exception;
* synchronize the CPU state in case it does.
*/
- gen_a64_set_pc_im(s->pc_curr);
+ gen_a64_update_pc(s, 0);
}
/* Handle special cases first */
@@ -2172,7 +2172,7 @@ static void disas_exc(DisasContext *s, uint32_t insn)
/* The pre HVC helper handles cases when HVC gets trapped
* as an undefined insn by runtime configuration.
*/
- gen_a64_set_pc_im(s->pc_curr);
+ gen_a64_update_pc(s, 0);
gen_helper_pre_hvc(cpu_env);
gen_ss_advance(s);
gen_exception_insn_el(s, s->base.pc_next, EXCP_HVC,
@@ -2183,7 +2183,7 @@ static void disas_exc(DisasContext *s, uint32_t insn)
unallocated_encoding(s);
break;
}
- gen_a64_set_pc_im(s->pc_curr);
+ gen_a64_update_pc(s, 0);
gen_helper_pre_smc(cpu_env, tcg_constant_i32(syn_aa64_smc(imm16)));
gen_ss_advance(s);
gen_exception_insn_el(s, s->base.pc_next, EXCP_SMC,
@@ -14936,7 +14936,7 @@ static void aarch64_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu)
*/
switch (dc->base.is_jmp) {
default:
- gen_a64_set_pc_im(dc->base.pc_next);
+ gen_a64_update_pc(dc, 4);
/* fall through */
case DISAS_EXIT:
case DISAS_JUMP:
@@ -14953,13 +14953,13 @@ static void aarch64_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu)
break;
default:
case DISAS_UPDATE_EXIT:
- gen_a64_set_pc_im(dc->base.pc_next);
+ gen_a64_update_pc(dc, 4);
/* fall through */
case DISAS_EXIT:
tcg_gen_exit_tb(NULL, 0);
break;
case DISAS_UPDATE_NOCHAIN:
- gen_a64_set_pc_im(dc->base.pc_next);
+ gen_a64_update_pc(dc, 4);
/* fall through */
case DISAS_JUMP:
tcg_gen_lookup_and_goto_ptr();
@@ -14968,11 +14968,11 @@ static void aarch64_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu)
case DISAS_SWI:
break;
case DISAS_WFE:
- gen_a64_set_pc_im(dc->base.pc_next);
+ gen_a64_update_pc(dc, 4);
gen_helper_wfe(cpu_env);
break;
case DISAS_YIELD:
- gen_a64_set_pc_im(dc->base.pc_next);
+ gen_a64_update_pc(dc, 4);
gen_helper_yield(cpu_env);
break;
case DISAS_WFI:
@@ -14980,7 +14980,7 @@ static void aarch64_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu)
* This is a special case because we don't want to just halt
* the CPU if trying to debug across a WFI.
*/
- gen_a64_set_pc_im(dc->base.pc_next);
+ gen_a64_update_pc(dc, 4);
gen_helper_wfi(cpu_env, tcg_constant_i32(4));
/*
* The helper doesn't necessarily throw an exception, but we
diff --git a/target/arm/translate-vfp.c b/target/arm/translate-vfp.c
index 94cc1e4b77..070f465b17 100644
--- a/target/arm/translate-vfp.c
+++ b/target/arm/translate-vfp.c
@@ -856,7 +856,7 @@ static bool trans_VMSR_VMRS(DisasContext *s, arg_VMSR_VMRS *a)
case ARM_VFP_FPSID:
if (s->current_el == 1) {
gen_set_condexec(s);
- gen_set_pc_im(s, s->pc_curr);
+ gen_update_pc(s, 0);
gen_helper_check_hcr_el2_trap(cpu_env,
tcg_constant_i32(a->rt),
tcg_constant_i32(a->reg));
diff --git a/target/arm/translate.c b/target/arm/translate.c
index ae30c26ca4..9863a08f49 100644
--- a/target/arm/translate.c
+++ b/target/arm/translate.c
@@ -768,9 +768,9 @@ void gen_set_condexec(DisasContext *s)
}
}
-void gen_set_pc_im(DisasContext *s, target_ulong val)
+void gen_update_pc(DisasContext *s, target_long diff)
{
- tcg_gen_movi_i32(cpu_R[15], val);
+ tcg_gen_movi_i32(cpu_R[15], s->pc_curr + diff);
}
/* Set PC and Thumb state from var. var is marked as dead. */
@@ -862,7 +862,7 @@ static inline void gen_bxns(DisasContext *s, int rm)
/* The bxns helper may raise an EXCEPTION_EXIT exception, so in theory
* we need to sync state before calling it, but:
- * - we don't need to do gen_set_pc_im() because the bxns helper will
+ * - we don't need to do gen_update_pc() because the bxns helper will
* always set the PC itself
* - we don't need to do gen_set_condexec() because BXNS is UNPREDICTABLE
* unless it's outside an IT block or the last insn in an IT block,
@@ -883,7 +883,7 @@ static inline void gen_blxns(DisasContext *s, int rm)
* 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->base.pc_next);
+ gen_update_pc(s, curr_insn_len(s));
gen_helper_v7m_blxns(cpu_env, var);
tcg_temp_free_i32(var);
s->base.is_jmp = DISAS_EXIT;
@@ -1051,7 +1051,7 @@ static inline void gen_hvc(DisasContext *s, int imm16)
* as an undefined insn by runtime configuration (ie before
* the insn really executes).
*/
- gen_set_pc_im(s, s->pc_curr);
+ gen_update_pc(s, 0);
gen_helper_pre_hvc(cpu_env);
/* Otherwise we will treat this as a real exception which
* happens after execution of the insn. (The distinction matters
@@ -1059,7 +1059,7 @@ static inline void gen_hvc(DisasContext *s, int imm16)
* for single stepping.)
*/
s->svc_imm = imm16;
- gen_set_pc_im(s, s->base.pc_next);
+ gen_update_pc(s, curr_insn_len(s));
s->base.is_jmp = DISAS_HVC;
}
@@ -1068,16 +1068,16 @@ static inline void gen_smc(DisasContext *s)
/* As with HVC, we may take an exception either before or after
* the insn executes.
*/
- gen_set_pc_im(s, s->pc_curr);
+ gen_update_pc(s, 0);
gen_helper_pre_smc(cpu_env, tcg_constant_i32(syn_aa32_smc()));
- gen_set_pc_im(s, s->base.pc_next);
+ gen_update_pc(s, curr_insn_len(s));
s->base.is_jmp = DISAS_SMC;
}
static void gen_exception_internal_insn(DisasContext *s, uint32_t pc, int excp)
{
gen_set_condexec(s);
- gen_set_pc_im(s, pc);
+ gen_update_pc(s, pc - s->pc_curr);
gen_exception_internal(excp);
s->base.is_jmp = DISAS_NORETURN;
}
@@ -1103,10 +1103,10 @@ static void gen_exception_insn_el_v(DisasContext *s, uint64_t pc, int excp,
uint32_t syn, TCGv_i32 tcg_el)
{
if (s->aarch64) {
- gen_a64_set_pc_im(pc);
+ gen_a64_update_pc(s, pc - s->pc_curr);
} else {
gen_set_condexec(s);
- gen_set_pc_im(s, pc);
+ gen_update_pc(s, pc - s->pc_curr);
}
gen_exception_el_v(excp, syn, tcg_el);
s->base.is_jmp = DISAS_NORETURN;
@@ -1121,10 +1121,10 @@ void gen_exception_insn_el(DisasContext *s, uint64_t pc, int excp,
void gen_exception_insn(DisasContext *s, uint64_t pc, int excp, uint32_t syn)
{
if (s->aarch64) {
- gen_a64_set_pc_im(pc);
+ gen_a64_update_pc(s, pc - s->pc_curr);
} else {
gen_set_condexec(s);
- gen_set_pc_im(s, pc);
+ gen_update_pc(s, pc - s->pc_curr);
}
gen_exception(excp, syn);
s->base.is_jmp = DISAS_NORETURN;
@@ -1133,7 +1133,7 @@ void gen_exception_insn(DisasContext *s, uint64_t pc, int excp, uint32_t syn)
static void gen_exception_bkpt_insn(DisasContext *s, uint32_t syn)
{
gen_set_condexec(s);
- gen_set_pc_im(s, s->pc_curr);
+ gen_update_pc(s, 0);
gen_helper_exception_bkpt_insn(cpu_env, tcg_constant_i32(syn));
s->base.is_jmp = DISAS_NORETURN;
}
@@ -2596,10 +2596,10 @@ static void gen_goto_tb(DisasContext *s, int n, int diff)
if (translator_use_goto_tb(&s->base, dest)) {
tcg_gen_goto_tb(n);
- gen_set_pc_im(s, dest);
+ gen_update_pc(s, diff);
tcg_gen_exit_tb(s->base.tb, n);
} else {
- gen_set_pc_im(s, dest);
+ gen_update_pc(s, diff);
gen_goto_ptr();
}
s->base.is_jmp = DISAS_NORETURN;
@@ -2608,9 +2608,11 @@ static void gen_goto_tb(DisasContext *s, int n, int diff)
/* Jump, specifying which TB number to use if we gen_goto_tb() */
static inline void gen_jmp_tb(DisasContext *s, uint32_t dest, int tbno)
{
+ int diff = dest - s->pc_curr;
+
if (unlikely(s->ss_active)) {
/* An indirect jump so that we still trigger the debug exception. */
- gen_set_pc_im(s, dest);
+ gen_update_pc(s, diff);
s->base.is_jmp = DISAS_JUMP;
return;
}
@@ -2627,7 +2629,7 @@ static inline void gen_jmp_tb(DisasContext *s, uint32_t dest, int tbno)
* gen_jmp();
* on the second call to gen_jmp().
*/
- gen_goto_tb(s, tbno, dest - s->pc_curr);
+ gen_goto_tb(s, tbno, diff);
break;
case DISAS_UPDATE_NOCHAIN:
case DISAS_UPDATE_EXIT:
@@ -2636,7 +2638,7 @@ static inline void gen_jmp_tb(DisasContext *s, uint32_t dest, int tbno)
* Avoid using goto_tb so we really do exit back to the main loop
* and don't chain to another TB.
*/
- gen_set_pc_im(s, dest);
+ gen_update_pc(s, diff);
gen_goto_ptr();
s->base.is_jmp = DISAS_NORETURN;
break;
@@ -2904,7 +2906,7 @@ static void gen_msr_banked(DisasContext *s, int r, int sysm, int rn)
/* Sync state because msr_banked() can raise exceptions */
gen_set_condexec(s);
- gen_set_pc_im(s, s->pc_curr);
+ gen_update_pc(s, 0);
tcg_reg = load_reg(s, rn);
gen_helper_msr_banked(cpu_env, tcg_reg,
tcg_constant_i32(tgtmode),
@@ -2924,7 +2926,7 @@ static void gen_mrs_banked(DisasContext *s, int r, int sysm, int rn)
/* Sync state because mrs_banked() can raise exceptions */
gen_set_condexec(s);
- gen_set_pc_im(s, s->pc_curr);
+ gen_update_pc(s, 0);
tcg_reg = tcg_temp_new_i32();
gen_helper_mrs_banked(tcg_reg, cpu_env,
tcg_constant_i32(tgtmode),
@@ -4745,7 +4747,7 @@ static void do_coproc_insn(DisasContext *s, int cpnum, int is64,
}
gen_set_condexec(s);
- gen_set_pc_im(s, s->pc_curr);
+ gen_update_pc(s, 0);
gen_helper_access_check_cp_reg(cpu_env,
tcg_constant_ptr(ri),
tcg_constant_i32(syndrome),
@@ -4756,7 +4758,7 @@ static void do_coproc_insn(DisasContext *s, int cpnum, int is64,
* synchronize the CPU state in case it does.
*/
gen_set_condexec(s);
- gen_set_pc_im(s, s->pc_curr);
+ gen_update_pc(s, 0);
}
/* Handle special cases first */
@@ -4770,7 +4772,7 @@ static void do_coproc_insn(DisasContext *s, int cpnum, int is64,
unallocated_encoding(s);
return;
}
- gen_set_pc_im(s, s->base.pc_next);
+ gen_update_pc(s, curr_insn_len(s));
s->base.is_jmp = DISAS_WFI;
return;
default:
@@ -5157,7 +5159,7 @@ static void gen_srs(DisasContext *s,
addr = tcg_temp_new_i32();
/* get_r13_banked() will raise an exception if called from System mode */
gen_set_condexec(s);
- gen_set_pc_im(s, s->pc_curr);
+ gen_update_pc(s, 0);
gen_helper_get_r13_banked(addr, cpu_env, tcg_constant_i32(mode));
switch (amode) {
case 0: /* DA */
@@ -6226,7 +6228,7 @@ static bool trans_YIELD(DisasContext *s, arg_YIELD *a)
* scheduling of other vCPUs.
*/
if (!(tb_cflags(s->base.tb) & CF_PARALLEL)) {
- gen_set_pc_im(s, s->base.pc_next);
+ gen_update_pc(s, curr_insn_len(s));
s->base.is_jmp = DISAS_YIELD;
}
return true;
@@ -6242,7 +6244,7 @@ static bool trans_WFE(DisasContext *s, arg_WFE *a)
* implemented so we can't sleep like WFI does.
*/
if (!(tb_cflags(s->base.tb) & CF_PARALLEL)) {
- gen_set_pc_im(s, s->base.pc_next);
+ gen_update_pc(s, curr_insn_len(s));
s->base.is_jmp = DISAS_WFE;
}
return true;
@@ -6251,7 +6253,7 @@ static bool trans_WFE(DisasContext *s, arg_WFE *a)
static bool trans_WFI(DisasContext *s, arg_WFI *a)
{
/* For WFI, halt the vCPU until an IRQ. */
- gen_set_pc_im(s, s->base.pc_next);
+ gen_update_pc(s, curr_insn_len(s));
s->base.is_jmp = DISAS_WFI;
return true;
}
@@ -8761,7 +8763,7 @@ static bool trans_SVC(DisasContext *s, arg_SVC *a)
(a->imm == semihost_imm)) {
gen_exception_internal_insn(s, s->pc_curr, EXCP_SEMIHOST);
} else {
- gen_set_pc_im(s, s->base.pc_next);
+ gen_update_pc(s, curr_insn_len(s));
s->svc_imm = a->imm;
s->base.is_jmp = DISAS_SWI;
}
@@ -9774,7 +9776,7 @@ static void arm_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu)
case DISAS_TOO_MANY:
case DISAS_UPDATE_EXIT:
case DISAS_UPDATE_NOCHAIN:
- gen_set_pc_im(dc, dc->base.pc_next);
+ gen_update_pc(dc, curr_insn_len(dc));
/* fall through */
default:
/* FIXME: Single stepping a WFI insn will not halt the CPU. */
@@ -9798,13 +9800,13 @@ static void arm_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu)
gen_goto_tb(dc, 1, curr_insn_len(dc));
break;
case DISAS_UPDATE_NOCHAIN:
- gen_set_pc_im(dc, dc->base.pc_next);
+ gen_update_pc(dc, curr_insn_len(dc));
/* fall through */
case DISAS_JUMP:
gen_goto_ptr();
break;
case DISAS_UPDATE_EXIT:
- gen_set_pc_im(dc, dc->base.pc_next);
+ gen_update_pc(dc, curr_insn_len(dc));
/* fall through */
default:
/* indicate that the hash table must be used to find the next TB */
@@ -9844,7 +9846,7 @@ static void arm_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu)
gen_set_label(dc->condlabel);
gen_set_condexec(dc);
if (unlikely(dc->ss_active)) {
- gen_set_pc_im(dc, dc->base.pc_next);
+ gen_update_pc(dc, curr_insn_len(dc));
gen_singlestep_exception(dc);
} else {
gen_goto_tb(dc, 1, curr_insn_len(dc));
--
2.34.1
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PATCH v8 4/9] target/arm: Change gen_exception_insn* to work on displacements
2022-10-20 3:06 [PATCH v8 0/9] target/arm: pc-relative translation blocks Richard Henderson
` (2 preceding siblings ...)
2022-10-20 3:06 ` [PATCH v8 3/9] target/arm: Change gen_*set_pc_im to gen_*update_pc Richard Henderson
@ 2022-10-20 3:06 ` Richard Henderson
2022-10-20 3:06 ` [PATCH v8 5/9] target/arm: Remove gen_exception_internal_insn pc argument Richard Henderson
` (5 subsequent siblings)
9 siblings, 0 replies; 11+ messages in thread
From: Richard Henderson @ 2022-10-20 3:06 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Philippe Mathieu-Daudé
In preparation for TARGET_TB_PCREL, reduce reliance on absolute values.
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/translate.h | 5 +++--
target/arm/translate-a64.c | 28 ++++++++++-------------
target/arm/translate-m-nocp.c | 6 ++---
target/arm/translate-mve.c | 2 +-
target/arm/translate-vfp.c | 6 ++---
target/arm/translate.c | 42 +++++++++++++++++------------------
6 files changed, 43 insertions(+), 46 deletions(-)
diff --git a/target/arm/translate.h b/target/arm/translate.h
index d651044855..4aa239e23c 100644
--- a/target/arm/translate.h
+++ b/target/arm/translate.h
@@ -281,9 +281,10 @@ void arm_jump_cc(DisasCompare *cmp, TCGLabel *label);
void arm_gen_test_cc(int cc, TCGLabel *label);
MemOp pow2_align(unsigned i);
void unallocated_encoding(DisasContext *s);
-void gen_exception_insn_el(DisasContext *s, uint64_t pc, int excp,
+void gen_exception_insn_el(DisasContext *s, target_long pc_diff, int excp,
uint32_t syn, uint32_t target_el);
-void gen_exception_insn(DisasContext *s, uint64_t pc, int excp, uint32_t syn);
+void gen_exception_insn(DisasContext *s, target_long pc_diff,
+ int excp, uint32_t syn);
/* Return state of Alternate Half-precision flag, caller frees result */
static inline TCGv_i32 get_ahp_flag(void)
diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c
index 585d42d5b2..49380e1cfe 100644
--- a/target/arm/translate-a64.c
+++ b/target/arm/translate-a64.c
@@ -1155,7 +1155,7 @@ static bool fp_access_check_only(DisasContext *s)
assert(!s->fp_access_checked);
s->fp_access_checked = true;
- gen_exception_insn_el(s, s->pc_curr, EXCP_UDEF,
+ gen_exception_insn_el(s, 0, EXCP_UDEF,
syn_fp_access_trap(1, 0xe, false, 0),
s->fp_excp_el);
return false;
@@ -1170,7 +1170,7 @@ static bool fp_access_check(DisasContext *s)
return false;
}
if (s->sme_trap_nonstreaming && s->is_nonstreaming) {
- gen_exception_insn(s, s->pc_curr, EXCP_UDEF,
+ gen_exception_insn(s, 0, EXCP_UDEF,
syn_smetrap(SME_ET_Streaming, false));
return false;
}
@@ -1190,7 +1190,7 @@ bool sve_access_check(DisasContext *s)
goto fail_exit;
}
} else if (s->sve_excp_el) {
- gen_exception_insn_el(s, s->pc_curr, EXCP_UDEF,
+ gen_exception_insn_el(s, 0, EXCP_UDEF,
syn_sve_access_trap(), s->sve_excp_el);
goto fail_exit;
}
@@ -1212,7 +1212,7 @@ bool sve_access_check(DisasContext *s)
static bool sme_access_check(DisasContext *s)
{
if (s->sme_excp_el) {
- gen_exception_insn_el(s, s->pc_curr, EXCP_UDEF,
+ gen_exception_insn_el(s, 0, EXCP_UDEF,
syn_smetrap(SME_ET_AccessTrap, false),
s->sme_excp_el);
return false;
@@ -1242,12 +1242,12 @@ bool sme_enabled_check_with_svcr(DisasContext *s, unsigned req)
return false;
}
if (FIELD_EX64(req, SVCR, SM) && !s->pstate_sm) {
- gen_exception_insn(s, s->pc_curr, EXCP_UDEF,
+ gen_exception_insn(s, 0, EXCP_UDEF,
syn_smetrap(SME_ET_NotStreaming, false));
return false;
}
if (FIELD_EX64(req, SVCR, ZA) && !s->pstate_za) {
- gen_exception_insn(s, s->pc_curr, EXCP_UDEF,
+ gen_exception_insn(s, 0, EXCP_UDEF,
syn_smetrap(SME_ET_InactiveZA, false));
return false;
}
@@ -1907,7 +1907,7 @@ static void gen_sysreg_undef(DisasContext *s, bool isread,
} else {
syndrome = syn_uncategorized();
}
- gen_exception_insn(s, s->pc_curr, EXCP_UDEF, syndrome);
+ gen_exception_insn(s, 0, EXCP_UDEF, syndrome);
}
/* MRS - move from system register
@@ -2161,8 +2161,7 @@ static void disas_exc(DisasContext *s, uint32_t insn)
switch (op2_ll) {
case 1: /* SVC */
gen_ss_advance(s);
- gen_exception_insn(s, s->base.pc_next, EXCP_SWI,
- syn_aa64_svc(imm16));
+ gen_exception_insn(s, 4, EXCP_SWI, syn_aa64_svc(imm16));
break;
case 2: /* HVC */
if (s->current_el == 0) {
@@ -2175,8 +2174,7 @@ static void disas_exc(DisasContext *s, uint32_t insn)
gen_a64_update_pc(s, 0);
gen_helper_pre_hvc(cpu_env);
gen_ss_advance(s);
- gen_exception_insn_el(s, s->base.pc_next, EXCP_HVC,
- syn_aa64_hvc(imm16), 2);
+ gen_exception_insn_el(s, 4, EXCP_HVC, syn_aa64_hvc(imm16), 2);
break;
case 3: /* SMC */
if (s->current_el == 0) {
@@ -2186,8 +2184,7 @@ static void disas_exc(DisasContext *s, uint32_t insn)
gen_a64_update_pc(s, 0);
gen_helper_pre_smc(cpu_env, tcg_constant_i32(syn_aa64_smc(imm16)));
gen_ss_advance(s);
- gen_exception_insn_el(s, s->base.pc_next, EXCP_SMC,
- syn_aa64_smc(imm16), 3);
+ gen_exception_insn_el(s, 4, EXCP_SMC, syn_aa64_smc(imm16), 3);
break;
default:
unallocated_encoding(s);
@@ -14825,7 +14822,7 @@ static void aarch64_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu)
* Illegal execution state. This has priority over BTI
* exceptions, but comes after instruction abort exceptions.
*/
- gen_exception_insn(s, s->pc_curr, EXCP_UDEF, syn_illegalstate());
+ gen_exception_insn(s, 0, EXCP_UDEF, syn_illegalstate());
return;
}
@@ -14856,8 +14853,7 @@ static void aarch64_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu)
if (s->btype != 0
&& s->guarded_page
&& !btype_destination_ok(insn, s->bt, s->btype)) {
- gen_exception_insn(s, s->pc_curr, EXCP_UDEF,
- syn_btitrap(s->btype));
+ gen_exception_insn(s, 0, EXCP_UDEF, syn_btitrap(s->btype));
return;
}
} else {
diff --git a/target/arm/translate-m-nocp.c b/target/arm/translate-m-nocp.c
index 4029d7fdd4..694fae7e2e 100644
--- a/target/arm/translate-m-nocp.c
+++ b/target/arm/translate-m-nocp.c
@@ -143,7 +143,7 @@ static bool trans_VSCCLRM(DisasContext *s, arg_VSCCLRM *a)
tcg_gen_brcondi_i32(TCG_COND_EQ, sfpa, 0, s->condlabel);
if (s->fp_excp_el != 0) {
- gen_exception_insn_el(s, s->pc_curr, EXCP_NOCP,
+ gen_exception_insn_el(s, 0, EXCP_NOCP,
syn_uncategorized(), s->fp_excp_el);
return true;
}
@@ -765,12 +765,12 @@ static bool trans_NOCP(DisasContext *s, arg_nocp *a)
}
if (a->cp != 10) {
- gen_exception_insn(s, s->pc_curr, EXCP_NOCP, syn_uncategorized());
+ gen_exception_insn(s, 0, EXCP_NOCP, syn_uncategorized());
return true;
}
if (s->fp_excp_el != 0) {
- gen_exception_insn_el(s, s->pc_curr, EXCP_NOCP,
+ gen_exception_insn_el(s, 0, EXCP_NOCP,
syn_uncategorized(), s->fp_excp_el);
return true;
}
diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c
index 0cf1b5ea4f..db7ea3f603 100644
--- a/target/arm/translate-mve.c
+++ b/target/arm/translate-mve.c
@@ -100,7 +100,7 @@ bool mve_eci_check(DisasContext *s)
return true;
default:
/* Reserved value: INVSTATE UsageFault */
- gen_exception_insn(s, s->pc_curr, EXCP_INVSTATE, syn_uncategorized());
+ gen_exception_insn(s, 0, EXCP_INVSTATE, syn_uncategorized());
return false;
}
}
diff --git a/target/arm/translate-vfp.c b/target/arm/translate-vfp.c
index 070f465b17..5c5d58d2c6 100644
--- a/target/arm/translate-vfp.c
+++ b/target/arm/translate-vfp.c
@@ -230,7 +230,7 @@ static bool vfp_access_check_a(DisasContext *s, bool ignore_vfp_enabled)
int coproc = arm_dc_feature(s, ARM_FEATURE_V8) ? 0 : 0xa;
uint32_t syn = syn_fp_access_trap(1, 0xe, false, coproc);
- gen_exception_insn_el(s, s->pc_curr, EXCP_UDEF, syn, s->fp_excp_el);
+ gen_exception_insn_el(s, 0, EXCP_UDEF, syn, s->fp_excp_el);
return false;
}
@@ -240,7 +240,7 @@ static bool vfp_access_check_a(DisasContext *s, bool ignore_vfp_enabled)
* appear to be any insns which touch VFP which are allowed.
*/
if (s->sme_trap_nonstreaming) {
- gen_exception_insn(s, s->pc_curr, EXCP_UDEF,
+ gen_exception_insn(s, 0, EXCP_UDEF,
syn_smetrap(SME_ET_Streaming,
curr_insn_len(s) == 2));
return false;
@@ -272,7 +272,7 @@ bool vfp_access_check_m(DisasContext *s, bool skip_context_update)
* the encoding space handled by the patterns in m-nocp.decode,
* and for them we may need to raise NOCP here.
*/
- gen_exception_insn_el(s, s->pc_curr, EXCP_NOCP,
+ gen_exception_insn_el(s, 0, EXCP_NOCP,
syn_uncategorized(), s->fp_excp_el);
return false;
}
diff --git a/target/arm/translate.c b/target/arm/translate.c
index 9863a08f49..350f991649 100644
--- a/target/arm/translate.c
+++ b/target/arm/translate.c
@@ -1099,32 +1099,34 @@ static void gen_exception(int excp, uint32_t syndrome)
tcg_constant_i32(syndrome));
}
-static void gen_exception_insn_el_v(DisasContext *s, uint64_t pc, int excp,
- uint32_t syn, TCGv_i32 tcg_el)
+static void gen_exception_insn_el_v(DisasContext *s, target_long pc_diff,
+ int excp, uint32_t syn, TCGv_i32 tcg_el)
{
if (s->aarch64) {
- gen_a64_update_pc(s, pc - s->pc_curr);
+ gen_a64_update_pc(s, pc_diff);
} else {
gen_set_condexec(s);
- gen_update_pc(s, pc - s->pc_curr);
+ gen_update_pc(s, pc_diff);
}
gen_exception_el_v(excp, syn, tcg_el);
s->base.is_jmp = DISAS_NORETURN;
}
-void gen_exception_insn_el(DisasContext *s, uint64_t pc, int excp,
+void gen_exception_insn_el(DisasContext *s, target_long pc_diff, int excp,
uint32_t syn, uint32_t target_el)
{
- gen_exception_insn_el_v(s, pc, excp, syn, tcg_constant_i32(target_el));
+ gen_exception_insn_el_v(s, pc_diff, excp, syn,
+ tcg_constant_i32(target_el));
}
-void gen_exception_insn(DisasContext *s, uint64_t pc, int excp, uint32_t syn)
+void gen_exception_insn(DisasContext *s, target_long pc_diff,
+ int excp, uint32_t syn)
{
if (s->aarch64) {
- gen_a64_update_pc(s, pc - s->pc_curr);
+ gen_a64_update_pc(s, pc_diff);
} else {
gen_set_condexec(s);
- gen_update_pc(s, pc - s->pc_curr);
+ gen_update_pc(s, pc_diff);
}
gen_exception(excp, syn);
s->base.is_jmp = DISAS_NORETURN;
@@ -1141,7 +1143,7 @@ static void gen_exception_bkpt_insn(DisasContext *s, uint32_t syn)
void unallocated_encoding(DisasContext *s)
{
/* Unallocated and reserved encodings are uncategorized */
- gen_exception_insn(s, s->pc_curr, EXCP_UDEF, syn_uncategorized());
+ gen_exception_insn(s, 0, EXCP_UDEF, syn_uncategorized());
}
/* Force a TB lookup after an instruction that changes the CPU state. */
@@ -2865,7 +2867,7 @@ static bool msr_banked_access_decode(DisasContext *s, int r, int sysm, int rn,
tcg_el = tcg_constant_i32(3);
}
- gen_exception_insn_el_v(s, s->pc_curr, EXCP_UDEF,
+ gen_exception_insn_el_v(s, 0, EXCP_UDEF,
syn_uncategorized(), tcg_el);
tcg_temp_free_i32(tcg_el);
return false;
@@ -2891,7 +2893,7 @@ static bool msr_banked_access_decode(DisasContext *s, int r, int sysm, int rn,
undef:
/* If we get here then some access check did not pass */
- gen_exception_insn(s, s->pc_curr, EXCP_UDEF, syn_uncategorized());
+ gen_exception_insn(s, 0, EXCP_UDEF, syn_uncategorized());
return false;
}
@@ -5115,8 +5117,7 @@ static void gen_srs(DisasContext *s,
* For the UNPREDICTABLE cases we choose to UNDEF.
*/
if (s->current_el == 1 && !s->ns && mode == ARM_CPU_MODE_MON) {
- gen_exception_insn_el(s, s->pc_curr, EXCP_UDEF,
- syn_uncategorized(), 3);
+ gen_exception_insn_el(s, 0, EXCP_UDEF, syn_uncategorized(), 3);
return;
}
@@ -8498,7 +8499,7 @@ static bool trans_WLS(DisasContext *s, arg_WLS *a)
* Do the check-and-raise-exception by hand.
*/
if (s->fp_excp_el) {
- gen_exception_insn_el(s, s->pc_curr, EXCP_NOCP,
+ gen_exception_insn_el(s, 0, EXCP_NOCP,
syn_uncategorized(), s->fp_excp_el);
return true;
}
@@ -8601,7 +8602,7 @@ static bool trans_LE(DisasContext *s, arg_LE *a)
tmp = load_cpu_field(v7m.ltpsize);
tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, 4, skipexc);
tcg_temp_free_i32(tmp);
- gen_exception_insn(s, s->pc_curr, EXCP_INVSTATE, syn_uncategorized());
+ gen_exception_insn(s, 0, EXCP_INVSTATE, syn_uncategorized());
gen_set_label(skipexc);
}
@@ -9069,7 +9070,7 @@ static void disas_arm_insn(DisasContext *s, unsigned int insn)
* UsageFault exception.
*/
if (arm_dc_feature(s, ARM_FEATURE_M)) {
- gen_exception_insn(s, s->pc_curr, EXCP_INVSTATE, syn_uncategorized());
+ gen_exception_insn(s, 0, EXCP_INVSTATE, syn_uncategorized());
return;
}
@@ -9078,7 +9079,7 @@ static void disas_arm_insn(DisasContext *s, unsigned int insn)
* Illegal execution state. This has priority over BTI
* exceptions, but comes after instruction abort exceptions.
*/
- gen_exception_insn(s, s->pc_curr, EXCP_UDEF, syn_illegalstate());
+ gen_exception_insn(s, 0, EXCP_UDEF, syn_illegalstate());
return;
}
@@ -9642,7 +9643,7 @@ static void thumb_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu)
* Illegal execution state. This has priority over BTI
* exceptions, but comes after instruction abort exceptions.
*/
- gen_exception_insn(dc, dc->pc_curr, EXCP_UDEF, syn_illegalstate());
+ gen_exception_insn(dc, 0, EXCP_UDEF, syn_illegalstate());
return;
}
@@ -9715,8 +9716,7 @@ static void thumb_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu)
*/
tcg_remove_ops_after(dc->insn_eci_rewind);
dc->condjmp = 0;
- gen_exception_insn(dc, dc->pc_curr, EXCP_INVSTATE,
- syn_uncategorized());
+ gen_exception_insn(dc, 0, EXCP_INVSTATE, syn_uncategorized());
}
arm_post_translate_insn(dc);
--
2.34.1
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PATCH v8 5/9] target/arm: Remove gen_exception_internal_insn pc argument
2022-10-20 3:06 [PATCH v8 0/9] target/arm: pc-relative translation blocks Richard Henderson
` (3 preceding siblings ...)
2022-10-20 3:06 ` [PATCH v8 4/9] target/arm: Change gen_exception_insn* to work on displacements Richard Henderson
@ 2022-10-20 3:06 ` Richard Henderson
2022-10-20 3:06 ` [PATCH v8 6/9] target/arm: Change gen_jmp* to work on displacements Richard Henderson
` (4 subsequent siblings)
9 siblings, 0 replies; 11+ messages in thread
From: Richard Henderson @ 2022-10-20 3:06 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Philippe Mathieu-Daudé
In preparation for TARGET_TB_PCREL, reduce reliance on absolute values.
Since we always pass dc->pc_curr, fold the arithmetic to zero displacement.
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/translate-a64.c | 6 +++---
target/arm/translate.c | 10 +++++-----
2 files changed, 8 insertions(+), 8 deletions(-)
diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c
index 49380e1cfe..623f7e2e96 100644
--- a/target/arm/translate-a64.c
+++ b/target/arm/translate-a64.c
@@ -332,9 +332,9 @@ static void gen_exception_internal(int excp)
gen_helper_exception_internal(cpu_env, tcg_constant_i32(excp));
}
-static void gen_exception_internal_insn(DisasContext *s, uint64_t pc, int excp)
+static void gen_exception_internal_insn(DisasContext *s, int excp)
{
- gen_a64_update_pc(s, pc - s->pc_curr);
+ gen_a64_update_pc(s, 0);
gen_exception_internal(excp);
s->base.is_jmp = DISAS_NORETURN;
}
@@ -2211,7 +2211,7 @@ static void disas_exc(DisasContext *s, uint32_t insn)
* Secondly, "HLT 0xf000" is the A64 semihosting syscall instruction.
*/
if (semihosting_enabled(s->current_el == 0) && imm16 == 0xf000) {
- gen_exception_internal_insn(s, s->pc_curr, EXCP_SEMIHOST);
+ gen_exception_internal_insn(s, EXCP_SEMIHOST);
} else {
unallocated_encoding(s);
}
diff --git a/target/arm/translate.c b/target/arm/translate.c
index 350f991649..9104ab8232 100644
--- a/target/arm/translate.c
+++ b/target/arm/translate.c
@@ -1074,10 +1074,10 @@ static inline void gen_smc(DisasContext *s)
s->base.is_jmp = DISAS_SMC;
}
-static void gen_exception_internal_insn(DisasContext *s, uint32_t pc, int excp)
+static void gen_exception_internal_insn(DisasContext *s, int excp)
{
gen_set_condexec(s);
- gen_update_pc(s, pc - s->pc_curr);
+ gen_update_pc(s, 0);
gen_exception_internal(excp);
s->base.is_jmp = DISAS_NORETURN;
}
@@ -1169,7 +1169,7 @@ static inline void gen_hlt(DisasContext *s, int imm)
*/
if (semihosting_enabled(s->current_el != 0) &&
(imm == (s->thumb ? 0x3c : 0xf000))) {
- gen_exception_internal_insn(s, s->pc_curr, EXCP_SEMIHOST);
+ gen_exception_internal_insn(s, EXCP_SEMIHOST);
return;
}
@@ -6556,7 +6556,7 @@ static bool trans_BKPT(DisasContext *s, arg_BKPT *a)
if (arm_dc_feature(s, ARM_FEATURE_M) &&
semihosting_enabled(s->current_el == 0) &&
(a->imm == 0xab)) {
- gen_exception_internal_insn(s, s->pc_curr, EXCP_SEMIHOST);
+ gen_exception_internal_insn(s, EXCP_SEMIHOST);
} else {
gen_exception_bkpt_insn(s, syn_aa32_bkpt(a->imm, false));
}
@@ -8762,7 +8762,7 @@ static bool trans_SVC(DisasContext *s, arg_SVC *a)
if (!arm_dc_feature(s, ARM_FEATURE_M) &&
semihosting_enabled(s->current_el == 0) &&
(a->imm == semihost_imm)) {
- gen_exception_internal_insn(s, s->pc_curr, EXCP_SEMIHOST);
+ gen_exception_internal_insn(s, EXCP_SEMIHOST);
} else {
gen_update_pc(s, curr_insn_len(s));
s->svc_imm = a->imm;
--
2.34.1
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PATCH v8 6/9] target/arm: Change gen_jmp* to work on displacements
2022-10-20 3:06 [PATCH v8 0/9] target/arm: pc-relative translation blocks Richard Henderson
` (4 preceding siblings ...)
2022-10-20 3:06 ` [PATCH v8 5/9] target/arm: Remove gen_exception_internal_insn pc argument Richard Henderson
@ 2022-10-20 3:06 ` Richard Henderson
2022-10-20 3:06 ` [PATCH v8 7/9] target/arm: Introduce gen_pc_plus_diff for aarch64 Richard Henderson
` (3 subsequent siblings)
9 siblings, 0 replies; 11+ messages in thread
From: Richard Henderson @ 2022-10-20 3:06 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Peter Maydell
In preparation for TARGET_TB_PCREL, reduce reliance on absolute values.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/translate.c | 37 +++++++++++++++++++++----------------
1 file changed, 21 insertions(+), 16 deletions(-)
diff --git a/target/arm/translate.c b/target/arm/translate.c
index 9104ab8232..ca128edab7 100644
--- a/target/arm/translate.c
+++ b/target/arm/translate.c
@@ -266,6 +266,12 @@ static uint32_t read_pc(DisasContext *s)
return s->pc_curr + (s->thumb ? 4 : 8);
}
+/* The pc_curr difference for an architectural jump. */
+static target_long jmp_diff(DisasContext *s, target_long diff)
+{
+ return diff + (s->thumb ? 4 : 8);
+}
+
/* Set a variable to the value of a CPU register. */
void load_reg_var(DisasContext *s, TCGv_i32 var, int reg)
{
@@ -2592,7 +2598,7 @@ static void gen_goto_ptr(void)
* cpu_loop_exec. Any live exit_requests will be processed as we
* enter the next TB.
*/
-static void gen_goto_tb(DisasContext *s, int n, int diff)
+static void gen_goto_tb(DisasContext *s, int n, target_long diff)
{
target_ulong dest = s->pc_curr + diff;
@@ -2608,10 +2614,8 @@ static void gen_goto_tb(DisasContext *s, int n, int diff)
}
/* Jump, specifying which TB number to use if we gen_goto_tb() */
-static inline void gen_jmp_tb(DisasContext *s, uint32_t dest, int tbno)
+static void gen_jmp_tb(DisasContext *s, target_long diff, int tbno)
{
- int diff = dest - s->pc_curr;
-
if (unlikely(s->ss_active)) {
/* An indirect jump so that we still trigger the debug exception. */
gen_update_pc(s, diff);
@@ -2653,9 +2657,9 @@ static inline void gen_jmp_tb(DisasContext *s, uint32_t dest, int tbno)
}
}
-static inline void gen_jmp(DisasContext *s, uint32_t dest)
+static inline void gen_jmp(DisasContext *s, target_long diff)
{
- gen_jmp_tb(s, dest, 0);
+ gen_jmp_tb(s, diff, 0);
}
static inline void gen_mulxy(TCGv_i32 t0, TCGv_i32 t1, int x, int y)
@@ -8322,7 +8326,7 @@ static bool trans_CLRM(DisasContext *s, arg_CLRM *a)
static bool trans_B(DisasContext *s, arg_i *a)
{
- gen_jmp(s, read_pc(s) + a->imm);
+ gen_jmp(s, jmp_diff(s, a->imm));
return true;
}
@@ -8337,14 +8341,14 @@ static bool trans_B_cond_thumb(DisasContext *s, arg_ci *a)
return true;
}
arm_skip_unless(s, a->cond);
- gen_jmp(s, read_pc(s) + a->imm);
+ gen_jmp(s, jmp_diff(s, a->imm));
return true;
}
static bool trans_BL(DisasContext *s, arg_i *a)
{
tcg_gen_movi_i32(cpu_R[14], s->base.pc_next | s->thumb);
- gen_jmp(s, read_pc(s) + a->imm);
+ gen_jmp(s, jmp_diff(s, a->imm));
return true;
}
@@ -8364,7 +8368,8 @@ static bool trans_BLX_i(DisasContext *s, arg_BLX_i *a)
}
tcg_gen_movi_i32(cpu_R[14], s->base.pc_next | s->thumb);
store_cpu_field_constant(!s->thumb, thumb);
- gen_jmp(s, (read_pc(s) & ~3) + a->imm);
+ /* This jump is computed from an aligned PC: subtract off the low bits. */
+ gen_jmp(s, jmp_diff(s, a->imm - (s->pc_curr & 3)));
return true;
}
@@ -8525,10 +8530,10 @@ static bool trans_WLS(DisasContext *s, arg_WLS *a)
* when we take this upcoming exit from this TB, so gen_jmp_tb() is OK.
*/
}
- gen_jmp_tb(s, s->base.pc_next, 1);
+ gen_jmp_tb(s, curr_insn_len(s), 1);
gen_set_label(nextlabel);
- gen_jmp(s, read_pc(s) + a->imm);
+ gen_jmp(s, jmp_diff(s, a->imm));
return true;
}
@@ -8608,7 +8613,7 @@ static bool trans_LE(DisasContext *s, arg_LE *a)
if (a->f) {
/* Loop-forever: just jump back to the loop start */
- gen_jmp(s, read_pc(s) - a->imm);
+ gen_jmp(s, jmp_diff(s, -a->imm));
return true;
}
@@ -8639,7 +8644,7 @@ static bool trans_LE(DisasContext *s, arg_LE *a)
tcg_temp_free_i32(decr);
}
/* Jump back to the loop start */
- gen_jmp(s, read_pc(s) - a->imm);
+ gen_jmp(s, jmp_diff(s, -a->imm));
gen_set_label(loopend);
if (a->tp) {
@@ -8647,7 +8652,7 @@ static bool trans_LE(DisasContext *s, arg_LE *a)
store_cpu_field(tcg_constant_i32(4), v7m.ltpsize);
}
/* End TB, continuing to following insn */
- gen_jmp_tb(s, s->base.pc_next, 1);
+ gen_jmp_tb(s, curr_insn_len(s), 1);
return true;
}
@@ -8746,7 +8751,7 @@ static bool trans_CBZ(DisasContext *s, arg_CBZ *a)
tcg_gen_brcondi_i32(a->nz ? TCG_COND_EQ : TCG_COND_NE,
tmp, 0, s->condlabel);
tcg_temp_free_i32(tmp);
- gen_jmp(s, read_pc(s) + a->imm);
+ gen_jmp(s, jmp_diff(s, a->imm));
return true;
}
--
2.34.1
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PATCH v8 7/9] target/arm: Introduce gen_pc_plus_diff for aarch64
2022-10-20 3:06 [PATCH v8 0/9] target/arm: pc-relative translation blocks Richard Henderson
` (5 preceding siblings ...)
2022-10-20 3:06 ` [PATCH v8 6/9] target/arm: Change gen_jmp* to work on displacements Richard Henderson
@ 2022-10-20 3:06 ` Richard Henderson
2022-10-20 3:06 ` [PATCH v8 8/9] target/arm: Introduce gen_pc_plus_diff for aarch32 Richard Henderson
` (2 subsequent siblings)
9 siblings, 0 replies; 11+ messages in thread
From: Richard Henderson @ 2022-10-20 3:06 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Peter Maydell
In preparation for TARGET_TB_PCREL, reduce reliance on absolute values.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/translate-a64.c | 41 +++++++++++++++++++++++++++-----------
1 file changed, 29 insertions(+), 12 deletions(-)
diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c
index 623f7e2e96..f9f8559c01 100644
--- a/target/arm/translate-a64.c
+++ b/target/arm/translate-a64.c
@@ -140,9 +140,14 @@ static void reset_btype(DisasContext *s)
}
}
+static void gen_pc_plus_diff(DisasContext *s, TCGv_i64 dest, target_long diff)
+{
+ tcg_gen_movi_i64(dest, s->pc_curr + diff);
+}
+
void gen_a64_update_pc(DisasContext *s, target_long diff)
{
- tcg_gen_movi_i64(cpu_pc, s->pc_curr + diff);
+ gen_pc_plus_diff(s, cpu_pc, diff);
}
/*
@@ -1360,7 +1365,7 @@ static void disas_uncond_b_imm(DisasContext *s, uint32_t insn)
if (insn & (1U << 31)) {
/* BL Branch with link */
- tcg_gen_movi_i64(cpu_reg(s, 30), s->base.pc_next);
+ gen_pc_plus_diff(s, cpu_reg(s, 30), curr_insn_len(s));
}
/* B Branch / BL Branch with link */
@@ -2301,11 +2306,17 @@ static void disas_uncond_b_reg(DisasContext *s, uint32_t insn)
default:
goto do_unallocated;
}
- gen_a64_set_pc(s, dst);
/* BLR also needs to load return address */
if (opc == 1) {
- tcg_gen_movi_i64(cpu_reg(s, 30), s->base.pc_next);
+ TCGv_i64 lr = cpu_reg(s, 30);
+ if (dst == lr) {
+ TCGv_i64 tmp = new_tmp_a64(s);
+ tcg_gen_mov_i64(tmp, dst);
+ dst = tmp;
+ }
+ gen_pc_plus_diff(s, lr, curr_insn_len(s));
}
+ gen_a64_set_pc(s, dst);
break;
case 8: /* BRAA */
@@ -2328,11 +2339,17 @@ static void disas_uncond_b_reg(DisasContext *s, uint32_t insn)
} else {
dst = cpu_reg(s, rn);
}
- gen_a64_set_pc(s, dst);
/* BLRAA also needs to load return address */
if (opc == 9) {
- tcg_gen_movi_i64(cpu_reg(s, 30), s->base.pc_next);
+ TCGv_i64 lr = cpu_reg(s, 30);
+ if (dst == lr) {
+ TCGv_i64 tmp = new_tmp_a64(s);
+ tcg_gen_mov_i64(tmp, dst);
+ dst = tmp;
+ }
+ gen_pc_plus_diff(s, lr, curr_insn_len(s));
}
+ gen_a64_set_pc(s, dst);
break;
case 4: /* ERET */
@@ -2900,7 +2917,8 @@ static void disas_ld_lit(DisasContext *s, uint32_t insn)
tcg_rt = cpu_reg(s, rt);
- clean_addr = tcg_constant_i64(s->pc_curr + imm);
+ clean_addr = new_tmp_a64(s);
+ gen_pc_plus_diff(s, clean_addr, imm);
if (is_vector) {
do_fp_ld(s, rt, clean_addr, size);
} else {
@@ -4244,23 +4262,22 @@ static void disas_ldst(DisasContext *s, uint32_t insn)
static void disas_pc_rel_adr(DisasContext *s, uint32_t insn)
{
unsigned int page, rd;
- uint64_t base;
- uint64_t offset;
+ int64_t offset;
page = extract32(insn, 31, 1);
/* SignExtend(immhi:immlo) -> offset */
offset = sextract64(insn, 5, 19);
offset = offset << 2 | extract32(insn, 29, 2);
rd = extract32(insn, 0, 5);
- base = s->pc_curr;
if (page) {
/* ADRP (page based) */
- base &= ~0xfff;
offset <<= 12;
+ /* The page offset is ok for TARGET_TB_PCREL. */
+ offset -= s->pc_curr & 0xfff;
}
- tcg_gen_movi_i64(cpu_reg(s, rd), base + offset);
+ gen_pc_plus_diff(s, cpu_reg(s, rd), offset);
}
/*
--
2.34.1
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PATCH v8 8/9] target/arm: Introduce gen_pc_plus_diff for aarch32
2022-10-20 3:06 [PATCH v8 0/9] target/arm: pc-relative translation blocks Richard Henderson
` (6 preceding siblings ...)
2022-10-20 3:06 ` [PATCH v8 7/9] target/arm: Introduce gen_pc_plus_diff for aarch64 Richard Henderson
@ 2022-10-20 3:06 ` Richard Henderson
2022-10-20 3:06 ` [PATCH v8 9/9] target/arm: Enable TARGET_TB_PCREL Richard Henderson
2022-10-20 10:29 ` [PATCH v8 0/9] target/arm: pc-relative translation blocks Peter Maydell
9 siblings, 0 replies; 11+ messages in thread
From: Richard Henderson @ 2022-10-20 3:06 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Philippe Mathieu-Daudé
In preparation for TARGET_TB_PCREL, reduce reliance on absolute values.
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/translate.c | 38 +++++++++++++++++++++-----------------
1 file changed, 21 insertions(+), 17 deletions(-)
diff --git a/target/arm/translate.c b/target/arm/translate.c
index ca128edab7..5f6bd9b5b7 100644
--- a/target/arm/translate.c
+++ b/target/arm/translate.c
@@ -260,23 +260,22 @@ static inline int get_a32_user_mem_index(DisasContext *s)
}
}
-/* The architectural value of PC. */
-static uint32_t read_pc(DisasContext *s)
-{
- return s->pc_curr + (s->thumb ? 4 : 8);
-}
-
/* The pc_curr difference for an architectural jump. */
static target_long jmp_diff(DisasContext *s, target_long diff)
{
return diff + (s->thumb ? 4 : 8);
}
+static void gen_pc_plus_diff(DisasContext *s, TCGv_i32 var, target_long diff)
+{
+ tcg_gen_movi_i32(var, s->pc_curr + diff);
+}
+
/* Set a variable to the value of a CPU register. */
void load_reg_var(DisasContext *s, TCGv_i32 var, int reg)
{
if (reg == 15) {
- tcg_gen_movi_i32(var, read_pc(s));
+ gen_pc_plus_diff(s, var, jmp_diff(s, 0));
} else {
tcg_gen_mov_i32(var, cpu_R[reg]);
}
@@ -292,7 +291,11 @@ TCGv_i32 add_reg_for_lit(DisasContext *s, int reg, int ofs)
TCGv_i32 tmp = tcg_temp_new_i32();
if (reg == 15) {
- tcg_gen_movi_i32(tmp, (read_pc(s) & ~3) + ofs);
+ /*
+ * This address is computed from an aligned PC:
+ * subtract off the low bits.
+ */
+ gen_pc_plus_diff(s, tmp, jmp_diff(s, ofs - (s->pc_curr & 3)));
} else {
tcg_gen_addi_i32(tmp, cpu_R[reg], ofs);
}
@@ -1155,7 +1158,7 @@ void unallocated_encoding(DisasContext *s)
/* Force a TB lookup after an instruction that changes the CPU state. */
void gen_lookup_tb(DisasContext *s)
{
- tcg_gen_movi_i32(cpu_R[15], s->base.pc_next);
+ gen_pc_plus_diff(s, cpu_R[15], curr_insn_len(s));
s->base.is_jmp = DISAS_EXIT;
}
@@ -6479,7 +6482,7 @@ static bool trans_BLX_r(DisasContext *s, arg_BLX_r *a)
return false;
}
tmp = load_reg(s, a->rm);
- tcg_gen_movi_i32(cpu_R[14], s->base.pc_next | s->thumb);
+ gen_pc_plus_diff(s, cpu_R[14], curr_insn_len(s) | s->thumb);
gen_bx(s, tmp);
return true;
}
@@ -8347,7 +8350,7 @@ static bool trans_B_cond_thumb(DisasContext *s, arg_ci *a)
static bool trans_BL(DisasContext *s, arg_i *a)
{
- tcg_gen_movi_i32(cpu_R[14], s->base.pc_next | s->thumb);
+ gen_pc_plus_diff(s, cpu_R[14], curr_insn_len(s) | s->thumb);
gen_jmp(s, jmp_diff(s, a->imm));
return true;
}
@@ -8366,7 +8369,7 @@ static bool trans_BLX_i(DisasContext *s, arg_BLX_i *a)
if (s->thumb && (a->imm & 2)) {
return false;
}
- tcg_gen_movi_i32(cpu_R[14], s->base.pc_next | s->thumb);
+ gen_pc_plus_diff(s, cpu_R[14], curr_insn_len(s) | s->thumb);
store_cpu_field_constant(!s->thumb, thumb);
/* This jump is computed from an aligned PC: subtract off the low bits. */
gen_jmp(s, jmp_diff(s, a->imm - (s->pc_curr & 3)));
@@ -8376,7 +8379,7 @@ static bool trans_BLX_i(DisasContext *s, arg_BLX_i *a)
static bool trans_BL_BLX_prefix(DisasContext *s, arg_BL_BLX_prefix *a)
{
assert(!arm_dc_feature(s, ARM_FEATURE_THUMB2));
- tcg_gen_movi_i32(cpu_R[14], read_pc(s) + (a->imm << 12));
+ gen_pc_plus_diff(s, cpu_R[14], jmp_diff(s, a->imm << 12));
return true;
}
@@ -8386,7 +8389,7 @@ static bool trans_BL_suffix(DisasContext *s, arg_BL_suffix *a)
assert(!arm_dc_feature(s, ARM_FEATURE_THUMB2));
tcg_gen_addi_i32(tmp, cpu_R[14], (a->imm << 1) | 1);
- tcg_gen_movi_i32(cpu_R[14], s->base.pc_next | 1);
+ gen_pc_plus_diff(s, cpu_R[14], curr_insn_len(s) | 1);
gen_bx(s, tmp);
return true;
}
@@ -8402,7 +8405,7 @@ static bool trans_BLX_suffix(DisasContext *s, arg_BLX_suffix *a)
tmp = tcg_temp_new_i32();
tcg_gen_addi_i32(tmp, cpu_R[14], a->imm << 1);
tcg_gen_andi_i32(tmp, tmp, 0xfffffffc);
- tcg_gen_movi_i32(cpu_R[14], s->base.pc_next | 1);
+ gen_pc_plus_diff(s, cpu_R[14], curr_insn_len(s) | 1);
gen_bx(s, tmp);
return true;
}
@@ -8725,10 +8728,11 @@ static bool op_tbranch(DisasContext *s, arg_tbranch *a, bool half)
tcg_gen_add_i32(addr, addr, tmp);
gen_aa32_ld_i32(s, tmp, addr, get_mem_index(s), half ? MO_UW : MO_UB);
- tcg_temp_free_i32(addr);
tcg_gen_add_i32(tmp, tmp, tmp);
- tcg_gen_addi_i32(tmp, tmp, read_pc(s));
+ gen_pc_plus_diff(s, addr, jmp_diff(s, 0));
+ tcg_gen_add_i32(tmp, tmp, addr);
+ tcg_temp_free_i32(addr);
store_reg(s, 15, tmp);
return true;
}
--
2.34.1
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PATCH v8 9/9] target/arm: Enable TARGET_TB_PCREL
2022-10-20 3:06 [PATCH v8 0/9] target/arm: pc-relative translation blocks Richard Henderson
` (7 preceding siblings ...)
2022-10-20 3:06 ` [PATCH v8 8/9] target/arm: Introduce gen_pc_plus_diff for aarch32 Richard Henderson
@ 2022-10-20 3:06 ` Richard Henderson
2022-10-20 10:29 ` [PATCH v8 0/9] target/arm: pc-relative translation blocks Peter Maydell
9 siblings, 0 replies; 11+ messages in thread
From: Richard Henderson @ 2022-10-20 3:06 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
v7: Introduce DisasLabel to clean up pc_save frobbing.
Adjust pc_save around tcg_remove_ops_after.
---
target/arm/cpu-param.h | 1 +
target/arm/translate.h | 50 +++++++++++++++-
target/arm/cpu.c | 23 ++++----
target/arm/translate-a64.c | 64 +++++++++++++-------
target/arm/translate-m-nocp.c | 2 +-
target/arm/translate.c | 108 +++++++++++++++++++++++-----------
6 files changed, 177 insertions(+), 71 deletions(-)
diff --git a/target/arm/cpu-param.h b/target/arm/cpu-param.h
index 08681828ac..ae472cf330 100644
--- a/target/arm/cpu-param.h
+++ b/target/arm/cpu-param.h
@@ -30,6 +30,7 @@
*/
# define TARGET_PAGE_BITS_VARY
# define TARGET_PAGE_BITS_MIN 10
+# define TARGET_TB_PCREL 1
#endif
#define NB_MMU_MODES 8
diff --git a/target/arm/translate.h b/target/arm/translate.h
index 4aa239e23c..3cdc7dbc2f 100644
--- a/target/arm/translate.h
+++ b/target/arm/translate.h
@@ -6,18 +6,42 @@
/* internal defines */
+
+/*
+ * Save pc_save across a branch, so that we may restore the value from
+ * before the branch at the point the label is emitted.
+ */
+typedef struct DisasLabel {
+ TCGLabel *label;
+ target_ulong pc_save;
+} DisasLabel;
+
typedef struct DisasContext {
DisasContextBase base;
const ARMISARegisters *isar;
/* The address of the current instruction being translated. */
target_ulong pc_curr;
+ /*
+ * For TARGET_TB_PCREL, the full value of cpu_pc is not known
+ * (although the page offset is known). For convenience, the
+ * translation loop uses the full virtual address that triggered
+ * the translation, from base.pc_start through pc_curr.
+ * For efficiency, we do not update cpu_pc for every instruction.
+ * Instead, pc_save has the value of pc_curr at the time of the
+ * last update to cpu_pc, which allows us to compute the addend
+ * needed to bring cpu_pc current: pc_curr - pc_save.
+ * If cpu_pc now contains the destination of an indirect branch,
+ * pc_save contains -1 to indicate that relative updates are no
+ * longer possible.
+ */
+ target_ulong pc_save;
target_ulong page_start;
uint32_t insn;
/* Nonzero if this instruction has been conditionally skipped. */
int condjmp;
/* The label that will be jumped to when the instruction is skipped. */
- TCGLabel *condlabel;
+ DisasLabel condlabel;
/* Thumb-2 conditional execution bits. */
int condexec_mask;
int condexec_cond;
@@ -28,8 +52,6 @@ typedef struct DisasContext {
* after decode (ie after any UNDEF checks)
*/
bool eci_handled;
- /* TCG op to rewind to if this turns out to be an invalid ECI state */
- TCGOp *insn_eci_rewind;
int sctlr_b;
MemOp be_data;
#if !defined(CONFIG_USER_ONLY)
@@ -566,6 +588,28 @@ static inline MemOp finalize_memop(DisasContext *s, MemOp opc)
*/
uint64_t asimd_imm_const(uint32_t imm, int cmode, int op);
+/*
+ * gen_disas_label:
+ * Create a label and cache a copy of pc_save.
+ */
+static inline DisasLabel gen_disas_label(DisasContext *s)
+{
+ return (DisasLabel){
+ .label = gen_new_label(),
+ .pc_save = s->pc_save,
+ };
+}
+
+/*
+ * set_disas_label:
+ * Emit a label and restore the cached copy of pc_save.
+ */
+static inline void set_disas_label(DisasContext *s, DisasLabel l)
+{
+ gen_set_label(l.label);
+ s->pc_save = l.pc_save;
+}
+
/*
* Helpers for implementing sets of trans_* functions.
* Defer the implementation of NAME to FUNC, with optional extra arguments.
diff --git a/target/arm/cpu.c b/target/arm/cpu.c
index 94ca6f163f..0bc5e9b125 100644
--- a/target/arm/cpu.c
+++ b/target/arm/cpu.c
@@ -76,17 +76,18 @@ static vaddr arm_cpu_get_pc(CPUState *cs)
void arm_cpu_synchronize_from_tb(CPUState *cs,
const TranslationBlock *tb)
{
- ARMCPU *cpu = ARM_CPU(cs);
- CPUARMState *env = &cpu->env;
-
- /*
- * It's OK to look at env for the current mode here, because it's
- * never possible for an AArch64 TB to chain to an AArch32 TB.
- */
- if (is_a64(env)) {
- env->pc = tb_pc(tb);
- } else {
- env->regs[15] = tb_pc(tb);
+ /* The program counter is always up to date with TARGET_TB_PCREL. */
+ if (!TARGET_TB_PCREL) {
+ CPUARMState *env = cs->env_ptr;
+ /*
+ * It's OK to look at env for the current mode here, because it's
+ * never possible for an AArch64 TB to chain to an AArch32 TB.
+ */
+ if (is_a64(env)) {
+ env->pc = tb_pc(tb);
+ } else {
+ env->regs[15] = tb_pc(tb);
+ }
}
}
#endif /* CONFIG_TCG */
diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c
index f9f8559c01..32e95cc2f4 100644
--- a/target/arm/translate-a64.c
+++ b/target/arm/translate-a64.c
@@ -142,12 +142,18 @@ static void reset_btype(DisasContext *s)
static void gen_pc_plus_diff(DisasContext *s, TCGv_i64 dest, target_long diff)
{
- tcg_gen_movi_i64(dest, s->pc_curr + diff);
+ assert(s->pc_save != -1);
+ if (TARGET_TB_PCREL) {
+ tcg_gen_addi_i64(dest, cpu_pc, (s->pc_curr - s->pc_save) + diff);
+ } else {
+ tcg_gen_movi_i64(dest, s->pc_curr + diff);
+ }
}
void gen_a64_update_pc(DisasContext *s, target_long diff)
{
gen_pc_plus_diff(s, cpu_pc, diff);
+ s->pc_save = s->pc_curr + diff;
}
/*
@@ -201,6 +207,7 @@ static void gen_a64_set_pc(DisasContext *s, TCGv_i64 src)
* then loading an address into the PC will clear out any tag.
*/
gen_top_byte_ignore(s, cpu_pc, src, s->tbii);
+ s->pc_save = -1;
}
/*
@@ -377,11 +384,22 @@ static inline bool use_goto_tb(DisasContext *s, uint64_t dest)
static void gen_goto_tb(DisasContext *s, int n, int64_t diff)
{
- uint64_t dest = s->pc_curr + diff;
-
- if (use_goto_tb(s, dest)) {
- tcg_gen_goto_tb(n);
- gen_a64_update_pc(s, diff);
+ if (use_goto_tb(s, s->pc_curr + diff)) {
+ /*
+ * For pcrel, the pc must always be up-to-date on entry to
+ * the linked TB, so that it can use simple additions for all
+ * further adjustments. For !pcrel, the linked TB is compiled
+ * to know its full virtual address, so we can delay the
+ * update to pc to the unlinked path. A long chain of links
+ * can thus avoid many updates to the PC.
+ */
+ if (TARGET_TB_PCREL) {
+ gen_a64_update_pc(s, diff);
+ tcg_gen_goto_tb(n);
+ } else {
+ tcg_gen_goto_tb(n);
+ gen_a64_update_pc(s, diff);
+ }
tcg_gen_exit_tb(s->base.tb, n);
s->base.is_jmp = DISAS_NORETURN;
} else {
@@ -1383,7 +1401,7 @@ static void disas_comp_b_imm(DisasContext *s, uint32_t insn)
{
unsigned int sf, op, rt;
int64_t diff;
- TCGLabel *label_match;
+ DisasLabel match;
TCGv_i64 tcg_cmp;
sf = extract32(insn, 31, 1);
@@ -1392,14 +1410,13 @@ static void disas_comp_b_imm(DisasContext *s, uint32_t insn)
diff = sextract32(insn, 5, 19) * 4;
tcg_cmp = read_cpu_reg(s, rt, sf);
- label_match = gen_new_label();
-
reset_btype(s);
- tcg_gen_brcondi_i64(op ? TCG_COND_NE : TCG_COND_EQ,
- tcg_cmp, 0, label_match);
+ match = gen_disas_label(s);
+ tcg_gen_brcondi_i64(op ? TCG_COND_NE : TCG_COND_EQ,
+ tcg_cmp, 0, match.label);
gen_goto_tb(s, 0, 4);
- gen_set_label(label_match);
+ set_disas_label(s, match);
gen_goto_tb(s, 1, diff);
}
@@ -1413,7 +1430,7 @@ static void disas_test_b_imm(DisasContext *s, uint32_t insn)
{
unsigned int bit_pos, op, rt;
int64_t diff;
- TCGLabel *label_match;
+ DisasLabel match;
TCGv_i64 tcg_cmp;
bit_pos = (extract32(insn, 31, 1) << 5) | extract32(insn, 19, 5);
@@ -1423,14 +1440,15 @@ static void disas_test_b_imm(DisasContext *s, uint32_t insn)
tcg_cmp = tcg_temp_new_i64();
tcg_gen_andi_i64(tcg_cmp, cpu_reg(s, rt), (1ULL << bit_pos));
- label_match = gen_new_label();
reset_btype(s);
+
+ match = gen_disas_label(s);
tcg_gen_brcondi_i64(op ? TCG_COND_NE : TCG_COND_EQ,
- tcg_cmp, 0, label_match);
+ tcg_cmp, 0, match.label);
tcg_temp_free_i64(tcg_cmp);
gen_goto_tb(s, 0, 4);
- gen_set_label(label_match);
+ set_disas_label(s, match);
gen_goto_tb(s, 1, diff);
}
@@ -1455,10 +1473,10 @@ static void disas_cond_b_imm(DisasContext *s, uint32_t insn)
reset_btype(s);
if (cond < 0x0e) {
/* genuinely conditional branches */
- TCGLabel *label_match = gen_new_label();
- arm_gen_test_cc(cond, label_match);
+ DisasLabel match = gen_disas_label(s);
+ arm_gen_test_cc(cond, match.label);
gen_goto_tb(s, 0, 4);
- gen_set_label(label_match);
+ set_disas_label(s, match);
gen_goto_tb(s, 1, diff);
} else {
/* 0xe and 0xf are both "always" conditions */
@@ -14699,7 +14717,7 @@ static void aarch64_tr_init_disas_context(DisasContextBase *dcbase,
dc->isar = &arm_cpu->isar;
dc->condjmp = 0;
-
+ dc->pc_save = dc->base.pc_first;
dc->aarch64 = true;
dc->thumb = false;
dc->sctlr_b = 0;
@@ -14781,8 +14799,12 @@ static void aarch64_tr_tb_start(DisasContextBase *db, CPUState *cpu)
static void aarch64_tr_insn_start(DisasContextBase *dcbase, CPUState *cpu)
{
DisasContext *dc = container_of(dcbase, DisasContext, base);
+ target_ulong pc_arg = dc->base.pc_next;
- tcg_gen_insn_start(dc->base.pc_next, 0, 0);
+ if (TARGET_TB_PCREL) {
+ pc_arg &= ~TARGET_PAGE_MASK;
+ }
+ tcg_gen_insn_start(pc_arg, 0, 0);
dc->insn_start = tcg_last_op();
}
diff --git a/target/arm/translate-m-nocp.c b/target/arm/translate-m-nocp.c
index 694fae7e2e..5df7d46120 100644
--- a/target/arm/translate-m-nocp.c
+++ b/target/arm/translate-m-nocp.c
@@ -140,7 +140,7 @@ static bool trans_VSCCLRM(DisasContext *s, arg_VSCCLRM *a)
tcg_gen_andi_i32(sfpa, sfpa, R_V7M_CONTROL_SFPA_MASK);
tcg_gen_or_i32(sfpa, sfpa, aspen);
arm_gen_condlabel(s);
- tcg_gen_brcondi_i32(TCG_COND_EQ, sfpa, 0, s->condlabel);
+ tcg_gen_brcondi_i32(TCG_COND_EQ, sfpa, 0, s->condlabel.label);
if (s->fp_excp_el != 0) {
gen_exception_insn_el(s, 0, EXCP_NOCP,
diff --git a/target/arm/translate.c b/target/arm/translate.c
index 5f6bd9b5b7..d1b868430e 100644
--- a/target/arm/translate.c
+++ b/target/arm/translate.c
@@ -162,7 +162,7 @@ uint64_t asimd_imm_const(uint32_t imm, int cmode, int op)
void arm_gen_condlabel(DisasContext *s)
{
if (!s->condjmp) {
- s->condlabel = gen_new_label();
+ s->condlabel = gen_disas_label(s);
s->condjmp = 1;
}
}
@@ -268,7 +268,12 @@ static target_long jmp_diff(DisasContext *s, target_long diff)
static void gen_pc_plus_diff(DisasContext *s, TCGv_i32 var, target_long diff)
{
- tcg_gen_movi_i32(var, s->pc_curr + diff);
+ assert(s->pc_save != -1);
+ if (TARGET_TB_PCREL) {
+ tcg_gen_addi_i32(var, cpu_R[15], (s->pc_curr - s->pc_save) + diff);
+ } else {
+ tcg_gen_movi_i32(var, s->pc_curr + diff);
+ }
}
/* Set a variable to the value of a CPU register. */
@@ -314,6 +319,7 @@ void store_reg(DisasContext *s, int reg, TCGv_i32 var)
*/
tcg_gen_andi_i32(var, var, s->thumb ? ~1 : ~3);
s->base.is_jmp = DISAS_JUMP;
+ s->pc_save = -1;
} else if (reg == 13 && arm_dc_feature(s, ARM_FEATURE_M)) {
/* For M-profile SP bits [1:0] are always zero */
tcg_gen_andi_i32(var, var, ~3);
@@ -779,7 +785,8 @@ void gen_set_condexec(DisasContext *s)
void gen_update_pc(DisasContext *s, target_long diff)
{
- tcg_gen_movi_i32(cpu_R[15], s->pc_curr + diff);
+ gen_pc_plus_diff(s, cpu_R[15], diff);
+ s->pc_save = s->pc_curr + diff;
}
/* Set PC and Thumb state from var. var is marked as dead. */
@@ -789,6 +796,7 @@ static inline void gen_bx(DisasContext *s, TCGv_i32 var)
tcg_gen_andi_i32(cpu_R[15], var, ~1);
tcg_gen_andi_i32(var, var, 1);
store_cpu_field(var, thumb);
+ s->pc_save = -1;
}
/*
@@ -830,7 +838,7 @@ static inline void gen_bx_excret(DisasContext *s, TCGv_i32 var)
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();
+ DisasLabel excret_label = gen_disas_label(s);
uint32_t min_magic;
if (arm_dc_feature(s, ARM_FEATURE_M_SECURITY)) {
@@ -842,14 +850,14 @@ static inline void gen_bx_excret_final_code(DisasContext *s)
}
/* Is the new PC value in the magic range indicating exception return? */
- tcg_gen_brcondi_i32(TCG_COND_GEU, cpu_R[15], min_magic, excret_label);
+ tcg_gen_brcondi_i32(TCG_COND_GEU, cpu_R[15], min_magic, excret_label.label);
/* No: end the TB as we would for a DISAS_JMP */
if (s->ss_active) {
gen_singlestep_exception(s);
} else {
tcg_gen_exit_tb(NULL, 0);
}
- gen_set_label(excret_label);
+ set_disas_label(s, excret_label);
/* Yes: this is an exception return.
* At this point in runtime env->regs[15] and env->thumb will hold
* the exception-return magic number, which do_v7m_exception_exit()
@@ -2603,11 +2611,22 @@ static void gen_goto_ptr(void)
*/
static void gen_goto_tb(DisasContext *s, int n, target_long diff)
{
- target_ulong dest = s->pc_curr + diff;
-
- if (translator_use_goto_tb(&s->base, dest)) {
- tcg_gen_goto_tb(n);
- gen_update_pc(s, diff);
+ if (translator_use_goto_tb(&s->base, s->pc_curr + diff)) {
+ /*
+ * For pcrel, the pc must always be up-to-date on entry to
+ * the linked TB, so that it can use simple additions for all
+ * further adjustments. For !pcrel, the linked TB is compiled
+ * to know its full virtual address, so we can delay the
+ * update to pc to the unlinked path. A long chain of links
+ * can thus avoid many updates to the PC.
+ */
+ if (TARGET_TB_PCREL) {
+ gen_update_pc(s, diff);
+ tcg_gen_goto_tb(n);
+ } else {
+ tcg_gen_goto_tb(n);
+ gen_update_pc(s, diff);
+ }
tcg_gen_exit_tb(s->base.tb, n);
} else {
gen_update_pc(s, diff);
@@ -5221,7 +5240,7 @@ static void gen_srs(DisasContext *s,
static void arm_skip_unless(DisasContext *s, uint32_t cond)
{
arm_gen_condlabel(s);
- arm_gen_test_cc(cond ^ 1, s->condlabel);
+ arm_gen_test_cc(cond ^ 1, s->condlabel.label);
}
@@ -8472,7 +8491,7 @@ static bool trans_WLS(DisasContext *s, arg_WLS *a)
{
/* M-profile low-overhead while-loop start */
TCGv_i32 tmp;
- TCGLabel *nextlabel;
+ DisasLabel nextlabel;
if (!dc_isar_feature(aa32_lob, s)) {
return false;
@@ -8513,8 +8532,8 @@ static bool trans_WLS(DisasContext *s, arg_WLS *a)
}
}
- nextlabel = gen_new_label();
- tcg_gen_brcondi_i32(TCG_COND_EQ, cpu_R[a->rn], 0, nextlabel);
+ nextlabel = gen_disas_label(s);
+ tcg_gen_brcondi_i32(TCG_COND_EQ, cpu_R[a->rn], 0, nextlabel.label);
tmp = load_reg(s, a->rn);
store_reg(s, 14, tmp);
if (a->size != 4) {
@@ -8535,7 +8554,7 @@ static bool trans_WLS(DisasContext *s, arg_WLS *a)
}
gen_jmp_tb(s, curr_insn_len(s), 1);
- gen_set_label(nextlabel);
+ set_disas_label(s, nextlabel);
gen_jmp(s, jmp_diff(s, a->imm));
return true;
}
@@ -8551,7 +8570,7 @@ static bool trans_LE(DisasContext *s, arg_LE *a)
* any faster.
*/
TCGv_i32 tmp;
- TCGLabel *loopend;
+ DisasLabel loopend;
bool fpu_active;
if (!dc_isar_feature(aa32_lob, s)) {
@@ -8606,12 +8625,12 @@ static bool trans_LE(DisasContext *s, arg_LE *a)
if (!a->tp && dc_isar_feature(aa32_mve, s) && fpu_active) {
/* Need to do a runtime check for LTPSIZE != 4 */
- TCGLabel *skipexc = gen_new_label();
+ DisasLabel skipexc = gen_disas_label(s);
tmp = load_cpu_field(v7m.ltpsize);
- tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, 4, skipexc);
+ tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, 4, skipexc.label);
tcg_temp_free_i32(tmp);
gen_exception_insn(s, 0, EXCP_INVSTATE, syn_uncategorized());
- gen_set_label(skipexc);
+ set_disas_label(s, skipexc);
}
if (a->f) {
@@ -8626,9 +8645,9 @@ static bool trans_LE(DisasContext *s, arg_LE *a)
* loop decrement value is 1. For LETP we need to calculate the decrement
* value from LTPSIZE.
*/
- loopend = gen_new_label();
+ loopend = gen_disas_label(s);
if (!a->tp) {
- tcg_gen_brcondi_i32(TCG_COND_LEU, cpu_R[14], 1, loopend);
+ tcg_gen_brcondi_i32(TCG_COND_LEU, cpu_R[14], 1, loopend.label);
tcg_gen_addi_i32(cpu_R[14], cpu_R[14], -1);
} else {
/*
@@ -8641,7 +8660,7 @@ static bool trans_LE(DisasContext *s, arg_LE *a)
tcg_gen_shl_i32(decr, tcg_constant_i32(1), decr);
tcg_temp_free_i32(ltpsize);
- tcg_gen_brcond_i32(TCG_COND_LEU, cpu_R[14], decr, loopend);
+ tcg_gen_brcond_i32(TCG_COND_LEU, cpu_R[14], decr, loopend.label);
tcg_gen_sub_i32(cpu_R[14], cpu_R[14], decr);
tcg_temp_free_i32(decr);
@@ -8649,7 +8668,7 @@ static bool trans_LE(DisasContext *s, arg_LE *a)
/* Jump back to the loop start */
gen_jmp(s, jmp_diff(s, -a->imm));
- gen_set_label(loopend);
+ set_disas_label(s, loopend);
if (a->tp) {
/* Exits from tail-pred loops must reset LTPSIZE to 4 */
store_cpu_field(tcg_constant_i32(4), v7m.ltpsize);
@@ -8753,7 +8772,7 @@ static bool trans_CBZ(DisasContext *s, arg_CBZ *a)
arm_gen_condlabel(s);
tcg_gen_brcondi_i32(a->nz ? TCG_COND_EQ : TCG_COND_NE,
- tmp, 0, s->condlabel);
+ tmp, 0, s->condlabel.label);
tcg_temp_free_i32(tmp);
gen_jmp(s, jmp_diff(s, a->imm));
return true;
@@ -9319,7 +9338,7 @@ static void arm_tr_init_disas_context(DisasContextBase *dcbase, CPUState *cs)
dc->isar = &cpu->isar;
dc->condjmp = 0;
-
+ dc->pc_save = dc->base.pc_first;
dc->aarch64 = false;
dc->thumb = EX_TBFLAG_AM32(tb_flags, THUMB);
dc->be_data = EX_TBFLAG_ANY(tb_flags, BE_DATA) ? MO_BE : MO_LE;
@@ -9337,7 +9356,6 @@ static void arm_tr_init_disas_context(DisasContextBase *dcbase, CPUState *cs)
*/
dc->eci = dc->condexec_mask = dc->condexec_cond = 0;
dc->eci_handled = false;
- dc->insn_eci_rewind = NULL;
if (condexec & 0xf) {
dc->condexec_mask = (condexec & 0xf) << 1;
dc->condexec_cond = condexec >> 4;
@@ -9473,13 +9491,17 @@ static void arm_tr_insn_start(DisasContextBase *dcbase, CPUState *cpu)
* fields here.
*/
uint32_t condexec_bits;
+ target_ulong pc_arg = dc->base.pc_next;
+ if (TARGET_TB_PCREL) {
+ pc_arg &= ~TARGET_PAGE_MASK;
+ }
if (dc->eci) {
condexec_bits = dc->eci << 4;
} else {
condexec_bits = (dc->condexec_cond << 4) | (dc->condexec_mask >> 1);
}
- tcg_gen_insn_start(dc->base.pc_next, condexec_bits, 0);
+ tcg_gen_insn_start(pc_arg, condexec_bits, 0);
dc->insn_start = tcg_last_op();
}
@@ -9522,8 +9544,11 @@ static bool arm_check_ss_active(DisasContext *dc)
static void arm_post_translate_insn(DisasContext *dc)
{
- if (dc->condjmp && !dc->base.is_jmp) {
- gen_set_label(dc->condlabel);
+ if (dc->condjmp && dc->base.is_jmp == DISAS_NEXT) {
+ if (dc->pc_save != dc->condlabel.pc_save) {
+ gen_update_pc(dc, dc->condlabel.pc_save - dc->pc_save);
+ }
+ gen_set_label(dc->condlabel.label);
dc->condjmp = 0;
}
translator_loop_temp_check(&dc->base);
@@ -9626,6 +9651,9 @@ static void thumb_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu)
uint32_t pc = dc->base.pc_next;
uint32_t insn;
bool is_16bit;
+ /* TCG op to rewind to if this turns out to be an invalid ECI state */
+ TCGOp *insn_eci_rewind = NULL;
+ target_ulong insn_eci_pc_save = -1;
/* Misaligned thumb PC is architecturally impossible. */
assert((dc->base.pc_next & 1) == 0);
@@ -9687,7 +9715,8 @@ static void thumb_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu)
* insn" case. We will rewind to the marker (ie throwing away
* all the generated code) and instead emit "take exception".
*/
- dc->insn_eci_rewind = tcg_last_op();
+ insn_eci_rewind = tcg_last_op();
+ insn_eci_pc_save = dc->pc_save;
}
if (dc->condexec_mask && !thumb_insn_is_unconditional(dc, insn)) {
@@ -9723,7 +9752,8 @@ static void thumb_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu)
* Insn wasn't valid for ECI/ICI at all: undo what we
* just generated and instead emit an exception
*/
- tcg_remove_ops_after(dc->insn_eci_rewind);
+ tcg_remove_ops_after(insn_eci_rewind);
+ dc->pc_save = insn_eci_pc_save;
dc->condjmp = 0;
gen_exception_insn(dc, 0, EXCP_INVSTATE, syn_uncategorized());
}
@@ -9852,7 +9882,7 @@ static void arm_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu)
if (dc->condjmp) {
/* "Condition failed" instruction codepath for the branch/trap insn */
- gen_set_label(dc->condlabel);
+ set_disas_label(dc, dc->condlabel);
gen_set_condexec(dc);
if (unlikely(dc->ss_active)) {
gen_update_pc(dc, curr_insn_len(dc));
@@ -9914,11 +9944,19 @@ void restore_state_to_opc(CPUARMState *env, TranslationBlock *tb,
target_ulong *data)
{
if (is_a64(env)) {
- env->pc = data[0];
+ if (TARGET_TB_PCREL) {
+ env->pc = (env->pc & TARGET_PAGE_MASK) | data[0];
+ } else {
+ env->pc = data[0];
+ }
env->condexec_bits = 0;
env->exception.syndrome = data[2] << ARM_INSN_START_WORD2_SHIFT;
} else {
- env->regs[15] = data[0];
+ if (TARGET_TB_PCREL) {
+ env->regs[15] = (env->regs[15] & TARGET_PAGE_MASK) | data[0];
+ } else {
+ env->regs[15] = data[0];
+ }
env->condexec_bits = data[1];
env->exception.syndrome = data[2] << ARM_INSN_START_WORD2_SHIFT;
}
--
2.34.1
^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [PATCH v8 0/9] target/arm: pc-relative translation blocks
2022-10-20 3:06 [PATCH v8 0/9] target/arm: pc-relative translation blocks Richard Henderson
` (8 preceding siblings ...)
2022-10-20 3:06 ` [PATCH v8 9/9] target/arm: Enable TARGET_TB_PCREL Richard Henderson
@ 2022-10-20 10:29 ` Peter Maydell
9 siblings, 0 replies; 11+ messages in thread
From: Peter Maydell @ 2022-10-20 10:29 UTC (permalink / raw)
To: Richard Henderson; +Cc: qemu-devel, qemu-arm
On Thu, 20 Oct 2022 at 04:14, Richard Henderson
<richard.henderson@linaro.org> wrote:
>
> This is the Arm specific changes required to reduce the
> amount of translation for address space randomization.
>
> Changes for v8:
> * Add a comment to both gen_goto_tb concerning the ordering
> of the pc update vs the goto_tb opcode.
Applied to target-arm.next, thanks.
-- PMM
^ permalink raw reply [flat|nested] 11+ messages in thread
end of thread, other threads:[~2022-10-20 11:11 UTC | newest]
Thread overview: 11+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2022-10-20 3:06 [PATCH v8 0/9] target/arm: pc-relative translation blocks Richard Henderson
2022-10-20 3:06 ` [PATCH v8 1/9] target/arm: Introduce curr_insn_len Richard Henderson
2022-10-20 3:06 ` [PATCH v8 2/9] target/arm: Change gen_goto_tb to work on displacements Richard Henderson
2022-10-20 3:06 ` [PATCH v8 3/9] target/arm: Change gen_*set_pc_im to gen_*update_pc Richard Henderson
2022-10-20 3:06 ` [PATCH v8 4/9] target/arm: Change gen_exception_insn* to work on displacements Richard Henderson
2022-10-20 3:06 ` [PATCH v8 5/9] target/arm: Remove gen_exception_internal_insn pc argument Richard Henderson
2022-10-20 3:06 ` [PATCH v8 6/9] target/arm: Change gen_jmp* to work on displacements Richard Henderson
2022-10-20 3:06 ` [PATCH v8 7/9] target/arm: Introduce gen_pc_plus_diff for aarch64 Richard Henderson
2022-10-20 3:06 ` [PATCH v8 8/9] target/arm: Introduce gen_pc_plus_diff for aarch32 Richard Henderson
2022-10-20 3:06 ` [PATCH v8 9/9] target/arm: Enable TARGET_TB_PCREL Richard Henderson
2022-10-20 10:29 ` [PATCH v8 0/9] target/arm: pc-relative translation blocks Peter Maydell
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).