* [PATCH v5 00/76] target/arm: Implement FEAT_GCS
@ 2025-09-22 18:48 Richard Henderson
2025-09-22 18:48 ` [PATCH v5 01/76] target/arm: Add isar feature test for FEAT_S1PIE, FEAT_S2PIE Richard Henderson
` (76 more replies)
0 siblings, 77 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm
Based-on: 20250916142238.664316-1-richard.henderson@linaro.org
("[PATCH v2 00/36] target/arm: Reorg VHE redirection")
Changes from v4:
- Master now includes FEAT_ATS1A.
- Rebase on VHE reorg
- Trivial adjustments to S1PIE and GCS regs for REG12 redirect.
- Simplifies EXLOCK access test.
- Migration fixes based on review.
Note that tests/functional/aarch64/test_rme_* will fail.
The TF-A images need updating again for at least FEAT_S1PIE
(EL1 accessing PIRE0_EL1 trapping to EL3 is the fatal fault).
I'm toying with updating the QEMU build to match the FVP build,
so that all features are runtime enabled.
r~
Richard Henderson (76):
target/arm: Add isar feature test for FEAT_S1PIE, FEAT_S2PIE
target/arm: Enable TCR2_ELx.PIE
target/arm: Implement PIR_ELx, PIRE0_ELx, S2PIR_EL2 registers
target/arm: Force HPD for stage2 translations
target/arm: Cache NV1 early in get_phys_addr_lpae
target/arm: Populate PIE in aa64_va_parameters
target/arm: Implement get_S1prot_indirect
target/arm: Implement get_S2prot_indirect
target/arm: Expand CPUARMState.exception.syndrome to 64 bits
target/arm: Expand syndrome parameter to raise_exception*
target/arm: Implement dirtybit check for PIE
target/arm: Enable FEAT_S1PIE and FEAT_S2PIE on -cpu max
include/hw/core/cpu: Introduce MMUIdxMap
include/hw/core/cpu: Introduce cpu_tlb_fast
include/hw/core/cpu: Invert the indexing into CPUTLBDescFast
target/hppa: Adjust mmu indexes to begin with 0
include/exec/memopidx: Adjust for 32 mmu indexes
include/hw/core/cpu: Widen MMUIdxMap
target/arm: Split out mmuidx.h from cpu.h
target/arm: Convert arm_mmu_idx_to_el from switch to table
target/arm: Remove unused env argument from regime_el
target/arm: Convert regime_el from switch to table
target/arm: Convert regime_has_2_ranges from switch to table
target/arm: Remove unused env argument from regime_is_pan
target/arm: Convert regime_is_pan from switch to table
target/arm: Remove unused env argument from regime_is_user
target/arm: Convert regime_is_user from switch to table
target/arm: Convert arm_mmu_idx_is_stage1_of_2 from switch to table
target/arm: Convert regime_is_stage2 to table
target/arm: Introduce mmu indexes for GCS
target/arm: Introduce regime_to_gcs
target/arm: Support page protections for GCS mmu indexes
target/arm: Implement gcs bit for data abort
target/arm: Add GCS cpregs
target/arm: Add GCS enable and trap levels to DisasContext
target/arm: Implement FEAT_CHK
target/arm: Make helper_exception_return system-only
target/arm: Export cpsr_{read_for,write_from}_spsr_elx
target/arm: Expand pstate to 64 bits
target/arm: Add syndrome data for EC_GCS
target/arm: Add arm_hcr_el2_nvx_eff
target/arm: Use arm_hcr_el2_nvx_eff in access_nv1
target/arm: Split out access_nv1_with_nvx
target/arm: Implement EXLOCKException for ELR_ELx and SPSR_ELx
target/arm: Split {full,core}_a64_user_mem_index
target/arm: Introduce delay_exception{_el}
target/arm: Emit HSTR trap exception out of line
target/arm: Emit v7m LTPSIZE exception out of line
target/arm: Implement GCSSTR, GCSSTTR
target/arm: Implement GCSB
target/arm: Implement GCSPUSHM
target/arm: Implement GCSPOPM
target/arm: Implement GCSPUSHX
target/arm: Implement GCSPOPX
target/arm: Implement GCSPOPCX
target/arm: Implement GCSSS1
target/arm: Implement GCSSS2
target/arm: Add gcs record for BL
target/arm: Add gcs record for BLR
target/arm: Add gcs record for BLR with PAuth
target/arm: Load gcs record for RET
target/arm: Load gcs record for RET with PAuth
target/arm: Copy EXLOCKEn to EXLOCK on exception to the same EL
target/arm: Implement EXLOCK check during exception return
target/arm: Enable FEAT_GCS with -cpu max
linux-user/aarch64: Implement prctls for GCS
linux-user/aarch64: Allocate new gcs stack on clone
linux-user/aarch64: Release gcs stack on thread exit
linux-user/aarch64: Implement map_shadow_stack syscall
target/arm: Enable GCSPR_EL0 for read in user-mode
linux-user/aarch64: Inject SIGSEGV for GCS faults
linux-user/aarch64: Generate GCS signal records
linux-user/aarch64: Enable GCS in HWCAP
tests/tcg/aarch64: Add gcsstr
tests/tcg/aarch64: Add gcspushm
tests/tcg/aarch64: Add gcsss
include/exec/cputlb.h | 32 +--
include/exec/memopidx.h | 9 +-
include/hw/core/cpu.h | 25 +-
linux-user/aarch64/gcs-internal.h | 38 +++
linux-user/aarch64/target_prctl.h | 96 +++++++
linux-user/aarch64/target_signal.h | 1 +
linux-user/qemu.h | 5 +
target/arm/cpregs.h | 22 ++
target/arm/cpu-features.h | 15 +
target/arm/cpu.h | 243 ++--------------
target/arm/internals.h | 148 ++--------
target/arm/mmuidx-internal.h | 113 ++++++++
target/arm/mmuidx.h | 241 ++++++++++++++++
target/arm/syndrome.h | 35 +++
target/arm/tcg/helper-a64.h | 5 +-
target/arm/tcg/translate.h | 46 ++-
target/hppa/cpu.h | 28 +-
tests/tcg/aarch64/gcs.h | 80 ++++++
accel/tcg/cputlb.c | 49 ++--
linux-user/aarch64/cpu_loop.c | 5 +
linux-user/aarch64/elfload.c | 1 +
linux-user/aarch64/signal.c | 138 ++++++++-
linux-user/syscall.c | 114 ++++++++
target/arm/cpregs-gcs.c | 156 ++++++++++
target/arm/cpu.c | 17 +-
target/arm/gdbstub64.c | 2 +
target/arm/helper.c | 283 +++++++++++++++----
target/arm/machine.c | 113 +++++++-
target/arm/mmuidx.c | 66 +++++
target/arm/ptw.c | 373 ++++++++++++++++++------
target/arm/tcg-stubs.c | 2 +-
target/arm/tcg/cpu64.c | 3 +
target/arm/tcg/helper-a64.c | 35 +--
target/arm/tcg/hflags.c | 38 +++
target/arm/tcg/mte_helper.c | 2 +-
target/arm/tcg/op_helper.c | 11 +-
target/arm/tcg/tlb-insns.c | 47 +++-
target/arm/tcg/tlb_helper.c | 18 +-
target/arm/tcg/translate-a64.c | 438 ++++++++++++++++++++++++++---
target/arm/tcg/translate.c | 78 +++--
tcg/tcg.c | 3 +-
tests/tcg/aarch64/gcspushm.c | 71 +++++
tests/tcg/aarch64/gcsss.c | 74 +++++
tests/tcg/aarch64/gcsstr.c | 48 ++++
docs/system/arm/emulation.rst | 4 +
target/arm/meson.build | 9 +-
target/arm/tcg/a64.decode | 5 +
tcg/aarch64/tcg-target.c.inc | 2 +-
tcg/arm/tcg-target.c.inc | 2 +-
tests/tcg/aarch64/Makefile.target | 5 +
50 files changed, 2734 insertions(+), 660 deletions(-)
create mode 100644 linux-user/aarch64/gcs-internal.h
create mode 100644 target/arm/mmuidx-internal.h
create mode 100644 target/arm/mmuidx.h
create mode 100644 tests/tcg/aarch64/gcs.h
create mode 100644 target/arm/cpregs-gcs.c
create mode 100644 target/arm/mmuidx.c
create mode 100644 tests/tcg/aarch64/gcspushm.c
create mode 100644 tests/tcg/aarch64/gcsss.c
create mode 100644 tests/tcg/aarch64/gcsstr.c
--
2.43.0
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH v5 01/76] target/arm: Add isar feature test for FEAT_S1PIE, FEAT_S2PIE
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 02/76] target/arm: Enable TCR2_ELx.PIE Richard Henderson
` (75 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/cpu-features.h | 10 ++++++++++
1 file changed, 10 insertions(+)
diff --git a/target/arm/cpu-features.h b/target/arm/cpu-features.h
index 512eeaf551..30226814bb 100644
--- a/target/arm/cpu-features.h
+++ b/target/arm/cpu-features.h
@@ -929,6 +929,16 @@ static inline bool isar_feature_aa64_sctlr2(const ARMISARegisters *id)
return FIELD_EX64_IDREG(id, ID_AA64MMFR3, SCTLRX) != 0;
}
+static inline bool isar_feature_aa64_s1pie(const ARMISARegisters *id)
+{
+ return FIELD_EX64_IDREG(id, ID_AA64MMFR3, S1PIE) != 0;
+}
+
+static inline bool isar_feature_aa64_s2pie(const ARMISARegisters *id)
+{
+ return FIELD_EX64_IDREG(id, ID_AA64MMFR3, S2PIE) != 0;
+}
+
static inline bool isar_feature_aa64_pmuv3p1(const ARMISARegisters *id)
{
return FIELD_EX64_IDREG(id, ID_AA64DFR0, PMUVER) >= 4 &&
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 02/76] target/arm: Enable TCR2_ELx.PIE
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
2025-09-22 18:48 ` [PATCH v5 01/76] target/arm: Add isar feature test for FEAT_S1PIE, FEAT_S2PIE Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 03/76] target/arm: Implement PIR_ELx, PIRE0_ELx, S2PIR_EL2 registers Richard Henderson
` (74 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/helper.c | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/target/arm/helper.c b/target/arm/helper.c
index c60da6013d..5c3805eaff 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -5891,8 +5891,12 @@ static CPAccessResult tcr2_el1_access(CPUARMState *env, const ARMCPRegInfo *ri,
static void tcr2_el1_write(CPUARMState *env, const ARMCPRegInfo *ri,
uint64_t value)
{
+ ARMCPU *cpu = env_archcpu(env);
uint64_t valid_mask = 0;
+ if (cpu_isar_feature(aa64_s1pie, cpu)) {
+ valid_mask |= TCR2_PIE;
+ }
value &= valid_mask;
raw_write(env, ri, value);
}
@@ -5900,8 +5904,12 @@ static void tcr2_el1_write(CPUARMState *env, const ARMCPRegInfo *ri,
static void tcr2_el2_write(CPUARMState *env, const ARMCPRegInfo *ri,
uint64_t value)
{
+ ARMCPU *cpu = env_archcpu(env);
uint64_t valid_mask = 0;
+ if (cpu_isar_feature(aa64_s1pie, cpu)) {
+ valid_mask |= TCR2_PIE;
+ }
value &= valid_mask;
raw_write(env, ri, value);
}
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 03/76] target/arm: Implement PIR_ELx, PIRE0_ELx, S2PIR_EL2 registers
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
2025-09-22 18:48 ` [PATCH v5 01/76] target/arm: Add isar feature test for FEAT_S1PIE, FEAT_S2PIE Richard Henderson
2025-09-22 18:48 ` [PATCH v5 02/76] target/arm: Enable TCR2_ELx.PIE Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 04/76] target/arm: Force HPD for stage2 translations Richard Henderson
` (73 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/cpregs.h | 2 ++
target/arm/cpu.h | 4 +++
target/arm/cpu.c | 4 +++
target/arm/helper.c | 69 +++++++++++++++++++++++++++++++++++++++++++++
4 files changed, 79 insertions(+)
diff --git a/target/arm/cpregs.h b/target/arm/cpregs.h
index 57fde5f57a..f48c4df30f 100644
--- a/target/arm/cpregs.h
+++ b/target/arm/cpregs.h
@@ -781,6 +781,8 @@ typedef enum FGTBit {
DO_BIT(HFGRTR, ERRIDR_EL1),
DO_REV_BIT(HFGRTR, NSMPRI_EL1),
DO_REV_BIT(HFGRTR, NTPIDR2_EL0),
+ DO_REV_BIT(HFGRTR, NPIRE0_EL1),
+ DO_REV_BIT(HFGRTR, NPIR_EL1),
/* Trap bits in HDFGRTR_EL2 / HDFGWTR_EL2, starting from bit 0. */
DO_BIT(HDFGRTR, DBGBCRN_EL1),
diff --git a/target/arm/cpu.h b/target/arm/cpu.h
index d5534e3580..d17252f734 100644
--- a/target/arm/cpu.h
+++ b/target/arm/cpu.h
@@ -368,6 +368,9 @@ typedef struct CPUArchState {
uint64_t tcr2_el[3];
uint64_t vtcr_el2; /* Virtualization Translation Control. */
uint64_t vstcr_el2; /* Secure Virtualization Translation Control. */
+ uint64_t pir_el[4]; /* PIRE0_EL1, PIR_EL1, PIR_EL2, PIR_EL3 */
+ uint64_t pire0_el2;
+ uint64_t s2pir_el2;
uint32_t c2_data; /* MPU data cacheable bits. */
uint32_t c2_insn; /* MPU instruction cacheable bits. */
union { /* MMU domain access control register
@@ -1720,6 +1723,7 @@ static inline void xpsr_write(CPUARMState *env, uint32_t val, uint32_t mask)
#define SCR_ENTP2 (1ULL << 41)
#define SCR_TCR2EN (1ULL << 43)
#define SCR_SCTLR2EN (1ULL << 44)
+#define SCR_PIEN (1ULL << 45)
#define SCR_GPF (1ULL << 48)
#define SCR_NSE (1ULL << 62)
diff --git a/target/arm/cpu.c b/target/arm/cpu.c
index 5aee4e6a91..d4bc2e6a92 100644
--- a/target/arm/cpu.c
+++ b/target/arm/cpu.c
@@ -635,6 +635,10 @@ void arm_emulate_firmware_reset(CPUState *cpustate, int target_el)
if (cpu_isar_feature(aa64_sctlr2, cpu)) {
env->cp15.scr_el3 |= SCR_SCTLR2EN;
}
+ if (cpu_isar_feature(aa64_s1pie, cpu) ||
+ cpu_isar_feature(aa64_s2pie, cpu)) {
+ env->cp15.scr_el3 |= SCR_PIEN;
+ }
}
if (target_el == 2) {
diff --git a/target/arm/helper.c b/target/arm/helper.c
index 5c3805eaff..c0b499946c 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -770,6 +770,10 @@ static void scr_write(CPUARMState *env, const ARMCPRegInfo *ri, uint64_t value)
if (cpu_isar_feature(aa64_sctlr2, cpu)) {
valid_mask |= SCR_SCTLR2EN;
}
+ if (cpu_isar_feature(aa64_s1pie, cpu) ||
+ cpu_isar_feature(aa64_s2pie, cpu)) {
+ valid_mask |= SCR_PIEN;
+ }
} else {
valid_mask &= ~(SCR_RW | SCR_ST);
if (cpu_isar_feature(aa32_ras, cpu)) {
@@ -5930,6 +5934,64 @@ static const ARMCPRegInfo tcr2_reginfo[] = {
.fieldoffset = offsetof(CPUARMState, cp15.tcr2_el[2]) },
};
+static CPAccessResult pien_access(CPUARMState *env, const ARMCPRegInfo *ri,
+ bool isread)
+{
+ if (arm_feature(env, ARM_FEATURE_EL3)
+ && !(env->cp15.scr_el3 & SCR_PIEN)
+ && arm_current_el(env) < 3) {
+ return CP_ACCESS_TRAP_EL3;
+ }
+ return CP_ACCESS_OK;
+}
+
+static CPAccessResult pien_el1_access(CPUARMState *env, const ARMCPRegInfo *ri,
+ bool isread)
+{
+ CPAccessResult ret = access_tvm_trvm(env, ri, isread);
+ if (ret == CP_ACCESS_OK) {
+ ret = pien_access(env, ri, isread);
+ }
+ return ret;
+}
+
+static const ARMCPRegInfo s1pie_reginfo[] = {
+ { .name = "PIR_EL1", .state = ARM_CP_STATE_AA64,
+ .opc0 = 3, .opc1 = 0, .opc2 = 3, .crn = 10, .crm = 2,
+ .access = PL1_RW, .accessfn = pien_el1_access,
+ .fgt = FGT_NPIR_EL1, .nv2_redirect_offset = 0x2a0 | NV2_REDIR_NV1,
+ .vhe_redir_to_el2 = ENCODE_AA64_CP_REG(3, 4, 10, 2, 3),
+ .vhe_redir_to_el01 = ENCODE_AA64_CP_REG(3, 5, 10, 2, 3),
+ .fieldoffset = offsetof(CPUARMState, cp15.pir_el[1]) },
+ { .name = "PIR_EL2", .state = ARM_CP_STATE_AA64,
+ .opc0 = 3, .opc1 = 4, .opc2 = 3, .crn = 10, .crm = 2,
+ .access = PL2_RW, .accessfn = pien_access,
+ .fieldoffset = offsetof(CPUARMState, cp15.pir_el[2]) },
+ { .name = "PIR_EL3", .state = ARM_CP_STATE_AA64,
+ .opc0 = 3, .opc1 = 6, .opc2 = 3, .crn = 10, .crm = 2,
+ .access = PL3_RW,
+ .fieldoffset = offsetof(CPUARMState, cp15.pir_el[3]) },
+ { .name = "PIRE0_EL1", .state = ARM_CP_STATE_AA64,
+ .opc0 = 3, .opc1 = 0, .opc2 = 2, .crn = 10, .crm = 2,
+ .access = PL1_RW, .accessfn = pien_el1_access,
+ .fgt = FGT_NPIRE0_EL1, .nv2_redirect_offset = 0x290 | NV2_REDIR_NV1,
+ .vhe_redir_to_el2 = ENCODE_AA64_CP_REG(3, 4, 10, 2, 2),
+ .vhe_redir_to_el01 = ENCODE_AA64_CP_REG(3, 5, 10, 2, 2),
+ .fieldoffset = offsetof(CPUARMState, cp15.pir_el[0]) },
+ { .name = "PIRE0_EL2", .state = ARM_CP_STATE_AA64,
+ .opc0 = 3, .opc1 = 4, .opc2 = 2, .crn = 10, .crm = 2,
+ .access = PL2_RW, .accessfn = pien_access,
+ .fieldoffset = offsetof(CPUARMState, cp15.pire0_el2) },
+};
+
+static const ARMCPRegInfo s2pie_reginfo[] = {
+ { .name = "S2PIR_EL2", .state = ARM_CP_STATE_AA64,
+ .opc0 = 3, .opc1 = 4, .opc2 = 5, .crn = 10, .crm = 2,
+ .access = PL2_RW, .accessfn = pien_access,
+ .nv2_redirect_offset = 0x2b0,
+ .fieldoffset = offsetof(CPUARMState, cp15.s2pir_el2) },
+};
+
void register_cp_regs_for_features(ARMCPU *cpu)
{
/* Register all the coprocessor registers based on feature bits */
@@ -7160,6 +7222,13 @@ void register_cp_regs_for_features(ARMCPU *cpu)
define_arm_cp_regs(cpu, tcr2_reginfo);
}
+ if (cpu_isar_feature(aa64_s1pie, cpu)) {
+ define_arm_cp_regs(cpu, s1pie_reginfo);
+ }
+ if (cpu_isar_feature(aa64_s2pie, cpu)) {
+ define_arm_cp_regs(cpu, s2pie_reginfo);
+ }
+
if (cpu_isar_feature(any_predinv, cpu)) {
define_arm_cp_regs(cpu, predinv_reginfo);
}
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 04/76] target/arm: Force HPD for stage2 translations
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (2 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 03/76] target/arm: Implement PIR_ELx, PIRE0_ELx, S2PIR_EL2 registers Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 05/76] target/arm: Cache NV1 early in get_phys_addr_lpae Richard Henderson
` (72 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Stage2 translations do not have hierarchial permissions.
Setting HPD means we can eliminate an extra check against
regime_is_stage2.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/helper.c | 7 +++++--
target/arm/ptw.c | 24 +++++++++++++-----------
2 files changed, 18 insertions(+), 13 deletions(-)
diff --git a/target/arm/helper.c b/target/arm/helper.c
index c0b499946c..f64543dc78 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -9383,8 +9383,11 @@ ARMVAParameters aa64_va_parameters(CPUARMState *env, uint64_t va,
tsz = extract32(tcr, 0, 6);
gran = tg0_to_gran_size(extract32(tcr, 14, 2));
if (stage2) {
- /* VTCR_EL2 */
- hpd = false;
+ /*
+ * Stage2 does not have hierarchical permissions.
+ * Thus disabling them makes things easier during ptw.
+ */
+ hpd = true;
} else {
hpd = extract32(tcr, 24, 1);
}
diff --git a/target/arm/ptw.c b/target/arm/ptw.c
index 6344971fa6..ebb260d54d 100644
--- a/target/arm/ptw.c
+++ b/target/arm/ptw.c
@@ -1471,8 +1471,12 @@ static ARMVAParameters aa32_va_parameters(CPUARMState *env, uint32_t va,
}
tsz = sextract32(tcr, 0, 4) + 8;
select = 0;
- hpd = false;
epd = false;
+ /*
+ * Stage2 does not have hierarchical permissions.
+ * Thus disabling them makes things easier during ptw.
+ */
+ hpd = true;
} else if (el == 2) {
/* HTCR */
tsz = extract32(tcr, 0, 3);
@@ -1978,16 +1982,14 @@ static bool get_phys_addr_lpae(CPUARMState *env, S1Translate *ptw,
* except NSTable (which we have already handled).
*/
attrs = new_descriptor & (MAKE_64BIT_MASK(2, 10) | MAKE_64BIT_MASK(50, 14));
- if (!regime_is_stage2(mmu_idx)) {
- if (!param.hpd) {
- attrs |= extract64(tableattrs, 0, 2) << 53; /* XN, PXN */
- /*
- * The sense of AP[1] vs APTable[0] is reversed, as APTable[0] == 1
- * means "force PL1 access only", which means forcing AP[1] to 0.
- */
- attrs &= ~(extract64(tableattrs, 2, 1) << 6); /* !APT[0] => AP[1] */
- attrs |= extract32(tableattrs, 3, 1) << 7; /* APT[1] => AP[2] */
- }
+ if (!param.hpd) {
+ attrs |= extract64(tableattrs, 0, 2) << 53; /* XN, PXN */
+ /*
+ * The sense of AP[1] vs APTable[0] is reversed, as APTable[0] == 1
+ * means "force PL1 access only", which means forcing AP[1] to 0.
+ */
+ attrs &= ~(extract64(tableattrs, 2, 1) << 6); /* !APT[0] => AP[1] */
+ attrs |= extract32(tableattrs, 3, 1) << 7; /* APT[1] => AP[2] */
}
ap = extract32(attrs, 6, 2);
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 05/76] target/arm: Cache NV1 early in get_phys_addr_lpae
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (3 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 04/76] target/arm: Force HPD for stage2 translations Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 06/76] target/arm: Populate PIE in aa64_va_parameters Richard Henderson
` (71 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
We were not using the correct security space in the existing call
to nv_nv1_enabled, because it may have been modified for NSTable.
Cache it early, as we will shortly need it elsewhere as well.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/ptw.c | 20 +++++++++++++-------
1 file changed, 13 insertions(+), 7 deletions(-)
diff --git a/target/arm/ptw.c b/target/arm/ptw.c
index ebb260d54d..d6d704e2a9 100644
--- a/target/arm/ptw.c
+++ b/target/arm/ptw.c
@@ -76,6 +76,8 @@ typedef struct S1Translate {
* may be suppressed for debug or AT insns.
*/
uint8_t in_prot_check;
+ /* Cached EffectiveHCR_EL2_NVx() bit */
+ bool in_nv1;
bool out_rw;
bool out_be;
ARMSecuritySpace out_space;
@@ -1641,12 +1643,6 @@ static bool lpae_block_desc_valid(ARMCPU *cpu, bool ds,
}
}
-static bool nv_nv1_enabled(CPUARMState *env, S1Translate *ptw)
-{
- uint64_t hcr = arm_hcr_el2_eff_secstate(env, ptw->in_space);
- return (hcr & (HCR_NV | HCR_NV1)) == (HCR_NV | HCR_NV1);
-}
-
/**
* get_phys_addr_lpae: perform one stage of page table walk, LPAE format
*
@@ -1698,6 +1694,16 @@ static bool get_phys_addr_lpae(CPUARMState *env, S1Translate *ptw,
!arm_el_is_aa64(env, 1));
level = 0;
+ /*
+ * Cache NV1 before we adjust ptw->in_space for NSTable.
+ * Note that this is only relevant for EL1&0, and that
+ * computing it would assert for ARMSS_Root.
+ */
+ if (el == 1) {
+ uint64_t hcr = arm_hcr_el2_eff_secstate(env, ptw->in_space);
+ ptw->in_nv1 = (hcr & (HCR_NV | HCR_NV1)) == (HCR_NV | HCR_NV1);
+ }
+
/*
* If TxSZ is programmed to a value larger than the maximum,
* or smaller than the effective minimum, it is IMPLEMENTATION
@@ -2073,7 +2079,7 @@ static bool get_phys_addr_lpae(CPUARMState *env, S1Translate *ptw,
xn = extract64(attrs, 54, 1);
pxn = extract64(attrs, 53, 1);
- if (el == 1 && nv_nv1_enabled(env, ptw)) {
+ if (el == 1 && ptw->in_nv1) {
/*
* With FEAT_NV, when HCR_EL2.{NV,NV1} == {1,1}, the block/page
* descriptor bit 54 holds PXN, 53 is RES0, and the effective value
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 06/76] target/arm: Populate PIE in aa64_va_parameters
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (4 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 05/76] target/arm: Cache NV1 early in get_phys_addr_lpae Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 07/76] target/arm: Implement get_S1prot_indirect Richard Henderson
` (70 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Select the PIE bit for the translation regime.
With PIE, the PTE layout changes, forcing HPD.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/internals.h | 3 ++-
target/arm/helper.c | 19 ++++++++++++++++++-
2 files changed, 20 insertions(+), 2 deletions(-)
diff --git a/target/arm/internals.h b/target/arm/internals.h
index b1263e4750..b7a1b69561 100644
--- a/target/arm/internals.h
+++ b/target/arm/internals.h
@@ -1486,7 +1486,7 @@ static inline int arm_granule_bits(ARMGranuleSize gran)
/*
* Parameters of a given virtual address, as extracted from the
- * translation control register (TCR) for a given regime.
+ * translation controls for a given regime.
*/
typedef struct ARMVAParameters {
unsigned tsz : 8;
@@ -1501,6 +1501,7 @@ typedef struct ARMVAParameters {
bool ha : 1;
bool hd : 1;
ARMGranuleSize gran : 2;
+ bool pie : 1;
} ARMVAParameters;
/**
diff --git a/target/arm/helper.c b/target/arm/helper.c
index f64543dc78..1111a16330 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -9372,11 +9372,12 @@ ARMVAParameters aa64_va_parameters(CPUARMState *env, uint64_t va,
bool el1_is_aa32)
{
uint64_t tcr = regime_tcr(env, mmu_idx);
- bool epd, hpd, tsz_oob, ds, ha, hd;
+ bool epd, hpd, tsz_oob, ds, ha, hd, pie = false;
int select, tsz, tbi, max_tsz, min_tsz, ps, sh;
ARMGranuleSize gran;
ARMCPU *cpu = env_archcpu(env);
bool stage2 = regime_is_stage2(mmu_idx);
+ int r_el = regime_el(env, mmu_idx);
if (!regime_has_2_ranges(mmu_idx)) {
select = 0;
@@ -9388,8 +9389,17 @@ ARMVAParameters aa64_va_parameters(CPUARMState *env, uint64_t va,
* Thus disabling them makes things easier during ptw.
*/
hpd = true;
+ pie = extract64(tcr, 36, 1) && cpu_isar_feature(aa64_s2pie, cpu);
} else {
hpd = extract32(tcr, 24, 1);
+ if (r_el == 3) {
+ pie = (extract64(tcr, 35, 1)
+ && cpu_isar_feature(aa64_s1pie, cpu));
+ } else {
+ pie = ((env->cp15.tcr2_el[2] & TCR2_PIE)
+ && (!arm_feature(env, ARM_FEATURE_EL3)
+ || (env->cp15.scr_el3 & SCR_TCR2EN)));
+ }
}
epd = false;
sh = extract32(tcr, 12, 2);
@@ -9429,7 +9439,13 @@ ARMVAParameters aa64_va_parameters(CPUARMState *env, uint64_t va,
regime_is_user(env, mmu_idx)) {
epd = true;
}
+
+ pie = ((env->cp15.tcr2_el[r_el] & TCR2_PIE)
+ && (!arm_feature(env, ARM_FEATURE_EL3)
+ || (env->cp15.scr_el3 & SCR_TCR2EN))
+ && (r_el == 2 || (arm_hcrx_el2_eff(env) & HCRX_TCR2EN)));
}
+ hpd |= pie;
gran = sanitize_gran_size(cpu, gran, stage2);
@@ -9508,6 +9524,7 @@ ARMVAParameters aa64_va_parameters(CPUARMState *env, uint64_t va,
.ha = ha,
.hd = ha && hd,
.gran = gran,
+ .pie = pie,
};
}
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 07/76] target/arm: Implement get_S1prot_indirect
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (5 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 06/76] target/arm: Populate PIE in aa64_va_parameters Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 08/76] target/arm: Implement get_S2prot_indirect Richard Henderson
` (69 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
This approximately corresponds to AArch64.S1IndirectBasePermissions
and the tail of AArch64.S1ComputePermissions which applies WXN.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/ptw.c | 169 +++++++++++++++++++++++++++++++++++++++--------
1 file changed, 143 insertions(+), 26 deletions(-)
diff --git a/target/arm/ptw.c b/target/arm/ptw.c
index d6d704e2a9..4fb4436db1 100644
--- a/target/arm/ptw.c
+++ b/target/arm/ptw.c
@@ -1448,6 +1448,106 @@ static int get_S1prot(CPUARMState *env, ARMMMUIdx mmu_idx, bool is_aa64,
return prot_rw | PAGE_EXEC;
}
+/* Extra page permission bits, during get_S1prot_indirect only. */
+#define PAGE_GCS (1 << 3)
+#define PAGE_WXN (1 << 4)
+#define PAGE_OVERLAY (1 << 5)
+QEMU_BUILD_BUG_ON(PAGE_RWX & (PAGE_GCS | PAGE_WXN | PAGE_OVERLAY));
+
+static int get_S1prot_indirect(CPUARMState *env, S1Translate *ptw,
+ ARMMMUIdx mmu_idx, int pi_index, int po_index,
+ ARMSecuritySpace in_pa, ARMSecuritySpace out_pa)
+{
+ static const uint8_t perm_table[16] = {
+ /* 0 */ PAGE_OVERLAY, /* no access */
+ /* 1 */ PAGE_OVERLAY | PAGE_READ,
+ /* 2 */ PAGE_OVERLAY | PAGE_EXEC,
+ /* 3 */ PAGE_OVERLAY | PAGE_READ | PAGE_EXEC,
+ /* 4 */ PAGE_OVERLAY, /* reserved */
+ /* 5 */ PAGE_OVERLAY | PAGE_READ | PAGE_WRITE,
+ /* 6 */ PAGE_OVERLAY | PAGE_READ | PAGE_WRITE | PAGE_EXEC | PAGE_WXN,
+ /* 7 */ PAGE_OVERLAY | PAGE_READ | PAGE_WRITE | PAGE_EXEC,
+ /* 8 */ PAGE_READ,
+ /* 9 */ PAGE_READ | PAGE_GCS,
+ /* A */ PAGE_READ | PAGE_EXEC,
+ /* B */ 0, /* reserved */
+ /* C */ PAGE_READ | PAGE_WRITE,
+ /* D */ 0, /* reserved */
+ /* E */ PAGE_READ | PAGE_WRITE | PAGE_EXEC,
+ /* F */ 0, /* reserved */
+ };
+
+ uint32_t el = regime_el(env, mmu_idx);
+ uint64_t pir = env->cp15.pir_el[el];
+ uint64_t pire0 = 0;
+ int perm;
+
+ if (el < 3) {
+ if (arm_feature(env, ARM_FEATURE_EL3)
+ && !(env->cp15.scr_el3 & SCR_PIEN)) {
+ pir = 0;
+ } else if (el == 2) {
+ pire0 = env->cp15.pire0_el2;
+ } else if (!ptw->in_nv1) {
+ pire0 = env->cp15.pir_el[0];
+ }
+ }
+ perm = perm_table[extract64(pir, pi_index * 4, 4)];
+
+ if (regime_has_2_ranges(mmu_idx)) {
+ int p_perm = perm;
+ int u_perm = perm_table[extract64(pire0, pi_index * 4, 4)];
+
+ if ((p_perm & (PAGE_EXEC | PAGE_GCS)) &&
+ (u_perm & (PAGE_WRITE | PAGE_GCS))) {
+ p_perm &= ~(PAGE_RWX | PAGE_GCS);
+ u_perm &= ~(PAGE_RWX | PAGE_GCS);
+ }
+ if ((u_perm & (PAGE_RWX | PAGE_GCS)) && regime_is_pan(env, mmu_idx)) {
+ p_perm &= ~(PAGE_READ | PAGE_WRITE);
+ }
+ perm = regime_is_user(env, mmu_idx) ? u_perm : p_perm;
+ }
+
+ if (in_pa != out_pa) {
+ switch (in_pa) {
+ case ARMSS_Root:
+ /*
+ * R_ZWRVD: permission fault for insn fetched from non-Root,
+ * I_WWBFB: SIF has no effect in EL3.
+ */
+ perm &= ~(PAGE_EXEC | PAGE_GCS);
+ break;
+ case ARMSS_Realm:
+ /*
+ * R_PKTDS: permission fault for insn fetched from non-Realm,
+ * for Realm EL2 or EL2&0. The corresponding fault for EL1&0
+ * happens during any stage2 translation.
+ */
+ if (el == 2) {
+ perm &= ~(PAGE_EXEC | PAGE_GCS);
+ }
+ break;
+ case ARMSS_Secure:
+ if (env->cp15.scr_el3 & SCR_SIF) {
+ perm &= ~(PAGE_EXEC | PAGE_GCS);
+ }
+ break;
+ default:
+ /* Input NonSecure must have output NonSecure. */
+ g_assert_not_reached();
+ }
+ }
+
+ if (perm & PAGE_WXN) {
+ perm &= ~PAGE_EXEC;
+ }
+
+ /* TODO: FEAT_GCS */
+
+ return perm & PAGE_RWX;
+}
+
static ARMVAParameters aa32_va_parameters(CPUARMState *env, uint32_t va,
ARMMMUIdx mmu_idx)
{
@@ -1677,7 +1777,7 @@ static bool get_phys_addr_lpae(CPUARMState *env, S1Translate *ptw,
int32_t stride;
int addrsize, inputsize, outputsize;
uint64_t tcr = regime_tcr(env, mmu_idx);
- int ap, xn, pxn;
+ int ap;
uint32_t el = regime_el(env, mmu_idx);
uint64_t descaddrmask;
bool aarch64 = arm_el_is_aa64(env, el);
@@ -2011,7 +2111,7 @@ static bool get_phys_addr_lpae(CPUARMState *env, S1Translate *ptw,
out_space = ARMSS_NonSecure;
result->f.prot = get_S2prot_noexecute(ap);
} else {
- xn = extract64(attrs, 53, 2);
+ int xn = extract64(attrs, 53, 2);
result->f.prot = get_S2prot(env, ap, xn, ptw->in_s1_is_el0);
}
@@ -2027,7 +2127,6 @@ static bool get_phys_addr_lpae(CPUARMState *env, S1Translate *ptw,
int nse, ns = extract32(attrs, 5, 1);
uint8_t attrindx;
uint64_t mair;
- int user_rw, prot_rw;
switch (out_space) {
case ARMSS_Root:
@@ -2076,33 +2175,51 @@ static bool get_phys_addr_lpae(CPUARMState *env, S1Translate *ptw,
default:
g_assert_not_reached();
}
- xn = extract64(attrs, 54, 1);
- pxn = extract64(attrs, 53, 1);
- if (el == 1 && ptw->in_nv1) {
+ if (param.pie) {
+ int pi = extract64(attrs, 6, 1)
+ | (extract64(attrs, 51, 1) << 1)
+ | (extract64(attrs, 53, 2) << 2);
+ int po = extract64(attrs, 60, 3);
/*
- * With FEAT_NV, when HCR_EL2.{NV,NV1} == {1,1}, the block/page
- * descriptor bit 54 holds PXN, 53 is RES0, and the effective value
- * of UXN is 0. Similarly for bits 59 and 60 in table descriptors
- * (which we have already folded into bits 53 and 54 of attrs).
- * AP[1] (descriptor bit 6, our ap bit 0) is treated as 0.
- * Similarly, APTable[0] from the table descriptor is treated as 0;
- * we already folded this into AP[1] and squashing that to 0 does
- * the right thing.
+ * Note that we modified ptw->in_space earlier for NSTable, but
+ * result->f.attrs retains a copy of the original security space.
*/
- pxn = xn;
- xn = 0;
- ap &= ~1;
- }
+ result->f.prot = get_S1prot_indirect(env, ptw, mmu_idx, pi, po,
+ result->f.attrs.space,
+ out_space);
+ } else {
+ int xn = extract64(attrs, 54, 1);
+ int pxn = extract64(attrs, 53, 1);
+ int user_rw, prot_rw;
- user_rw = simple_ap_to_rw_prot_is_user(ap, true);
- prot_rw = simple_ap_to_rw_prot_is_user(ap, false);
- /*
- * Note that we modified ptw->in_space earlier for NSTable, but
- * result->f.attrs retains a copy of the original security space.
- */
- result->f.prot = get_S1prot(env, mmu_idx, aarch64, user_rw, prot_rw,
- xn, pxn, result->f.attrs.space, out_space);
+ if (el == 1 && ptw->in_nv1) {
+ /*
+ * With FEAT_NV, when HCR_EL2.{NV,NV1} == {1,1},
+ * the block/page descriptor bit 54 holds PXN,
+ * 53 is RES0, and the effective value of UXN is 0.
+ * Similarly for bits 59 and 60 in table descriptors
+ * (which we have already folded into bits 53 and 54 of attrs).
+ * AP[1] (descriptor bit 6, our ap bit 0) is treated as 0.
+ * Similarly, APTable[0] from the table descriptor is treated
+ * as 0; we already folded this into AP[1] and squashing
+ * that to 0 does the right thing.
+ */
+ pxn = xn;
+ xn = 0;
+ ap &= ~1;
+ }
+
+ user_rw = simple_ap_to_rw_prot_is_user(ap, true);
+ prot_rw = simple_ap_to_rw_prot_is_user(ap, false);
+ /*
+ * Note that we modified ptw->in_space earlier for NSTable, but
+ * result->f.attrs retains a copy of the original security space.
+ */
+ result->f.prot = get_S1prot(env, mmu_idx, aarch64,
+ user_rw, prot_rw, xn, pxn,
+ result->f.attrs.space, out_space);
+ }
/* Index into MAIR registers for cache attributes */
attrindx = extract32(attrs, 2, 3);
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 08/76] target/arm: Implement get_S2prot_indirect
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (6 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 07/76] target/arm: Implement get_S1prot_indirect Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 09/76] target/arm: Expand CPUARMState.exception.syndrome to 64 bits Richard Henderson
` (68 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Move the stage2 permissions for normal accesses to
GetPhysAddrResult.s2prot. Put the stage2 permissions
for page table walking in CPUTLBEntryFull.prot.
This allows the permission checks in S1_ptw_translate
and arm_casq_ptw to see the right permission.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/internals.h | 7 ++++
target/arm/ptw.c | 81 +++++++++++++++++++++++++++++++-----------
2 files changed, 68 insertions(+), 20 deletions(-)
diff --git a/target/arm/internals.h b/target/arm/internals.h
index b7a1b69561..aa41054af0 100644
--- a/target/arm/internals.h
+++ b/target/arm/internals.h
@@ -1572,6 +1572,13 @@ typedef struct ARMCacheAttrs {
typedef struct GetPhysAddrResult {
CPUTLBEntryFull f;
ARMCacheAttrs cacheattrs;
+ /*
+ * For ARMMMUIdx_Stage2*, the protection installed into f.prot
+ * is the result for AccessType_TTW, i.e. the page table walk itself.
+ * The protection installed info s2prot is the one to be merged
+ * with the stage1 protection.
+ */
+ int s2prot;
} GetPhysAddrResult;
/**
diff --git a/target/arm/ptw.c b/target/arm/ptw.c
index 4fb4436db1..6dcd3adbdf 100644
--- a/target/arm/ptw.c
+++ b/target/arm/ptw.c
@@ -1280,7 +1280,7 @@ do_fault:
* @xn: XN (execute-never) bits
* @s1_is_el0: true if this is S2 of an S1+2 walk for EL0
*/
-static int get_S2prot_noexecute(int s2ap)
+static int get_S2prot(CPUARMState *env, int s2ap, int xn, bool s1_is_el0)
{
int prot = 0;
@@ -1290,12 +1290,6 @@ static int get_S2prot_noexecute(int s2ap)
if (s2ap & 2) {
prot |= PAGE_WRITE;
}
- return prot;
-}
-
-static int get_S2prot(CPUARMState *env, int s2ap, int xn, bool s1_is_el0)
-{
- int prot = get_S2prot_noexecute(s2ap);
if (cpu_isar_feature(any_tts2uxn, env_archcpu(env))) {
switch (xn) {
@@ -1327,6 +1321,44 @@ static int get_S2prot(CPUARMState *env, int s2ap, int xn, bool s1_is_el0)
return prot;
}
+static int get_S2prot_indirect(CPUARMState *env, GetPhysAddrResult *result,
+ int pi_index, int po_index, bool s1_is_el0)
+{
+ /* Last index is (priv, unpriv, ttw) */
+ static const uint8_t perm_table[16][3] = {
+ /* 0 */ { 0, 0, 0 }, /* no access */
+ /* 1 */ { 0, 0, 0 }, /* reserved */
+ /* 2 */ { PAGE_READ, PAGE_READ, PAGE_READ | PAGE_WRITE },
+ /* 3 */ { PAGE_READ, PAGE_READ, PAGE_READ | PAGE_WRITE },
+ /* 4 */ { PAGE_WRITE, PAGE_WRITE, 0 },
+ /* 5 */ { 0, 0, 0 }, /* reserved */
+ /* 6 */ { PAGE_READ, PAGE_READ, PAGE_READ | PAGE_WRITE },
+ /* 7 */ { PAGE_READ, PAGE_READ, PAGE_READ | PAGE_WRITE },
+ /* 8 */ { PAGE_READ, PAGE_READ, PAGE_READ },
+ /* 9 */ { PAGE_READ, PAGE_READ | PAGE_EXEC, PAGE_READ },
+ /* A */ { PAGE_READ | PAGE_EXEC, PAGE_READ, PAGE_READ },
+ /* B */ { PAGE_READ | PAGE_EXEC, PAGE_READ | PAGE_EXEC, PAGE_READ },
+ /* C */ { PAGE_READ | PAGE_WRITE,
+ PAGE_READ | PAGE_WRITE,
+ PAGE_READ | PAGE_WRITE },
+ /* D */ { PAGE_READ | PAGE_WRITE,
+ PAGE_READ | PAGE_WRITE | PAGE_EXEC,
+ PAGE_READ | PAGE_WRITE },
+ /* E */ { PAGE_READ | PAGE_WRITE | PAGE_EXEC,
+ PAGE_READ | PAGE_WRITE,
+ PAGE_READ | PAGE_WRITE },
+ /* F */ { PAGE_READ | PAGE_WRITE | PAGE_EXEC,
+ PAGE_READ | PAGE_WRITE | PAGE_EXEC,
+ PAGE_READ | PAGE_WRITE },
+ };
+
+ uint64_t pir = (env->cp15.scr_el3 & SCR_PIEN ? env->cp15.s2pir_el2 : 0);
+ int s2pi = extract64(pir, pi_index * 4, 4);
+
+ result->f.prot = perm_table[s2pi][2];
+ return perm_table[s2pi][s1_is_el0];
+}
+
/*
* Translate section/page access permissions to protection flags
* @env: CPUARMState
@@ -1777,7 +1809,7 @@ static bool get_phys_addr_lpae(CPUARMState *env, S1Translate *ptw,
int32_t stride;
int addrsize, inputsize, outputsize;
uint64_t tcr = regime_tcr(env, mmu_idx);
- int ap;
+ int ap, prot;
uint32_t el = regime_el(env, mmu_idx);
uint64_t descaddrmask;
bool aarch64 = arm_el_is_aa64(env, el);
@@ -2101,6 +2133,18 @@ static bool get_phys_addr_lpae(CPUARMState *env, S1Translate *ptw,
ap = extract32(attrs, 6, 2);
out_space = ptw->in_space;
if (regime_is_stage2(mmu_idx)) {
+ if (param.pie) {
+ int pi = extract64(attrs, 6, 1)
+ | (extract64(attrs, 51, 1) << 1)
+ | (extract64(attrs, 53, 2) << 2);
+ int po = extract64(attrs, 60, 3);
+ prot = get_S2prot_indirect(env, result, pi, po, ptw->in_s1_is_el0);
+ } else {
+ int xn = extract64(attrs, 53, 2);
+ prot = get_S2prot(env, ap, xn, ptw->in_s1_is_el0);
+ /* Install TTW permissions in f.prot. */
+ result->f.prot = prot & (PAGE_READ | PAGE_WRITE);
+ }
/*
* R_GYNXY: For stage2 in Realm security state, bit 55 is NS.
* The bit remains ignored for other security states.
@@ -2109,11 +2153,9 @@ static bool get_phys_addr_lpae(CPUARMState *env, S1Translate *ptw,
*/
if (out_space == ARMSS_Realm && extract64(attrs, 55, 1)) {
out_space = ARMSS_NonSecure;
- result->f.prot = get_S2prot_noexecute(ap);
- } else {
- int xn = extract64(attrs, 53, 2);
- result->f.prot = get_S2prot(env, ap, xn, ptw->in_s1_is_el0);
+ prot &= ~PAGE_EXEC;
}
+ result->s2prot = prot;
result->cacheattrs.is_s2_format = true;
result->cacheattrs.attrs = extract32(attrs, 2, 4);
@@ -2185,9 +2227,8 @@ static bool get_phys_addr_lpae(CPUARMState *env, S1Translate *ptw,
* Note that we modified ptw->in_space earlier for NSTable, but
* result->f.attrs retains a copy of the original security space.
*/
- result->f.prot = get_S1prot_indirect(env, ptw, mmu_idx, pi, po,
- result->f.attrs.space,
- out_space);
+ prot = get_S1prot_indirect(env, ptw, mmu_idx, pi, po,
+ result->f.attrs.space, out_space);
} else {
int xn = extract64(attrs, 54, 1);
int pxn = extract64(attrs, 53, 1);
@@ -2216,10 +2257,10 @@ static bool get_phys_addr_lpae(CPUARMState *env, S1Translate *ptw,
* Note that we modified ptw->in_space earlier for NSTable, but
* result->f.attrs retains a copy of the original security space.
*/
- result->f.prot = get_S1prot(env, mmu_idx, aarch64,
- user_rw, prot_rw, xn, pxn,
- result->f.attrs.space, out_space);
+ prot = get_S1prot(env, mmu_idx, aarch64, user_rw, prot_rw, xn, pxn,
+ result->f.attrs.space, out_space);
}
+ result->f.prot = prot;
/* Index into MAIR registers for cache attributes */
attrindx = extract32(attrs, 2, 3);
@@ -2265,7 +2306,7 @@ static bool get_phys_addr_lpae(CPUARMState *env, S1Translate *ptw,
result->f.tlb_fill_flags = 0;
}
- if (ptw->in_prot_check & ~result->f.prot) {
+ if (ptw->in_prot_check & ~prot) {
fi->type = ARMFault_Permission;
goto do_fault;
}
@@ -3463,7 +3504,7 @@ static bool get_phys_addr_twostage(CPUARMState *env, S1Translate *ptw,
fi->s2addr = ipa;
/* Combine the S1 and S2 perms. */
- result->f.prot &= s1_prot;
+ result->f.prot = s1_prot & result->s2prot;
/* If S2 fails, return early. */
if (ret) {
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 09/76] target/arm: Expand CPUARMState.exception.syndrome to 64 bits
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (7 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 08/76] target/arm: Implement get_S2prot_indirect Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 10/76] target/arm: Expand syndrome parameter to raise_exception* Richard Henderson
` (67 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
This will be used for storing the ISS2 portion of the
ESR_ELx registers in aarch64 state. Re-order the fsr
member to eliminate two structure holes.
Drop the comment about "if we implement EL2" since we
have already done so.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/cpu.h | 7 ++-----
target/arm/helper.c | 2 +-
target/arm/machine.c | 32 +++++++++++++++++++++++++++++++-
3 files changed, 34 insertions(+), 7 deletions(-)
diff --git a/target/arm/cpu.h b/target/arm/cpu.h
index d17252f734..4dd4c6d4bf 100644
--- a/target/arm/cpu.h
+++ b/target/arm/cpu.h
@@ -633,13 +633,10 @@ typedef struct CPUArchState {
* entry process.
*/
struct {
- uint32_t syndrome; /* AArch64 format syndrome register */
- uint32_t fsr; /* AArch32 format fault status register info */
+ uint64_t syndrome; /* AArch64 format syndrome register */
uint64_t vaddress; /* virtual addr associated with exception, if any */
+ uint32_t fsr; /* AArch32 format fault status register info */
uint32_t target_el; /* EL the exception should be targeted for */
- /* If we implement EL2 we will also need to store information
- * about the intermediate physical address for stage 2 faults.
- */
} exception;
/* Information associated with an SError */
diff --git a/target/arm/helper.c b/target/arm/helper.c
index 1111a16330..f66868f9ef 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -9182,7 +9182,7 @@ void arm_cpu_do_interrupt(CPUState *cs)
new_el);
if (qemu_loglevel_mask(CPU_LOG_INT)
&& !excp_is_internal(cs->exception_index)) {
- qemu_log_mask(CPU_LOG_INT, "...with ESR 0x%x/0x%" PRIx32 "\n",
+ qemu_log_mask(CPU_LOG_INT, "...with ESR 0x%x/0x%" PRIx64 "\n",
syn_get_ec(env->exception.syndrome),
env->exception.syndrome);
}
diff --git a/target/arm/machine.c b/target/arm/machine.c
index 6666a0c50c..ce20b46f50 100644
--- a/target/arm/machine.c
+++ b/target/arm/machine.c
@@ -848,6 +848,23 @@ static const VMStateInfo vmstate_powered_off = {
.put = put_power,
};
+static bool syndrome64_needed(void *opaque)
+{
+ ARMCPU *cpu = opaque;
+ return cpu->env.exception.syndrome > UINT32_MAX;
+}
+
+static const VMStateDescription vmstate_syndrome64 = {
+ .name = "cpu/syndrome64",
+ .version_id = 1,
+ .minimum_version_id = 1,
+ .needed = syndrome64_needed,
+ .fields = (const VMStateField[]) {
+ VMSTATE_UINT64(env.exception.syndrome, ARMCPU),
+ VMSTATE_END_OF_LIST()
+ },
+};
+
static int cpu_pre_save(void *opaque)
{
ARMCPU *cpu = opaque;
@@ -1065,7 +1082,19 @@ const VMStateDescription vmstate_arm_cpu = {
VMSTATE_UINT64(env.exclusive_val, ARMCPU),
VMSTATE_UINT64(env.exclusive_high, ARMCPU),
VMSTATE_UNUSED(sizeof(uint64_t)),
- VMSTATE_UINT32(env.exception.syndrome, ARMCPU),
+ /*
+ * If any bits are set in the upper 32 bits of syndrome,
+ * then the cpu/syndrome64 subsection will override this
+ * with the full 64 bit state.
+ */
+ {
+ .name = "env.exception.syndrome",
+ .version_id = 0,
+ .size = sizeof(uint32_t),
+ .info = &vmstate_info_uint32,
+ .flags = VMS_SINGLE,
+ .offset = offsetoflow32(ARMCPU, env.exception.syndrome),
+ },
VMSTATE_UINT32(env.exception.fsr, ARMCPU),
VMSTATE_UINT64(env.exception.vaddress, ARMCPU),
VMSTATE_TIMER_PTR(gt_timer[GTIMER_PHYS], ARMCPU),
@@ -1098,6 +1127,7 @@ const VMStateDescription vmstate_arm_cpu = {
&vmstate_serror,
&vmstate_irq_line_state,
&vmstate_wfxt_timer,
+ &vmstate_syndrome64,
NULL
}
};
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 10/76] target/arm: Expand syndrome parameter to raise_exception*
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (8 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 09/76] target/arm: Expand CPUARMState.exception.syndrome to 64 bits Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 11/76] target/arm: Implement dirtybit check for PIE Richard Henderson
` (66 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Prepare for raising exceptions with 64-bit syndromes.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/internals.h | 6 +++---
target/arm/tcg-stubs.c | 2 +-
target/arm/tcg/op_helper.c | 4 ++--
3 files changed, 6 insertions(+), 6 deletions(-)
diff --git a/target/arm/internals.h b/target/arm/internals.h
index aa41054af0..0ebcf566ec 100644
--- a/target/arm/internals.h
+++ b/target/arm/internals.h
@@ -304,14 +304,14 @@ FIELD(CNTHCTL, CNTPMASK, 19, 1)
* and never returns because we will longjump back up to the CPU main loop.
*/
G_NORETURN void raise_exception(CPUARMState *env, uint32_t excp,
- uint32_t syndrome, uint32_t target_el);
+ uint64_t syndrome, uint32_t target_el);
/*
* Similarly, but also use unwinding to restore cpu state.
*/
G_NORETURN void raise_exception_ra(CPUARMState *env, uint32_t excp,
- uint32_t syndrome, uint32_t target_el,
- uintptr_t ra);
+ uint64_t syndrome, uint32_t target_el,
+ uintptr_t ra);
/*
* For AArch64, map a given EL to an index in the banked_spsr array.
diff --git a/target/arm/tcg-stubs.c b/target/arm/tcg-stubs.c
index aac99b2672..aeeede8066 100644
--- a/target/arm/tcg-stubs.c
+++ b/target/arm/tcg-stubs.c
@@ -16,7 +16,7 @@ void write_v7m_exception(CPUARMState *env, uint32_t new_exc)
g_assert_not_reached();
}
-void raise_exception_ra(CPUARMState *env, uint32_t excp, uint32_t syndrome,
+void raise_exception_ra(CPUARMState *env, uint32_t excp, uint64_t syndrome,
uint32_t target_el, uintptr_t ra)
{
g_assert_not_reached();
diff --git a/target/arm/tcg/op_helper.c b/target/arm/tcg/op_helper.c
index 5373e0e998..dd3700dc6f 100644
--- a/target/arm/tcg/op_helper.c
+++ b/target/arm/tcg/op_helper.c
@@ -46,7 +46,7 @@ int exception_target_el(CPUARMState *env)
}
void raise_exception(CPUARMState *env, uint32_t excp,
- uint32_t syndrome, uint32_t target_el)
+ uint64_t syndrome, uint32_t target_el)
{
CPUState *cs = env_cpu(env);
@@ -70,7 +70,7 @@ void raise_exception(CPUARMState *env, uint32_t excp,
cpu_loop_exit(cs);
}
-void raise_exception_ra(CPUARMState *env, uint32_t excp, uint32_t syndrome,
+void raise_exception_ra(CPUARMState *env, uint32_t excp, uint64_t syndrome,
uint32_t target_el, uintptr_t ra)
{
CPUState *cs = env_cpu(env);
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 11/76] target/arm: Implement dirtybit check for PIE
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (9 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 10/76] target/arm: Expand syndrome parameter to raise_exception* Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 12/76] target/arm: Enable FEAT_S1PIE and FEAT_S2PIE on -cpu max Richard Henderson
` (65 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Both S1PIE and S2PIE have a bit to make software tracking
of dirty pages easier.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/internals.h | 1 +
target/arm/ptw.c | 16 ++++++++++++++++
target/arm/tcg/tlb_helper.c | 12 +++++++++---
3 files changed, 26 insertions(+), 3 deletions(-)
diff --git a/target/arm/internals.h b/target/arm/internals.h
index 0ebcf566ec..a2ee5f8739 100644
--- a/target/arm/internals.h
+++ b/target/arm/internals.h
@@ -752,6 +752,7 @@ struct ARMMMUFaultInfo {
bool s1ptw;
bool s1ns;
bool ea;
+ bool dirtybit; /* FEAT_S1PIE, FEAT_S2PIE */
};
/**
diff --git a/target/arm/ptw.c b/target/arm/ptw.c
index 6dcd3adbdf..be3629a745 100644
--- a/target/arm/ptw.c
+++ b/target/arm/ptw.c
@@ -2311,6 +2311,22 @@ static bool get_phys_addr_lpae(CPUARMState *env, S1Translate *ptw,
goto do_fault;
}
+ /* S1PIE and S2PIE both have a bit for software dirty page tracking. */
+ if (access_type == MMU_DATA_STORE && param.pie) {
+ /*
+ * For S1PIE, bit 7 is nDirty and both HA and HD are checked.
+ * For S2PIE, bit 7 is Dirty and only HD is checked.
+ */
+ bool bit7 = extract64(attrs, 7, 1);
+ if (regime_is_stage2(mmu_idx)
+ ? !bit7 && !param.hd
+ : bit7 && !(param.ha && param.hd)) {
+ fi->type = ARMFault_Permission;
+ fi->dirtybit = true;
+ goto do_fault;
+ }
+ }
+
/* If FEAT_HAFDBS has made changes, update the PTE. */
if (new_descriptor != descriptor) {
new_descriptor = arm_casq_ptw(env, descriptor, new_descriptor, ptw, fi);
diff --git a/target/arm/tcg/tlb_helper.c b/target/arm/tcg/tlb_helper.c
index 23c72a99f5..ae2acd6727 100644
--- a/target/arm/tcg/tlb_helper.c
+++ b/target/arm/tcg/tlb_helper.c
@@ -24,13 +24,13 @@ bool arm_s1_regime_using_lpae_format(CPUARMState *env, ARMMMUIdx mmu_idx)
return regime_using_lpae_format(env, mmu_idx);
}
-static inline uint32_t merge_syn_data_abort(uint32_t template_syn,
+static inline uint64_t merge_syn_data_abort(uint32_t template_syn,
ARMMMUFaultInfo *fi,
unsigned int target_el,
bool same_el, bool is_write,
int fsc)
{
- uint32_t syn;
+ uint64_t syn;
/*
* ISV is only set for stage-2 data aborts routed to EL2 and
@@ -75,6 +75,10 @@ static inline uint32_t merge_syn_data_abort(uint32_t template_syn,
/* Merge the runtime syndrome with the template syndrome. */
syn |= template_syn;
}
+
+ /* Form ISS2 at the top of the syndrome. */
+ syn |= (uint64_t)fi->dirtybit << 37;
+
return syn;
}
@@ -176,7 +180,9 @@ void arm_deliver_fault(ARMCPU *cpu, vaddr addr,
int target_el = exception_target_el(env);
int current_el = arm_current_el(env);
bool same_el;
- uint32_t syn, exc, fsr, fsc;
+ uint32_t exc, fsr, fsc;
+ uint64_t syn;
+
/*
* We know this must be a data or insn abort, and that
* env->exception.syndrome contains the template syndrome set
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 12/76] target/arm: Enable FEAT_S1PIE and FEAT_S2PIE on -cpu max
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (10 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 11/76] target/arm: Implement dirtybit check for PIE Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 13/76] include/hw/core/cpu: Introduce MMUIdxMap Richard Henderson
` (64 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/tcg/cpu64.c | 2 ++
docs/system/arm/emulation.rst | 2 ++
2 files changed, 4 insertions(+)
diff --git a/target/arm/tcg/cpu64.c b/target/arm/tcg/cpu64.c
index abef6a246e..76193826ce 100644
--- a/target/arm/tcg/cpu64.c
+++ b/target/arm/tcg/cpu64.c
@@ -1253,6 +1253,8 @@ void aarch64_max_tcg_initfn(Object *obj)
t = FIELD_DP64(t, ID_AA64MMFR3, TCRX, 1); /* FEAT_TCR2 */
t = FIELD_DP64(t, ID_AA64MMFR3, SCTLRX, 1); /* FEAT_SCTLR2 */
t = FIELD_DP64(t, ID_AA64MMFR3, SPEC_FPACC, 1); /* FEAT_FPACC_SPEC */
+ t = FIELD_DP64(t, ID_AA64MMFR3, S1PIE, 1); /* FEAT_S1PIE */
+ t = FIELD_DP64(t, ID_AA64MMFR3, S2PIE, 1); /* FEAT_S2PIE */
SET_IDREG(isar, ID_AA64MMFR3, t);
t = GET_IDREG(isar, ID_AA64ZFR0);
diff --git a/docs/system/arm/emulation.rst b/docs/system/arm/emulation.rst
index 6b04c96c8c..94a6192fa9 100644
--- a/docs/system/arm/emulation.rst
+++ b/docs/system/arm/emulation.rst
@@ -122,6 +122,8 @@ the following architecture extensions:
- FEAT_RME (Realm Management Extension) (NB: support status in QEMU is experimental)
- FEAT_RNG (Random number generator)
- FEAT_RPRES (Increased precision of FRECPE and FRSQRTE)
+- FEAT_S1PIE (Stage 1 permission indirections)
+- FEAT_S2PIE (Stage 2 permission indirections)
- FEAT_S2FWB (Stage 2 forced Write-Back)
- FEAT_SB (Speculation Barrier)
- FEAT_SCTLR2 (Extension to SCTLR_ELx)
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 13/76] include/hw/core/cpu: Introduce MMUIdxMap
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (11 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 12/76] target/arm: Enable FEAT_S1PIE and FEAT_S2PIE on -cpu max Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 14/76] include/hw/core/cpu: Introduce cpu_tlb_fast Richard Henderson
` (63 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Philippe Mathieu-Daudé, Pierrick Bouvier
Use a typedef instead of uint16_t directly when
describing sets of mmu indexes.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
include/exec/cputlb.h | 32 ++++++++++++++++----------------
include/hw/core/cpu.h | 6 +++---
accel/tcg/cputlb.c | 30 +++++++++++++++---------------
3 files changed, 34 insertions(+), 34 deletions(-)
diff --git a/include/exec/cputlb.h b/include/exec/cputlb.h
index 03ed7e2165..9bec0e7890 100644
--- a/include/exec/cputlb.h
+++ b/include/exec/cputlb.h
@@ -150,7 +150,7 @@ void tlb_flush_all_cpus_synced(CPUState *src_cpu);
* MMU indexes.
*/
void tlb_flush_page_by_mmuidx(CPUState *cpu, vaddr addr,
- uint16_t idxmap);
+ MMUIdxMap idxmap);
/**
* tlb_flush_page_by_mmuidx_all_cpus_synced:
@@ -165,7 +165,7 @@ void tlb_flush_page_by_mmuidx(CPUState *cpu, vaddr addr,
* translations using the flushed TLBs.
*/
void tlb_flush_page_by_mmuidx_all_cpus_synced(CPUState *cpu, vaddr addr,
- uint16_t idxmap);
+ MMUIdxMap idxmap);
/**
* tlb_flush_by_mmuidx:
@@ -176,7 +176,7 @@ void tlb_flush_page_by_mmuidx_all_cpus_synced(CPUState *cpu, vaddr addr,
* Flush all entries from the TLB of the specified CPU, for the specified
* MMU indexes.
*/
-void tlb_flush_by_mmuidx(CPUState *cpu, uint16_t idxmap);
+void tlb_flush_by_mmuidx(CPUState *cpu, MMUIdxMap idxmap);
/**
* tlb_flush_by_mmuidx_all_cpus_synced:
@@ -189,7 +189,7 @@ void tlb_flush_by_mmuidx(CPUState *cpu, uint16_t idxmap);
* When this function returns, no CPUs will subsequently perform
* translations using the flushed TLBs.
*/
-void tlb_flush_by_mmuidx_all_cpus_synced(CPUState *cpu, uint16_t idxmap);
+void tlb_flush_by_mmuidx_all_cpus_synced(CPUState *cpu, MMUIdxMap idxmap);
/**
* tlb_flush_page_bits_by_mmuidx
@@ -201,11 +201,11 @@ void tlb_flush_by_mmuidx_all_cpus_synced(CPUState *cpu, uint16_t idxmap);
* Similar to tlb_flush_page_mask, but with a bitmap of indexes.
*/
void tlb_flush_page_bits_by_mmuidx(CPUState *cpu, vaddr addr,
- uint16_t idxmap, unsigned bits);
+ MMUIdxMap idxmap, unsigned bits);
/* Similarly, with broadcast and syncing. */
void tlb_flush_page_bits_by_mmuidx_all_cpus_synced(CPUState *cpu, vaddr addr,
- uint16_t idxmap,
+ MMUIdxMap idxmap,
unsigned bits);
/**
@@ -220,14 +220,14 @@ void tlb_flush_page_bits_by_mmuidx_all_cpus_synced(CPUState *cpu, vaddr addr,
* comparing only the low @bits worth of each virtual page.
*/
void tlb_flush_range_by_mmuidx(CPUState *cpu, vaddr addr,
- vaddr len, uint16_t idxmap,
+ vaddr len, MMUIdxMap idxmap,
unsigned bits);
/* Similarly, with broadcast and syncing. */
void tlb_flush_range_by_mmuidx_all_cpus_synced(CPUState *cpu,
vaddr addr,
vaddr len,
- uint16_t idxmap,
+ MMUIdxMap idxmap,
unsigned bits);
#else
static inline void tlb_flush_page(CPUState *cpu, vaddr addr)
@@ -243,42 +243,42 @@ static inline void tlb_flush_all_cpus_synced(CPUState *src_cpu)
{
}
static inline void tlb_flush_page_by_mmuidx(CPUState *cpu,
- vaddr addr, uint16_t idxmap)
+ vaddr addr, MMUIdxMap idxmap)
{
}
-static inline void tlb_flush_by_mmuidx(CPUState *cpu, uint16_t idxmap)
+static inline void tlb_flush_by_mmuidx(CPUState *cpu, MMUIdxMap idxmap)
{
}
static inline void tlb_flush_page_by_mmuidx_all_cpus_synced(CPUState *cpu,
vaddr addr,
- uint16_t idxmap)
+ MMUIdxMap idxmap)
{
}
static inline void tlb_flush_by_mmuidx_all_cpus_synced(CPUState *cpu,
- uint16_t idxmap)
+ MMUIdxMap idxmap)
{
}
static inline void tlb_flush_page_bits_by_mmuidx(CPUState *cpu,
vaddr addr,
- uint16_t idxmap,
+ MMUIdxMap idxmap,
unsigned bits)
{
}
static inline void
tlb_flush_page_bits_by_mmuidx_all_cpus_synced(CPUState *cpu, vaddr addr,
- uint16_t idxmap, unsigned bits)
+ MMUIdxMap idxmap, unsigned bits)
{
}
static inline void tlb_flush_range_by_mmuidx(CPUState *cpu, vaddr addr,
- vaddr len, uint16_t idxmap,
+ vaddr len, MMUIdxMap idxmap,
unsigned bits)
{
}
static inline void tlb_flush_range_by_mmuidx_all_cpus_synced(CPUState *cpu,
vaddr addr,
vaddr len,
- uint16_t idxmap,
+ MMUIdxMap idxmap,
unsigned bits)
{
}
diff --git a/include/hw/core/cpu.h b/include/hw/core/cpu.h
index b01a0cffd6..a7038b8c07 100644
--- a/include/hw/core/cpu.h
+++ b/include/hw/core/cpu.h
@@ -198,10 +198,10 @@ struct CPUClass {
};
/*
- * Fix the number of mmu modes to 16, which is also the maximum
- * supported by the softmmu tlb api.
+ * Fix the number of mmu modes to 16.
*/
#define NB_MMU_MODES 16
+typedef uint16_t MMUIdxMap;
/* Use a fully associative victim tlb of 8 entries. */
#define CPU_VTLB_SIZE 8
@@ -306,7 +306,7 @@ typedef struct CPUTLBCommon {
* mmu_idx N since the last time that mmu_idx was flushed.
* Protected by tlb_c.lock.
*/
- uint16_t dirty;
+ MMUIdxMap dirty;
/*
* Statistics. These are not lock protected, but are read and
* written atomically. This allows the monitor to print a snapshot
diff --git a/accel/tcg/cputlb.c b/accel/tcg/cputlb.c
index 87e14bde4f..d324f33339 100644
--- a/accel/tcg/cputlb.c
+++ b/accel/tcg/cputlb.c
@@ -370,8 +370,8 @@ static void flush_all_helper(CPUState *src, run_on_cpu_func fn,
static void tlb_flush_by_mmuidx_async_work(CPUState *cpu, run_on_cpu_data data)
{
- uint16_t asked = data.host_int;
- uint16_t all_dirty, work, to_clean;
+ MMUIdxMap asked = data.host_int;
+ MMUIdxMap all_dirty, work, to_clean;
int64_t now = get_clock_realtime();
assert_cpu_is_self(cpu);
@@ -408,7 +408,7 @@ static void tlb_flush_by_mmuidx_async_work(CPUState *cpu, run_on_cpu_data data)
}
}
-void tlb_flush_by_mmuidx(CPUState *cpu, uint16_t idxmap)
+void tlb_flush_by_mmuidx(CPUState *cpu, MMUIdxMap idxmap)
{
tlb_debug("mmu_idx: 0x%" PRIx16 "\n", idxmap);
@@ -422,7 +422,7 @@ void tlb_flush(CPUState *cpu)
tlb_flush_by_mmuidx(cpu, ALL_MMUIDX_BITS);
}
-void tlb_flush_by_mmuidx_all_cpus_synced(CPUState *src_cpu, uint16_t idxmap)
+void tlb_flush_by_mmuidx_all_cpus_synced(CPUState *src_cpu, MMUIdxMap idxmap)
{
const run_on_cpu_func fn = tlb_flush_by_mmuidx_async_work;
@@ -531,7 +531,7 @@ static void tlb_flush_page_locked(CPUState *cpu, int midx, vaddr page)
*/
static void tlb_flush_page_by_mmuidx_async_0(CPUState *cpu,
vaddr addr,
- uint16_t idxmap)
+ MMUIdxMap idxmap)
{
int mmu_idx;
@@ -570,14 +570,14 @@ static void tlb_flush_page_by_mmuidx_async_1(CPUState *cpu,
{
vaddr addr_and_idxmap = data.target_ptr;
vaddr addr = addr_and_idxmap & TARGET_PAGE_MASK;
- uint16_t idxmap = addr_and_idxmap & ~TARGET_PAGE_MASK;
+ MMUIdxMap idxmap = addr_and_idxmap & ~TARGET_PAGE_MASK;
tlb_flush_page_by_mmuidx_async_0(cpu, addr, idxmap);
}
typedef struct {
vaddr addr;
- uint16_t idxmap;
+ MMUIdxMap idxmap;
} TLBFlushPageByMMUIdxData;
/**
@@ -599,7 +599,7 @@ static void tlb_flush_page_by_mmuidx_async_2(CPUState *cpu,
g_free(d);
}
-void tlb_flush_page_by_mmuidx(CPUState *cpu, vaddr addr, uint16_t idxmap)
+void tlb_flush_page_by_mmuidx(CPUState *cpu, vaddr addr, MMUIdxMap idxmap)
{
tlb_debug("addr: %016" VADDR_PRIx " mmu_idx:%" PRIx16 "\n", addr, idxmap);
@@ -618,7 +618,7 @@ void tlb_flush_page(CPUState *cpu, vaddr addr)
void tlb_flush_page_by_mmuidx_all_cpus_synced(CPUState *src_cpu,
vaddr addr,
- uint16_t idxmap)
+ MMUIdxMap idxmap)
{
tlb_debug("addr: %016" VADDR_PRIx " mmu_idx:%"PRIx16"\n", addr, idxmap);
@@ -715,8 +715,8 @@ static void tlb_flush_range_locked(CPUState *cpu, int midx,
typedef struct {
vaddr addr;
vaddr len;
- uint16_t idxmap;
- uint16_t bits;
+ MMUIdxMap idxmap;
+ unsigned bits;
} TLBFlushRangeData;
static void tlb_flush_range_by_mmuidx_async_0(CPUState *cpu,
@@ -766,7 +766,7 @@ static void tlb_flush_range_by_mmuidx_async_1(CPUState *cpu,
}
void tlb_flush_range_by_mmuidx(CPUState *cpu, vaddr addr,
- vaddr len, uint16_t idxmap,
+ vaddr len, MMUIdxMap idxmap,
unsigned bits)
{
TLBFlushRangeData d;
@@ -797,7 +797,7 @@ void tlb_flush_range_by_mmuidx(CPUState *cpu, vaddr addr,
}
void tlb_flush_page_bits_by_mmuidx(CPUState *cpu, vaddr addr,
- uint16_t idxmap, unsigned bits)
+ MMUIdxMap idxmap, unsigned bits)
{
tlb_flush_range_by_mmuidx(cpu, addr, TARGET_PAGE_SIZE, idxmap, bits);
}
@@ -805,7 +805,7 @@ void tlb_flush_page_bits_by_mmuidx(CPUState *cpu, vaddr addr,
void tlb_flush_range_by_mmuidx_all_cpus_synced(CPUState *src_cpu,
vaddr addr,
vaddr len,
- uint16_t idxmap,
+ MMUIdxMap idxmap,
unsigned bits)
{
TLBFlushRangeData d, *p;
@@ -847,7 +847,7 @@ void tlb_flush_range_by_mmuidx_all_cpus_synced(CPUState *src_cpu,
void tlb_flush_page_bits_by_mmuidx_all_cpus_synced(CPUState *src_cpu,
vaddr addr,
- uint16_t idxmap,
+ MMUIdxMap idxmap,
unsigned bits)
{
tlb_flush_range_by_mmuidx_all_cpus_synced(src_cpu, addr, TARGET_PAGE_SIZE,
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 14/76] include/hw/core/cpu: Introduce cpu_tlb_fast
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (12 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 13/76] include/hw/core/cpu: Introduce MMUIdxMap Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 15/76] include/hw/core/cpu: Invert the indexing into CPUTLBDescFast Richard Henderson
` (62 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Philippe Mathieu-Daudé, Pierrick Bouvier
Encapsulate access to cpu->neg.tlb.f[] in a function.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
include/hw/core/cpu.h | 7 +++++++
accel/tcg/cputlb.c | 16 ++++++++--------
tcg/aarch64/tcg-target.c.inc | 2 +-
tcg/arm/tcg-target.c.inc | 2 +-
4 files changed, 17 insertions(+), 10 deletions(-)
diff --git a/include/hw/core/cpu.h b/include/hw/core/cpu.h
index a7038b8c07..30e335d7d0 100644
--- a/include/hw/core/cpu.h
+++ b/include/hw/core/cpu.h
@@ -594,6 +594,13 @@ static inline CPUArchState *cpu_env(CPUState *cpu)
return (CPUArchState *)(cpu + 1);
}
+#ifdef CONFIG_TCG
+static inline CPUTLBDescFast *cpu_tlb_fast(CPUState *cpu, int mmu_idx)
+{
+ return &cpu->neg.tlb.f[mmu_idx];
+}
+#endif
+
typedef QTAILQ_HEAD(CPUTailQ, CPUState) CPUTailQ;
extern CPUTailQ cpus_queue;
diff --git a/accel/tcg/cputlb.c b/accel/tcg/cputlb.c
index d324f33339..2a6aa01c57 100644
--- a/accel/tcg/cputlb.c
+++ b/accel/tcg/cputlb.c
@@ -129,7 +129,7 @@ static inline uint64_t tlb_addr_write(const CPUTLBEntry *entry)
static inline uintptr_t tlb_index(CPUState *cpu, uintptr_t mmu_idx,
vaddr addr)
{
- uintptr_t size_mask = cpu->neg.tlb.f[mmu_idx].mask >> CPU_TLB_ENTRY_BITS;
+ uintptr_t size_mask = cpu_tlb_fast(cpu, mmu_idx)->mask >> CPU_TLB_ENTRY_BITS;
return (addr >> TARGET_PAGE_BITS) & size_mask;
}
@@ -138,7 +138,7 @@ static inline uintptr_t tlb_index(CPUState *cpu, uintptr_t mmu_idx,
static inline CPUTLBEntry *tlb_entry(CPUState *cpu, uintptr_t mmu_idx,
vaddr addr)
{
- return &cpu->neg.tlb.f[mmu_idx].table[tlb_index(cpu, mmu_idx, addr)];
+ return &cpu_tlb_fast(cpu, mmu_idx)->table[tlb_index(cpu, mmu_idx, addr)];
}
static void tlb_window_reset(CPUTLBDesc *desc, int64_t ns,
@@ -292,7 +292,7 @@ static void tlb_flush_one_mmuidx_locked(CPUState *cpu, int mmu_idx,
int64_t now)
{
CPUTLBDesc *desc = &cpu->neg.tlb.d[mmu_idx];
- CPUTLBDescFast *fast = &cpu->neg.tlb.f[mmu_idx];
+ CPUTLBDescFast *fast = cpu_tlb_fast(cpu, mmu_idx);
tlb_mmu_resize_locked(desc, fast, now);
tlb_mmu_flush_locked(desc, fast);
@@ -331,7 +331,7 @@ void tlb_init(CPUState *cpu)
cpu->neg.tlb.c.dirty = 0;
for (i = 0; i < NB_MMU_MODES; i++) {
- tlb_mmu_init(&cpu->neg.tlb.d[i], &cpu->neg.tlb.f[i], now);
+ tlb_mmu_init(&cpu->neg.tlb.d[i], cpu_tlb_fast(cpu, i), now);
}
}
@@ -342,7 +342,7 @@ void tlb_destroy(CPUState *cpu)
qemu_spin_destroy(&cpu->neg.tlb.c.lock);
for (i = 0; i < NB_MMU_MODES; i++) {
CPUTLBDesc *desc = &cpu->neg.tlb.d[i];
- CPUTLBDescFast *fast = &cpu->neg.tlb.f[i];
+ CPUTLBDescFast *fast = cpu_tlb_fast(cpu, i);
g_free(fast->table);
g_free(desc->fulltlb);
@@ -667,7 +667,7 @@ static void tlb_flush_range_locked(CPUState *cpu, int midx,
unsigned bits)
{
CPUTLBDesc *d = &cpu->neg.tlb.d[midx];
- CPUTLBDescFast *f = &cpu->neg.tlb.f[midx];
+ CPUTLBDescFast *f = cpu_tlb_fast(cpu, midx);
vaddr mask = MAKE_64BIT_MASK(0, bits);
/*
@@ -923,7 +923,7 @@ void tlb_reset_dirty(CPUState *cpu, uintptr_t start, uintptr_t length)
qemu_spin_lock(&cpu->neg.tlb.c.lock);
for (mmu_idx = 0; mmu_idx < NB_MMU_MODES; mmu_idx++) {
CPUTLBDesc *desc = &cpu->neg.tlb.d[mmu_idx];
- CPUTLBDescFast *fast = &cpu->neg.tlb.f[mmu_idx];
+ CPUTLBDescFast *fast = cpu_tlb_fast(cpu, mmu_idx);
unsigned int n = tlb_n_entries(fast);
unsigned int i;
@@ -1316,7 +1316,7 @@ static bool victim_tlb_hit(CPUState *cpu, size_t mmu_idx, size_t index,
if (cmp == page) {
/* Found entry in victim tlb, swap tlb and iotlb. */
- CPUTLBEntry tmptlb, *tlb = &cpu->neg.tlb.f[mmu_idx].table[index];
+ CPUTLBEntry tmptlb, *tlb = &cpu_tlb_fast(cpu, mmu_idx)->table[index];
qemu_spin_lock(&cpu->neg.tlb.c.lock);
copy_tlb_helper_locked(&tmptlb, tlb);
diff --git a/tcg/aarch64/tcg-target.c.inc b/tcg/aarch64/tcg-target.c.inc
index 3b088b7bd9..caf79c742d 100644
--- a/tcg/aarch64/tcg-target.c.inc
+++ b/tcg/aarch64/tcg-target.c.inc
@@ -1668,7 +1668,7 @@ static TCGLabelQemuLdst *prepare_host_addr(TCGContext *s, HostAddress *h,
ldst->oi = oi;
ldst->addr_reg = addr_reg;
- /* Load cpu->neg.tlb.f[mmu_idx].{mask,table} into {tmp0,tmp1}. */
+ /* Load CPUTLBDescFast.{mask,table} into {tmp0,tmp1}. */
QEMU_BUILD_BUG_ON(offsetof(CPUTLBDescFast, mask) != 0);
QEMU_BUILD_BUG_ON(offsetof(CPUTLBDescFast, table) != 8);
tcg_out_insn(s, 3314, LDP, TCG_REG_TMP0, TCG_REG_TMP1, TCG_AREG0,
diff --git a/tcg/arm/tcg-target.c.inc b/tcg/arm/tcg-target.c.inc
index 338c57b061..87ca66bb02 100644
--- a/tcg/arm/tcg-target.c.inc
+++ b/tcg/arm/tcg-target.c.inc
@@ -1421,7 +1421,7 @@ static TCGLabelQemuLdst *prepare_host_addr(TCGContext *s, HostAddress *h,
ldst->oi = oi;
ldst->addr_reg = addr;
- /* Load cpu->neg.tlb.f[mmu_idx].{mask,table} into {r0,r1}. */
+ /* Load CPUTLBDescFast.{mask,table} into {r0,r1}. */
QEMU_BUILD_BUG_ON(offsetof(CPUTLBDescFast, mask) != 0);
QEMU_BUILD_BUG_ON(offsetof(CPUTLBDescFast, table) != 4);
tcg_out_ldrd_8(s, COND_AL, TCG_REG_R0, TCG_AREG0, fast_off);
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 15/76] include/hw/core/cpu: Invert the indexing into CPUTLBDescFast
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (13 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 14/76] include/hw/core/cpu: Introduce cpu_tlb_fast Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 16/76] target/hppa: Adjust mmu indexes to begin with 0 Richard Henderson
` (61 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Philippe Mathieu-Daudé, Pierrick Bouvier
This array is within CPUNegativeOffsetState, which means the
last element of the array has an offset from env with the
smallest magnitude. This can be encoded into fewer bits
when generating TCG fast path memory references.
When we changed the NB_MMU_MODES to be a global constant,
rather than a per-target value, we pessimized the code
generated for targets which use only a few mmu indexes.
By inverting the array index, we counteract that.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
include/hw/core/cpu.h | 11 ++++++++++-
tcg/tcg.c | 3 ++-
2 files changed, 12 insertions(+), 2 deletions(-)
diff --git a/include/hw/core/cpu.h b/include/hw/core/cpu.h
index 30e335d7d0..89fe937a6f 100644
--- a/include/hw/core/cpu.h
+++ b/include/hw/core/cpu.h
@@ -595,9 +595,18 @@ static inline CPUArchState *cpu_env(CPUState *cpu)
}
#ifdef CONFIG_TCG
+/*
+ * Invert the index order of the CPUTLBDescFast array so that lower
+ * mmu_idx have offsets from env with smaller magnitude.
+ */
+static inline int mmuidx_to_fast_index(int mmu_idx)
+{
+ return NB_MMU_MODES - 1 - mmu_idx;
+}
+
static inline CPUTLBDescFast *cpu_tlb_fast(CPUState *cpu, int mmu_idx)
{
- return &cpu->neg.tlb.f[mmu_idx];
+ return &cpu->neg.tlb.f[mmuidx_to_fast_index(mmu_idx)];
}
#endif
diff --git a/tcg/tcg.c b/tcg/tcg.c
index afac55a203..294762c283 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -425,7 +425,8 @@ static uintptr_t G_GNUC_UNUSED get_jmp_target_addr(TCGContext *s, int which)
static int __attribute__((unused))
tlb_mask_table_ofs(TCGContext *s, int which)
{
- return (offsetof(CPUNegativeOffsetState, tlb.f[which]) -
+ int fi = mmuidx_to_fast_index(which);
+ return (offsetof(CPUNegativeOffsetState, tlb.f[fi]) -
sizeof(CPUNegativeOffsetState));
}
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 16/76] target/hppa: Adjust mmu indexes to begin with 0
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (14 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 15/76] include/hw/core/cpu: Invert the indexing into CPUTLBDescFast Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 17/76] include/exec/memopidx: Adjust for 32 mmu indexes Richard Henderson
` (60 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
This is a logical reversion of 2ad04500543, though
there have been additions to the set of mmu indexes
since then. The impetus to that original patch,
"9-15 will use shorter assembler instructions when
run on a x86-64 host" is now handled generically.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/hppa/cpu.h | 28 ++++++++++++++--------------
1 file changed, 14 insertions(+), 14 deletions(-)
diff --git a/target/hppa/cpu.h b/target/hppa/cpu.h
index 11d59d11ca..672ab3750c 100644
--- a/target/hppa/cpu.h
+++ b/target/hppa/cpu.h
@@ -29,21 +29,21 @@
#include "qemu/interval-tree.h"
#include "hw/registerfields.h"
-#define MMU_ABS_W_IDX 6
-#define MMU_ABS_IDX 7
-#define MMU_KERNEL_IDX 8
-#define MMU_KERNEL_P_IDX 9
-#define MMU_PL1_IDX 10
-#define MMU_PL1_P_IDX 11
-#define MMU_PL2_IDX 12
-#define MMU_PL2_P_IDX 13
-#define MMU_USER_IDX 14
-#define MMU_USER_P_IDX 15
+#define MMU_KERNEL_IDX 0
+#define MMU_KERNEL_P_IDX 1
+#define MMU_PL1_IDX 2
+#define MMU_PL1_P_IDX 3
+#define MMU_PL2_IDX 4
+#define MMU_PL2_P_IDX 5
+#define MMU_USER_IDX 6
+#define MMU_USER_P_IDX 7
+#define MMU_ABS_IDX 8
+#define MMU_ABS_W_IDX 9
-#define MMU_IDX_MMU_DISABLED(MIDX) ((MIDX) < MMU_KERNEL_IDX)
-#define MMU_IDX_TO_PRIV(MIDX) (((MIDX) - MMU_KERNEL_IDX) / 2)
-#define MMU_IDX_TO_P(MIDX) (((MIDX) - MMU_KERNEL_IDX) & 1)
-#define PRIV_P_TO_MMU_IDX(PRIV, P) ((PRIV) * 2 + !!(P) + MMU_KERNEL_IDX)
+#define MMU_IDX_MMU_DISABLED(MIDX) ((MIDX) >= MMU_ABS_IDX)
+#define MMU_IDX_TO_PRIV(MIDX) ((MIDX) / 2)
+#define MMU_IDX_TO_P(MIDX) ((MIDX) & 1)
+#define PRIV_P_TO_MMU_IDX(PRIV, P) ((PRIV) * 2 + !!(P))
#define PRIV_KERNEL 0
#define PRIV_USER 3
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 17/76] include/exec/memopidx: Adjust for 32 mmu indexes
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (15 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 16/76] target/hppa: Adjust mmu indexes to begin with 0 Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 18/76] include/hw/core/cpu: Widen MMUIdxMap Richard Henderson
` (59 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
include/exec/memopidx.h | 9 +++++----
1 file changed, 5 insertions(+), 4 deletions(-)
diff --git a/include/exec/memopidx.h b/include/exec/memopidx.h
index eb7f1591a3..66d9c58b3a 100644
--- a/include/exec/memopidx.h
+++ b/include/exec/memopidx.h
@@ -25,9 +25,10 @@ typedef uint32_t MemOpIdx;
static inline MemOpIdx make_memop_idx(MemOp op, unsigned idx)
{
#ifdef CONFIG_DEBUG_TCG
- assert(idx <= 15);
+ assert(idx <= 31);
+ assert(clz32(op) >= 5);
#endif
- return (op << 4) | idx;
+ return (op << 5) | idx;
}
/**
@@ -38,7 +39,7 @@ static inline MemOpIdx make_memop_idx(MemOp op, unsigned idx)
*/
static inline MemOp get_memop(MemOpIdx oi)
{
- return oi >> 4;
+ return oi >> 5;
}
/**
@@ -49,7 +50,7 @@ static inline MemOp get_memop(MemOpIdx oi)
*/
static inline unsigned get_mmuidx(MemOpIdx oi)
{
- return oi & 15;
+ return oi & 31;
}
#endif
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 18/76] include/hw/core/cpu: Widen MMUIdxMap
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (16 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 17/76] include/exec/memopidx: Adjust for 32 mmu indexes Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 19/76] target/arm: Split out mmuidx.h from cpu.h Richard Henderson
` (58 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Widen MMUIdxMap to 32 bits. Do not yet expand NB_MMU_MODES,
but widen the map type in preparation.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
include/hw/core/cpu.h | 2 +-
accel/tcg/cputlb.c | 3 ---
2 files changed, 1 insertion(+), 4 deletions(-)
diff --git a/include/hw/core/cpu.h b/include/hw/core/cpu.h
index 89fe937a6f..83d6e9c24e 100644
--- a/include/hw/core/cpu.h
+++ b/include/hw/core/cpu.h
@@ -201,7 +201,7 @@ struct CPUClass {
* Fix the number of mmu modes to 16.
*/
#define NB_MMU_MODES 16
-typedef uint16_t MMUIdxMap;
+typedef uint32_t MMUIdxMap;
/* Use a fully associative victim tlb of 8 entries. */
#define CPU_VTLB_SIZE 8
diff --git a/accel/tcg/cputlb.c b/accel/tcg/cputlb.c
index 2a6aa01c57..416aaa1040 100644
--- a/accel/tcg/cputlb.c
+++ b/accel/tcg/cputlb.c
@@ -89,9 +89,6 @@
*/
QEMU_BUILD_BUG_ON(sizeof(vaddr) > sizeof(run_on_cpu_data));
-/* We currently can't handle more than 16 bits in the MMUIDX bitmask.
- */
-QEMU_BUILD_BUG_ON(NB_MMU_MODES > 16);
#define ALL_MMUIDX_BITS ((1 << NB_MMU_MODES) - 1)
static inline size_t tlb_n_entries(CPUTLBDescFast *fast)
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 19/76] target/arm: Split out mmuidx.h from cpu.h
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (17 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 18/76] include/hw/core/cpu: Widen MMUIdxMap Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 20/76] target/arm: Convert arm_mmu_idx_to_el from switch to table Richard Henderson
` (57 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Philippe Mathieu-Daudé, Pierrick Bouvier
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/cpu.h | 207 +-----------------------------------------
target/arm/mmuidx.h | 216 ++++++++++++++++++++++++++++++++++++++++++++
2 files changed, 217 insertions(+), 206 deletions(-)
create mode 100644 target/arm/mmuidx.h
diff --git a/target/arm/cpu.h b/target/arm/cpu.h
index 4dd4c6d4bf..d5a45e9b43 100644
--- a/target/arm/cpu.h
+++ b/target/arm/cpu.h
@@ -33,6 +33,7 @@
#include "target/arm/multiprocessing.h"
#include "target/arm/gtimer.h"
#include "target/arm/cpu-sysregs.h"
+#include "target/arm/mmuidx.h"
#define EXCP_UDEF 1 /* undefined instruction */
#define EXCP_SWI 2 /* software interrupt */
@@ -2705,212 +2706,6 @@ bool write_cpustate_to_list(ARMCPU *cpu, bool kvm_sync);
#define TYPE_ARM_HOST_CPU "host-" TYPE_ARM_CPU
-/* ARM has the following "translation regimes" (as the ARM ARM calls them):
- *
- * If EL3 is 64-bit:
- * + NonSecure EL1 & 0 stage 1
- * + NonSecure EL1 & 0 stage 2
- * + NonSecure EL2
- * + NonSecure EL2 & 0 (ARMv8.1-VHE)
- * + Secure EL1 & 0 stage 1
- * + Secure EL1 & 0 stage 2 (FEAT_SEL2)
- * + Secure EL2 (FEAT_SEL2)
- * + Secure EL2 & 0 (FEAT_SEL2)
- * + Realm EL1 & 0 stage 1 (FEAT_RME)
- * + Realm EL1 & 0 stage 2 (FEAT_RME)
- * + Realm EL2 (FEAT_RME)
- * + EL3
- * If EL3 is 32-bit:
- * + NonSecure PL1 & 0 stage 1
- * + NonSecure PL1 & 0 stage 2
- * + NonSecure PL2
- * + Secure PL1 & 0
- * (reminder: for 32 bit EL3, Secure PL1 is *EL3*, not EL1.)
- *
- * For QEMU, an mmu_idx is not quite the same as a translation regime because:
- * 1. we need to split the "EL1 & 0" and "EL2 & 0" regimes into two mmu_idxes,
- * because they may differ in access permissions even if the VA->PA map is
- * the same
- * 2. we want to cache in our TLB the full VA->IPA->PA lookup for a stage 1+2
- * translation, which means that we have one mmu_idx that deals with two
- * concatenated translation regimes [this sort of combined s1+2 TLB is
- * architecturally permitted]
- * 3. we don't need to allocate an mmu_idx to translations that we won't be
- * handling via the TLB. The only way to do a stage 1 translation without
- * the immediate stage 2 translation is via the ATS or AT system insns,
- * which can be slow-pathed and always do a page table walk.
- * The only use of stage 2 translations is either as part of an s1+2
- * lookup or when loading the descriptors during a stage 1 page table walk,
- * and in both those cases we don't use the TLB.
- * 4. we can also safely fold together the "32 bit EL3" and "64 bit EL3"
- * translation regimes, because they map reasonably well to each other
- * and they can't both be active at the same time.
- * 5. we want to be able to use the TLB for accesses done as part of a
- * stage1 page table walk, rather than having to walk the stage2 page
- * table over and over.
- * 6. we need separate EL1/EL2 mmu_idx for handling the Privileged Access
- * Never (PAN) bit within PSTATE.
- * 7. we fold together most secure and non-secure regimes for A-profile,
- * because there are no banked system registers for aarch64, so the
- * process of switching between secure and non-secure is
- * already heavyweight.
- * 8. we cannot fold together Stage 2 Secure and Stage 2 NonSecure,
- * because both are in use simultaneously for Secure EL2.
- *
- * This gives us the following list of cases:
- *
- * EL0 EL1&0 stage 1+2 (aka NS PL0 PL1&0 stage 1+2)
- * EL1 EL1&0 stage 1+2 (aka NS PL1 PL1&0 stage 1+2)
- * EL1 EL1&0 stage 1+2 +PAN (aka NS PL1 P1&0 stage 1+2 +PAN)
- * EL0 EL2&0
- * EL2 EL2&0
- * EL2 EL2&0 +PAN
- * EL2 (aka NS PL2)
- * EL3 (aka AArch32 S PL1 PL1&0)
- * AArch32 S PL0 PL1&0 (we call this EL30_0)
- * AArch32 S PL1 PL1&0 +PAN (we call this EL30_3_PAN)
- * Stage2 Secure
- * Stage2 NonSecure
- * plus one TLB per Physical address space: S, NS, Realm, Root
- *
- * for a total of 16 different mmu_idx.
- *
- * R profile CPUs have an MPU, but can use the same set of MMU indexes
- * as A profile. They only need to distinguish EL0 and EL1 (and
- * EL2 for cores like the Cortex-R52).
- *
- * M profile CPUs are rather different as they do not have a true MMU.
- * They have the following different MMU indexes:
- * User
- * Privileged
- * User, execution priority negative (ie the MPU HFNMIENA bit may apply)
- * Privileged, execution priority negative (ditto)
- * If the CPU supports the v8M Security Extension then there are also:
- * Secure User
- * Secure Privileged
- * Secure User, execution priority negative
- * Secure Privileged, execution priority negative
- *
- * The ARMMMUIdx and the mmu index value used by the core QEMU TLB code
- * are not quite the same -- different CPU types (most notably M profile
- * vs A/R profile) would like to use MMU indexes with different semantics,
- * but since we don't ever need to use all of those in a single CPU we
- * can avoid having to set NB_MMU_MODES to "total number of A profile MMU
- * modes + total number of M profile MMU modes". The lower bits of
- * ARMMMUIdx are the core TLB mmu index, and the higher bits are always
- * the same for any particular CPU.
- * Variables of type ARMMUIdx are always full values, and the core
- * index values are in variables of type 'int'.
- *
- * Our enumeration includes at the end some entries which are not "true"
- * mmu_idx values in that they don't have corresponding TLBs and are only
- * valid for doing slow path page table walks.
- *
- * The constant names here are patterned after the general style of the names
- * of the AT/ATS operations.
- * The values used are carefully arranged to make mmu_idx => EL lookup easy.
- * For M profile we arrange them to have a bit for priv, a bit for negpri
- * and a bit for secure.
- */
-#define ARM_MMU_IDX_A 0x10 /* A profile */
-#define ARM_MMU_IDX_NOTLB 0x20 /* does not have a TLB */
-#define ARM_MMU_IDX_M 0x40 /* M profile */
-
-/* Meanings of the bits for M profile mmu idx values */
-#define ARM_MMU_IDX_M_PRIV 0x1
-#define ARM_MMU_IDX_M_NEGPRI 0x2
-#define ARM_MMU_IDX_M_S 0x4 /* Secure */
-
-#define ARM_MMU_IDX_TYPE_MASK \
- (ARM_MMU_IDX_A | ARM_MMU_IDX_M | ARM_MMU_IDX_NOTLB)
-#define ARM_MMU_IDX_COREIDX_MASK 0xf
-
-typedef enum ARMMMUIdx {
- /*
- * A-profile.
- */
- ARMMMUIdx_E10_0 = 0 | ARM_MMU_IDX_A,
- ARMMMUIdx_E20_0 = 1 | ARM_MMU_IDX_A,
- ARMMMUIdx_E10_1 = 2 | ARM_MMU_IDX_A,
- ARMMMUIdx_E20_2 = 3 | ARM_MMU_IDX_A,
- ARMMMUIdx_E10_1_PAN = 4 | ARM_MMU_IDX_A,
- ARMMMUIdx_E20_2_PAN = 5 | ARM_MMU_IDX_A,
- ARMMMUIdx_E2 = 6 | ARM_MMU_IDX_A,
- ARMMMUIdx_E3 = 7 | ARM_MMU_IDX_A,
- ARMMMUIdx_E30_0 = 8 | ARM_MMU_IDX_A,
- ARMMMUIdx_E30_3_PAN = 9 | ARM_MMU_IDX_A,
-
- /*
- * Used for second stage of an S12 page table walk, or for descriptor
- * loads during first stage of an S1 page table walk. Note that both
- * are in use simultaneously for SecureEL2: the security state for
- * the S2 ptw is selected by the NS bit from the S1 ptw.
- */
- ARMMMUIdx_Stage2_S = 10 | ARM_MMU_IDX_A,
- ARMMMUIdx_Stage2 = 11 | ARM_MMU_IDX_A,
-
- /* TLBs with 1-1 mapping to the physical address spaces. */
- ARMMMUIdx_Phys_S = 12 | ARM_MMU_IDX_A,
- ARMMMUIdx_Phys_NS = 13 | ARM_MMU_IDX_A,
- ARMMMUIdx_Phys_Root = 14 | ARM_MMU_IDX_A,
- ARMMMUIdx_Phys_Realm = 15 | ARM_MMU_IDX_A,
-
- /*
- * These are not allocated TLBs and are used only for AT system
- * instructions or for the first stage of an S12 page table walk.
- */
- ARMMMUIdx_Stage1_E0 = 0 | ARM_MMU_IDX_NOTLB,
- ARMMMUIdx_Stage1_E1 = 1 | ARM_MMU_IDX_NOTLB,
- ARMMMUIdx_Stage1_E1_PAN = 2 | ARM_MMU_IDX_NOTLB,
-
- /*
- * M-profile.
- */
- ARMMMUIdx_MUser = ARM_MMU_IDX_M,
- ARMMMUIdx_MPriv = ARM_MMU_IDX_M | ARM_MMU_IDX_M_PRIV,
- ARMMMUIdx_MUserNegPri = ARMMMUIdx_MUser | ARM_MMU_IDX_M_NEGPRI,
- ARMMMUIdx_MPrivNegPri = ARMMMUIdx_MPriv | ARM_MMU_IDX_M_NEGPRI,
- ARMMMUIdx_MSUser = ARMMMUIdx_MUser | ARM_MMU_IDX_M_S,
- ARMMMUIdx_MSPriv = ARMMMUIdx_MPriv | ARM_MMU_IDX_M_S,
- ARMMMUIdx_MSUserNegPri = ARMMMUIdx_MUserNegPri | ARM_MMU_IDX_M_S,
- ARMMMUIdx_MSPrivNegPri = ARMMMUIdx_MPrivNegPri | ARM_MMU_IDX_M_S,
-} ARMMMUIdx;
-
-/*
- * Bit macros for the core-mmu-index values for each index,
- * for use when calling tlb_flush_by_mmuidx() and friends.
- */
-#define TO_CORE_BIT(NAME) \
- ARMMMUIdxBit_##NAME = 1 << (ARMMMUIdx_##NAME & ARM_MMU_IDX_COREIDX_MASK)
-
-typedef enum ARMMMUIdxBit {
- TO_CORE_BIT(E10_0),
- TO_CORE_BIT(E20_0),
- TO_CORE_BIT(E10_1),
- TO_CORE_BIT(E10_1_PAN),
- TO_CORE_BIT(E2),
- TO_CORE_BIT(E20_2),
- TO_CORE_BIT(E20_2_PAN),
- TO_CORE_BIT(E3),
- TO_CORE_BIT(E30_0),
- TO_CORE_BIT(E30_3_PAN),
- TO_CORE_BIT(Stage2),
- TO_CORE_BIT(Stage2_S),
-
- TO_CORE_BIT(MUser),
- TO_CORE_BIT(MPriv),
- TO_CORE_BIT(MUserNegPri),
- TO_CORE_BIT(MPrivNegPri),
- TO_CORE_BIT(MSUser),
- TO_CORE_BIT(MSPriv),
- TO_CORE_BIT(MSUserNegPri),
- TO_CORE_BIT(MSPrivNegPri),
-} ARMMMUIdxBit;
-
-#undef TO_CORE_BIT
-
-#define MMU_USER_IDX 0
-
/* Indexes used when registering address spaces with cpu_address_space_init */
typedef enum ARMASIdx {
ARMASIdx_NS = 0,
diff --git a/target/arm/mmuidx.h b/target/arm/mmuidx.h
new file mode 100644
index 0000000000..5b9b4bc84f
--- /dev/null
+++ b/target/arm/mmuidx.h
@@ -0,0 +1,216 @@
+/*
+ * QEMU Arm software mmu index definitions
+ * SPDX-License-Identifier: GPL-2.0-or-later
+ */
+
+#ifndef TARGET_ARM_MMUIDX_H
+#define TARGET_ARM_MMUIDX_H
+
+/*
+ * Arm has the following "translation regimes" (as the Arm ARM calls them):
+ *
+ * If EL3 is 64-bit:
+ * + NonSecure EL1 & 0 stage 1
+ * + NonSecure EL1 & 0 stage 2
+ * + NonSecure EL2
+ * + NonSecure EL2 & 0 (ARMv8.1-VHE)
+ * + Secure EL1 & 0 stage 1
+ * + Secure EL1 & 0 stage 2 (FEAT_SEL2)
+ * + Secure EL2 (FEAT_SEL2)
+ * + Secure EL2 & 0 (FEAT_SEL2)
+ * + Realm EL1 & 0 stage 1 (FEAT_RME)
+ * + Realm EL1 & 0 stage 2 (FEAT_RME)
+ * + Realm EL2 (FEAT_RME)
+ * + EL3
+ * If EL3 is 32-bit:
+ * + NonSecure PL1 & 0 stage 1
+ * + NonSecure PL1 & 0 stage 2
+ * + NonSecure PL2
+ * + Secure PL1 & 0
+ * (reminder: for 32 bit EL3, Secure PL1 is *EL3*, not EL1.)
+ *
+ * For QEMU, an mmu_idx is not quite the same as a translation regime because:
+ * 1. we need to split the "EL1 & 0" and "EL2 & 0" regimes into two mmu_idxes,
+ * because they may differ in access permissions even if the VA->PA map is
+ * the same
+ * 2. we want to cache in our TLB the full VA->IPA->PA lookup for a stage 1+2
+ * translation, which means that we have one mmu_idx that deals with two
+ * concatenated translation regimes [this sort of combined s1+2 TLB is
+ * architecturally permitted]
+ * 3. we don't need to allocate an mmu_idx to translations that we won't be
+ * handling via the TLB. The only way to do a stage 1 translation without
+ * the immediate stage 2 translation is via the ATS or AT system insns,
+ * which can be slow-pathed and always do a page table walk.
+ * The only use of stage 2 translations is either as part of an s1+2
+ * lookup or when loading the descriptors during a stage 1 page table walk,
+ * and in both those cases we don't use the TLB.
+ * 4. we can also safely fold together the "32 bit EL3" and "64 bit EL3"
+ * translation regimes, because they map reasonably well to each other
+ * and they can't both be active at the same time.
+ * 5. we want to be able to use the TLB for accesses done as part of a
+ * stage1 page table walk, rather than having to walk the stage2 page
+ * table over and over.
+ * 6. we need separate EL1/EL2 mmu_idx for handling the Privileged Access
+ * Never (PAN) bit within PSTATE.
+ * 7. we fold together most secure and non-secure regimes for A-profile,
+ * because there are no banked system registers for aarch64, so the
+ * process of switching between secure and non-secure is
+ * already heavyweight.
+ * 8. we cannot fold together Stage 2 Secure and Stage 2 NonSecure,
+ * because both are in use simultaneously for Secure EL2.
+ *
+ * This gives us the following list of cases:
+ *
+ * EL0 EL1&0 stage 1+2 (aka NS PL0 PL1&0 stage 1+2)
+ * EL1 EL1&0 stage 1+2 (aka NS PL1 PL1&0 stage 1+2)
+ * EL1 EL1&0 stage 1+2 +PAN (aka NS PL1 P1&0 stage 1+2 +PAN)
+ * EL0 EL2&0
+ * EL2 EL2&0
+ * EL2 EL2&0 +PAN
+ * EL2 (aka NS PL2)
+ * EL3 (aka AArch32 S PL1 PL1&0)
+ * AArch32 S PL0 PL1&0 (we call this EL30_0)
+ * AArch32 S PL1 PL1&0 +PAN (we call this EL30_3_PAN)
+ * Stage2 Secure
+ * Stage2 NonSecure
+ * plus one TLB per Physical address space: S, NS, Realm, Root
+ *
+ * for a total of 16 different mmu_idx.
+ *
+ * R profile CPUs have an MPU, but can use the same set of MMU indexes
+ * as A profile. They only need to distinguish EL0 and EL1 (and
+ * EL2 for cores like the Cortex-R52).
+ *
+ * M profile CPUs are rather different as they do not have a true MMU.
+ * They have the following different MMU indexes:
+ * User
+ * Privileged
+ * User, execution priority negative (ie the MPU HFNMIENA bit may apply)
+ * Privileged, execution priority negative (ditto)
+ * If the CPU supports the v8M Security Extension then there are also:
+ * Secure User
+ * Secure Privileged
+ * Secure User, execution priority negative
+ * Secure Privileged, execution priority negative
+ *
+ * The ARMMMUIdx and the mmu index value used by the core QEMU TLB code
+ * are not quite the same -- different CPU types (most notably M profile
+ * vs A/R profile) would like to use MMU indexes with different semantics,
+ * but since we don't ever need to use all of those in a single CPU we
+ * can avoid having to set NB_MMU_MODES to "total number of A profile MMU
+ * modes + total number of M profile MMU modes". The lower bits of
+ * ARMMMUIdx are the core TLB mmu index, and the higher bits are always
+ * the same for any particular CPU.
+ * Variables of type ARMMUIdx are always full values, and the core
+ * index values are in variables of type 'int'.
+ *
+ * Our enumeration includes at the end some entries which are not "true"
+ * mmu_idx values in that they don't have corresponding TLBs and are only
+ * valid for doing slow path page table walks.
+ *
+ * The constant names here are patterned after the general style of the names
+ * of the AT/ATS operations.
+ * The values used are carefully arranged to make mmu_idx => EL lookup easy.
+ * For M profile we arrange them to have a bit for priv, a bit for negpri
+ * and a bit for secure.
+ */
+#define ARM_MMU_IDX_A 0x10 /* A profile */
+#define ARM_MMU_IDX_NOTLB 0x20 /* does not have a TLB */
+#define ARM_MMU_IDX_M 0x40 /* M profile */
+
+/* Meanings of the bits for M profile mmu idx values */
+#define ARM_MMU_IDX_M_PRIV 0x1
+#define ARM_MMU_IDX_M_NEGPRI 0x2
+#define ARM_MMU_IDX_M_S 0x4 /* Secure */
+
+#define ARM_MMU_IDX_TYPE_MASK \
+ (ARM_MMU_IDX_A | ARM_MMU_IDX_M | ARM_MMU_IDX_NOTLB)
+#define ARM_MMU_IDX_COREIDX_MASK 0xf
+
+typedef enum ARMMMUIdx {
+ /*
+ * A-profile.
+ */
+ ARMMMUIdx_E10_0 = 0 | ARM_MMU_IDX_A,
+ ARMMMUIdx_E20_0 = 1 | ARM_MMU_IDX_A,
+ ARMMMUIdx_E10_1 = 2 | ARM_MMU_IDX_A,
+ ARMMMUIdx_E20_2 = 3 | ARM_MMU_IDX_A,
+ ARMMMUIdx_E10_1_PAN = 4 | ARM_MMU_IDX_A,
+ ARMMMUIdx_E20_2_PAN = 5 | ARM_MMU_IDX_A,
+ ARMMMUIdx_E2 = 6 | ARM_MMU_IDX_A,
+ ARMMMUIdx_E3 = 7 | ARM_MMU_IDX_A,
+ ARMMMUIdx_E30_0 = 8 | ARM_MMU_IDX_A,
+ ARMMMUIdx_E30_3_PAN = 9 | ARM_MMU_IDX_A,
+
+ /*
+ * Used for second stage of an S12 page table walk, or for descriptor
+ * loads during first stage of an S1 page table walk. Note that both
+ * are in use simultaneously for SecureEL2: the security state for
+ * the S2 ptw is selected by the NS bit from the S1 ptw.
+ */
+ ARMMMUIdx_Stage2_S = 10 | ARM_MMU_IDX_A,
+ ARMMMUIdx_Stage2 = 11 | ARM_MMU_IDX_A,
+
+ /* TLBs with 1-1 mapping to the physical address spaces. */
+ ARMMMUIdx_Phys_S = 12 | ARM_MMU_IDX_A,
+ ARMMMUIdx_Phys_NS = 13 | ARM_MMU_IDX_A,
+ ARMMMUIdx_Phys_Root = 14 | ARM_MMU_IDX_A,
+ ARMMMUIdx_Phys_Realm = 15 | ARM_MMU_IDX_A,
+
+ /*
+ * These are not allocated TLBs and are used only for AT system
+ * instructions or for the first stage of an S12 page table walk.
+ */
+ ARMMMUIdx_Stage1_E0 = 0 | ARM_MMU_IDX_NOTLB,
+ ARMMMUIdx_Stage1_E1 = 1 | ARM_MMU_IDX_NOTLB,
+ ARMMMUIdx_Stage1_E1_PAN = 2 | ARM_MMU_IDX_NOTLB,
+
+ /*
+ * M-profile.
+ */
+ ARMMMUIdx_MUser = ARM_MMU_IDX_M,
+ ARMMMUIdx_MPriv = ARM_MMU_IDX_M | ARM_MMU_IDX_M_PRIV,
+ ARMMMUIdx_MUserNegPri = ARMMMUIdx_MUser | ARM_MMU_IDX_M_NEGPRI,
+ ARMMMUIdx_MPrivNegPri = ARMMMUIdx_MPriv | ARM_MMU_IDX_M_NEGPRI,
+ ARMMMUIdx_MSUser = ARMMMUIdx_MUser | ARM_MMU_IDX_M_S,
+ ARMMMUIdx_MSPriv = ARMMMUIdx_MPriv | ARM_MMU_IDX_M_S,
+ ARMMMUIdx_MSUserNegPri = ARMMMUIdx_MUserNegPri | ARM_MMU_IDX_M_S,
+ ARMMMUIdx_MSPrivNegPri = ARMMMUIdx_MPrivNegPri | ARM_MMU_IDX_M_S,
+} ARMMMUIdx;
+
+/*
+ * Bit macros for the core-mmu-index values for each index,
+ * for use when calling tlb_flush_by_mmuidx() and friends.
+ */
+#define TO_CORE_BIT(NAME) \
+ ARMMMUIdxBit_##NAME = 1 << (ARMMMUIdx_##NAME & ARM_MMU_IDX_COREIDX_MASK)
+
+typedef enum ARMMMUIdxBit {
+ TO_CORE_BIT(E10_0),
+ TO_CORE_BIT(E20_0),
+ TO_CORE_BIT(E10_1),
+ TO_CORE_BIT(E10_1_PAN),
+ TO_CORE_BIT(E2),
+ TO_CORE_BIT(E20_2),
+ TO_CORE_BIT(E20_2_PAN),
+ TO_CORE_BIT(E3),
+ TO_CORE_BIT(E30_0),
+ TO_CORE_BIT(E30_3_PAN),
+ TO_CORE_BIT(Stage2),
+ TO_CORE_BIT(Stage2_S),
+
+ TO_CORE_BIT(MUser),
+ TO_CORE_BIT(MPriv),
+ TO_CORE_BIT(MUserNegPri),
+ TO_CORE_BIT(MPrivNegPri),
+ TO_CORE_BIT(MSUser),
+ TO_CORE_BIT(MSPriv),
+ TO_CORE_BIT(MSUserNegPri),
+ TO_CORE_BIT(MSPrivNegPri),
+} ARMMMUIdxBit;
+
+#undef TO_CORE_BIT
+
+#define MMU_USER_IDX 0
+
+#endif /* TARGET_ARM_MMUIDX_H */
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 20/76] target/arm: Convert arm_mmu_idx_to_el from switch to table
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (18 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 19/76] target/arm: Split out mmuidx.h from cpu.h Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 21/76] target/arm: Remove unused env argument from regime_el Richard Henderson
` (56 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Philippe Mathieu-Daudé, Pierrick Bouvier
In an effort to keep all ARMMMUIdx data in one place, begin construction
of an info table describing all of the properties of the mmu_idx. Begin
with the access EL.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/internals.h | 3 +--
target/arm/mmuidx-internal.h | 29 +++++++++++++++++++++++++
target/arm/helper.c | 27 ------------------------
target/arm/mmuidx.c | 41 ++++++++++++++++++++++++++++++++++++
target/arm/meson.build | 7 +++++-
5 files changed, 77 insertions(+), 30 deletions(-)
create mode 100644 target/arm/mmuidx-internal.h
create mode 100644 target/arm/mmuidx.c
diff --git a/target/arm/internals.h b/target/arm/internals.h
index a2ee5f8739..c29682da7d 100644
--- a/target/arm/internals.h
+++ b/target/arm/internals.h
@@ -34,6 +34,7 @@
#include "system/memory.h"
#include "syndrome.h"
#include "cpu-features.h"
+#include "mmuidx-internal.h"
/* register banks for CPU modes */
#define BANK_USRSYS 0
@@ -984,8 +985,6 @@ static inline ARMMMUIdx core_to_aa64_mmu_idx(int mmu_idx)
return mmu_idx | ARM_MMU_IDX_A;
}
-int arm_mmu_idx_to_el(ARMMMUIdx mmu_idx);
-
/* Return the MMU index for a v7M CPU in the specified security state */
ARMMMUIdx arm_v7m_mmu_idx_for_secstate(CPUARMState *env, bool secstate);
diff --git a/target/arm/mmuidx-internal.h b/target/arm/mmuidx-internal.h
new file mode 100644
index 0000000000..29bba4ecb5
--- /dev/null
+++ b/target/arm/mmuidx-internal.h
@@ -0,0 +1,29 @@
+/*
+ * QEMU Arm software mmu index internal definitions
+ * SPDX-License-Identifier: GPL-2.0-or-later
+ */
+
+#ifndef TARGET_ARM_MMUIDX_INTERNAL_H
+#define TARGET_ARM_MMUIDX_INTERNAL_H
+
+#include "mmuidx.h"
+#include "tcg/debug-assert.h"
+#include "hw/registerfields.h"
+
+
+FIELD(MMUIDXINFO, EL, 0, 2)
+FIELD(MMUIDXINFO, ELVALID, 2, 1)
+
+extern const uint32_t arm_mmuidx_table[ARM_MMU_IDX_M + 8];
+
+#define arm_mmuidx_is_valid(x) ((unsigned)(x) < ARRAY_SIZE(arm_mmuidx_table))
+
+/* Return the exception level associated with this mmu index. */
+static inline int arm_mmu_idx_to_el(ARMMMUIdx idx)
+{
+ tcg_debug_assert(arm_mmuidx_is_valid(idx));
+ tcg_debug_assert(FIELD_EX32(arm_mmuidx_table[idx], MMUIDXINFO, ELVALID));
+ return FIELD_EX32(arm_mmuidx_table[idx], MMUIDXINFO, EL);
+}
+
+#endif /* TARGET_ARM_MMUIDX_INTERNAL_H */
diff --git a/target/arm/helper.c b/target/arm/helper.c
index f66868f9ef..ea343e9d48 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -9639,33 +9639,6 @@ int fp_exception_el(CPUARMState *env, int cur_el)
return 0;
}
-/* Return the exception level we're running at if this is our mmu_idx */
-int arm_mmu_idx_to_el(ARMMMUIdx mmu_idx)
-{
- if (mmu_idx & ARM_MMU_IDX_M) {
- return mmu_idx & ARM_MMU_IDX_M_PRIV;
- }
-
- switch (mmu_idx) {
- case ARMMMUIdx_E10_0:
- case ARMMMUIdx_E20_0:
- case ARMMMUIdx_E30_0:
- return 0;
- case ARMMMUIdx_E10_1:
- case ARMMMUIdx_E10_1_PAN:
- return 1;
- case ARMMMUIdx_E2:
- case ARMMMUIdx_E20_2:
- case ARMMMUIdx_E20_2_PAN:
- return 2;
- case ARMMMUIdx_E3:
- case ARMMMUIdx_E30_3_PAN:
- return 3;
- default:
- g_assert_not_reached();
- }
-}
-
#ifndef CONFIG_TCG
ARMMMUIdx arm_v7m_mmu_idx_for_secstate(CPUARMState *env, bool secstate)
{
diff --git a/target/arm/mmuidx.c b/target/arm/mmuidx.c
new file mode 100644
index 0000000000..309b1d68df
--- /dev/null
+++ b/target/arm/mmuidx.c
@@ -0,0 +1,41 @@
+/*
+ * QEMU Arm software mmu index definitions
+ * SPDX-License-Identifier: GPL-2.0-or-later
+ */
+
+#include "qemu/osdep.h"
+#include "mmuidx-internal.h"
+
+
+#define EL(X) ((X << R_MMUIDXINFO_EL_SHIFT) | R_MMUIDXINFO_ELVALID_MASK)
+
+const uint32_t arm_mmuidx_table[ARM_MMU_IDX_M + 8] = {
+ /*
+ * A-profile.
+ */
+ [ARMMMUIdx_E10_0] = EL(0),
+ [ARMMMUIdx_E10_1] = EL(1),
+ [ARMMMUIdx_E10_1_PAN] = EL(1),
+
+ [ARMMMUIdx_E20_0] = EL(0),
+ [ARMMMUIdx_E20_2] = EL(2),
+ [ARMMMUIdx_E20_2_PAN] = EL(2),
+
+ [ARMMMUIdx_E2] = EL(2),
+
+ [ARMMMUIdx_E3] = EL(3),
+ [ARMMMUIdx_E30_0] = EL(0),
+ [ARMMMUIdx_E30_3_PAN] = EL(3),
+
+ /*
+ * M-profile.
+ */
+ [ARMMMUIdx_MUser] = EL(0),
+ [ARMMMUIdx_MPriv] = EL(1),
+ [ARMMMUIdx_MUserNegPri] = EL(0),
+ [ARMMMUIdx_MPrivNegPri] = EL(1),
+ [ARMMMUIdx_MSUser] = EL(0),
+ [ARMMMUIdx_MSPriv] = EL(1),
+ [ARMMMUIdx_MSUserNegPri] = EL(0),
+ [ARMMMUIdx_MSPrivNegPri] = EL(1),
+};
diff --git a/target/arm/meson.build b/target/arm/meson.build
index 07d9271aa4..91630a1f72 100644
--- a/target/arm/meson.build
+++ b/target/arm/meson.build
@@ -6,7 +6,12 @@ arm_ss.add(files(
arm_ss.add(when: 'TARGET_AARCH64', if_true: files(
'cpu64.c',
- 'gdbstub64.c'))
+ 'gdbstub64.c'
+))
+
+arm_common_ss.add(files(
+ 'mmuidx.c',
+))
arm_system_ss = ss.source_set()
arm_common_system_ss = ss.source_set()
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 21/76] target/arm: Remove unused env argument from regime_el
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (19 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 20/76] target/arm: Convert arm_mmu_idx_to_el from switch to table Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 22/76] target/arm: Convert regime_el from switch to table Richard Henderson
` (55 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Philippe Mathieu-Daudé, Pierrick Bouvier
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/internals.h | 8 ++++----
target/arm/helper.c | 2 +-
target/arm/ptw.c | 34 +++++++++++++++++-----------------
target/arm/tcg/mte_helper.c | 2 +-
4 files changed, 23 insertions(+), 23 deletions(-)
diff --git a/target/arm/internals.h b/target/arm/internals.h
index c29682da7d..245c60652f 100644
--- a/target/arm/internals.h
+++ b/target/arm/internals.h
@@ -1074,7 +1074,7 @@ static inline bool regime_is_stage2(ARMMMUIdx mmu_idx)
}
/* Return the exception level which controls this address translation regime */
-static inline uint32_t regime_el(CPUARMState *env, ARMMMUIdx mmu_idx)
+static inline uint32_t regime_el(ARMMMUIdx mmu_idx)
{
switch (mmu_idx) {
case ARMMMUIdx_E20_0:
@@ -1128,7 +1128,7 @@ static inline bool regime_is_user(CPUARMState *env, ARMMMUIdx mmu_idx)
/* Return the SCTLR value which controls this address translation regime */
static inline uint64_t regime_sctlr(CPUARMState *env, ARMMMUIdx mmu_idx)
{
- return env->cp15.sctlr_el[regime_el(env, mmu_idx)];
+ return env->cp15.sctlr_el[regime_el(mmu_idx)];
}
/*
@@ -1160,13 +1160,13 @@ static inline uint64_t regime_tcr(CPUARMState *env, ARMMMUIdx mmu_idx)
v |= env->cp15.vtcr_el2 & VTCR_SHARED_FIELD_MASK;
return v;
}
- return env->cp15.tcr_el[regime_el(env, mmu_idx)];
+ return env->cp15.tcr_el[regime_el(mmu_idx)];
}
/* Return true if the translation regime is using LPAE format page tables */
static inline bool regime_using_lpae_format(CPUARMState *env, ARMMMUIdx mmu_idx)
{
- int el = regime_el(env, mmu_idx);
+ int el = regime_el(mmu_idx);
if (el == 2 || arm_el_is_aa64(env, el)) {
return true;
}
diff --git a/target/arm/helper.c b/target/arm/helper.c
index ea343e9d48..83a675cc22 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -9377,7 +9377,7 @@ ARMVAParameters aa64_va_parameters(CPUARMState *env, uint64_t va,
ARMGranuleSize gran;
ARMCPU *cpu = env_archcpu(env);
bool stage2 = regime_is_stage2(mmu_idx);
- int r_el = regime_el(env, mmu_idx);
+ int r_el = regime_el(mmu_idx);
if (!regime_has_2_ranges(mmu_idx)) {
select = 0;
diff --git a/target/arm/ptw.c b/target/arm/ptw.c
index be3629a745..622b5cf520 100644
--- a/target/arm/ptw.c
+++ b/target/arm/ptw.c
@@ -232,9 +232,9 @@ static uint64_t regime_ttbr(CPUARMState *env, ARMMMUIdx mmu_idx, int ttbrn)
return env->cp15.vsttbr_el2;
}
if (ttbrn == 0) {
- return env->cp15.ttbr0_el[regime_el(env, mmu_idx)];
+ return env->cp15.ttbr0_el[regime_el(mmu_idx)];
} else {
- return env->cp15.ttbr1_el[regime_el(env, mmu_idx)];
+ return env->cp15.ttbr1_el[regime_el(mmu_idx)];
}
}
@@ -1023,7 +1023,7 @@ static bool get_phys_addr_v5(CPUARMState *env, S1Translate *ptw,
}
type = (desc & 3);
domain = (desc >> 5) & 0x0f;
- if (regime_el(env, ptw->in_mmu_idx) == 1) {
+ if (regime_el(ptw->in_mmu_idx) == 1) {
dacr = env->cp15.dacr_ns;
} else {
dacr = env->cp15.dacr_s;
@@ -1162,7 +1162,7 @@ static bool get_phys_addr_v6(CPUARMState *env, S1Translate *ptw,
/* Page or Section. */
domain = (desc >> 5) & 0x0f;
}
- if (regime_el(env, mmu_idx) == 1) {
+ if (regime_el(mmu_idx) == 1) {
dacr = env->cp15.dacr_ns;
} else {
dacr = env->cp15.dacr_s;
@@ -1453,7 +1453,7 @@ static int get_S1prot(CPUARMState *env, ARMMMUIdx mmu_idx, bool is_aa64,
xn = pxn || (user_rw & PAGE_WRITE);
}
} else if (arm_feature(env, ARM_FEATURE_V7)) {
- switch (regime_el(env, mmu_idx)) {
+ switch (regime_el(mmu_idx)) {
case 1:
case 3:
if (is_user) {
@@ -1509,7 +1509,7 @@ static int get_S1prot_indirect(CPUARMState *env, S1Translate *ptw,
/* F */ 0, /* reserved */
};
- uint32_t el = regime_el(env, mmu_idx);
+ uint32_t el = regime_el(mmu_idx);
uint64_t pir = env->cp15.pir_el[el];
uint64_t pire0 = 0;
int perm;
@@ -1584,7 +1584,7 @@ static ARMVAParameters aa32_va_parameters(CPUARMState *env, uint32_t va,
ARMMMUIdx mmu_idx)
{
uint64_t tcr = regime_tcr(env, mmu_idx);
- uint32_t el = regime_el(env, mmu_idx);
+ uint32_t el = regime_el(mmu_idx);
int select, tsz;
bool epd, hpd;
@@ -1810,7 +1810,7 @@ static bool get_phys_addr_lpae(CPUARMState *env, S1Translate *ptw,
int addrsize, inputsize, outputsize;
uint64_t tcr = regime_tcr(env, mmu_idx);
int ap, prot;
- uint32_t el = regime_el(env, mmu_idx);
+ uint32_t el = regime_el(mmu_idx);
uint64_t descaddrmask;
bool aarch64 = arm_el_is_aa64(env, el);
uint64_t descriptor, new_descriptor;
@@ -2264,7 +2264,7 @@ static bool get_phys_addr_lpae(CPUARMState *env, S1Translate *ptw,
/* Index into MAIR registers for cache attributes */
attrindx = extract32(attrs, 2, 3);
- mair = env->cp15.mair_el[regime_el(env, mmu_idx)];
+ mair = env->cp15.mair_el[regime_el(mmu_idx)];
assert(attrindx <= 7);
result->cacheattrs.is_s2_format = false;
result->cacheattrs.attrs = extract64(mair, attrindx * 8, 8);
@@ -2742,7 +2742,7 @@ static bool get_phys_addr_pmsav7(CPUARMState *env,
static uint32_t *regime_rbar(CPUARMState *env, ARMMMUIdx mmu_idx,
uint32_t secure)
{
- if (regime_el(env, mmu_idx) == 2) {
+ if (regime_el(mmu_idx) == 2) {
return env->pmsav8.hprbar;
} else {
return env->pmsav8.rbar[secure];
@@ -2752,7 +2752,7 @@ static uint32_t *regime_rbar(CPUARMState *env, ARMMMUIdx mmu_idx,
static uint32_t *regime_rlar(CPUARMState *env, ARMMMUIdx mmu_idx,
uint32_t secure)
{
- if (regime_el(env, mmu_idx) == 2) {
+ if (regime_el(mmu_idx) == 2) {
return env->pmsav8.hprlar;
} else {
return env->pmsav8.rlar[secure];
@@ -2784,7 +2784,7 @@ bool pmsav8_mpu_lookup(CPUARMState *env, uint32_t address,
uint32_t addr_page_limit = addr_page_base + (TARGET_PAGE_SIZE - 1);
int region_counter;
- if (regime_el(env, mmu_idx) == 2) {
+ if (regime_el(mmu_idx) == 2) {
region_counter = cpu->pmsav8r_hdregion;
} else {
region_counter = cpu->pmsav7_dregion;
@@ -2910,7 +2910,7 @@ bool pmsav8_mpu_lookup(CPUARMState *env, uint32_t address,
xn = 1;
}
- if (regime_el(env, mmu_idx) == 2) {
+ if (regime_el(mmu_idx) == 2) {
result->f.prot = simple_ap_to_rw_prot_is_user(ap,
mmu_idx != ARMMMUIdx_E2);
} else {
@@ -2919,7 +2919,7 @@ bool pmsav8_mpu_lookup(CPUARMState *env, uint32_t address,
if (!arm_feature(env, ARM_FEATURE_M)) {
uint8_t attrindx = extract32(matched_rlar, 1, 3);
- uint64_t mair = env->cp15.mair_el[regime_el(env, mmu_idx)];
+ uint64_t mair = env->cp15.mair_el[regime_el(mmu_idx)];
uint8_t sh = extract32(matched_rlar, 3, 2);
if (regime_sctlr(env, mmu_idx) & SCTLR_WXN &&
@@ -2927,7 +2927,7 @@ bool pmsav8_mpu_lookup(CPUARMState *env, uint32_t address,
xn = 0x1;
}
- if ((regime_el(env, mmu_idx) == 1) &&
+ if ((regime_el(mmu_idx) == 1) &&
regime_sctlr(env, mmu_idx) & SCTLR_UWXN && ap == 0x1) {
pxn = 0x1;
}
@@ -3412,7 +3412,7 @@ static bool get_phys_addr_disabled(CPUARMState *env,
break;
default:
- r_el = regime_el(env, mmu_idx);
+ r_el = regime_el(mmu_idx);
if (arm_el_is_aa64(env, r_el)) {
int pamax = arm_pamax(env_archcpu(env));
uint64_t tcr = env->cp15.tcr_el[r_el];
@@ -3664,7 +3664,7 @@ static bool get_phys_addr_nogpc(CPUARMState *env, S1Translate *ptw,
*/
if (address < 0x02000000 && mmu_idx != ARMMMUIdx_Stage2
&& !arm_feature(env, ARM_FEATURE_V8)) {
- if (regime_el(env, mmu_idx) == 3) {
+ if (regime_el(mmu_idx) == 3) {
address += env->cp15.fcseidr_s;
} else {
address += env->cp15.fcseidr_ns;
diff --git a/target/arm/tcg/mte_helper.c b/target/arm/tcg/mte_helper.c
index 0efc18a181..ed30026b7d 100644
--- a/target/arm/tcg/mte_helper.c
+++ b/target/arm/tcg/mte_helper.c
@@ -604,7 +604,7 @@ void mte_check_fail(CPUARMState *env, uint32_t desc,
int el, reg_el, tcf;
uint64_t sctlr;
- reg_el = regime_el(env, arm_mmu_idx);
+ reg_el = regime_el(arm_mmu_idx);
sctlr = env->cp15.sctlr_el[reg_el];
switch (arm_mmu_idx) {
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 22/76] target/arm: Convert regime_el from switch to table
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (20 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 21/76] target/arm: Remove unused env argument from regime_el Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 23/76] target/arm: Convert regime_has_2_ranges " Richard Henderson
` (54 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/internals.h | 35 ----------------------------
target/arm/mmuidx-internal.h | 13 +++++++++++
target/arm/mmuidx.c | 44 +++++++++++++++++++++---------------
3 files changed, 39 insertions(+), 53 deletions(-)
diff --git a/target/arm/internals.h b/target/arm/internals.h
index 245c60652f..a886282957 100644
--- a/target/arm/internals.h
+++ b/target/arm/internals.h
@@ -1073,41 +1073,6 @@ static inline bool regime_is_stage2(ARMMMUIdx mmu_idx)
return mmu_idx == ARMMMUIdx_Stage2 || mmu_idx == ARMMMUIdx_Stage2_S;
}
-/* Return the exception level which controls this address translation regime */
-static inline uint32_t regime_el(ARMMMUIdx mmu_idx)
-{
- switch (mmu_idx) {
- case ARMMMUIdx_E20_0:
- case ARMMMUIdx_E20_2:
- case ARMMMUIdx_E20_2_PAN:
- case ARMMMUIdx_Stage2:
- case ARMMMUIdx_Stage2_S:
- case ARMMMUIdx_E2:
- return 2;
- case ARMMMUIdx_E3:
- case ARMMMUIdx_E30_0:
- case ARMMMUIdx_E30_3_PAN:
- return 3;
- case ARMMMUIdx_E10_0:
- case ARMMMUIdx_Stage1_E0:
- case ARMMMUIdx_Stage1_E1:
- case ARMMMUIdx_Stage1_E1_PAN:
- case ARMMMUIdx_E10_1:
- case ARMMMUIdx_E10_1_PAN:
- case ARMMMUIdx_MPrivNegPri:
- case ARMMMUIdx_MUserNegPri:
- case ARMMMUIdx_MPriv:
- case ARMMMUIdx_MUser:
- case ARMMMUIdx_MSPrivNegPri:
- case ARMMMUIdx_MSUserNegPri:
- case ARMMMUIdx_MSPriv:
- case ARMMMUIdx_MSUser:
- return 1;
- default:
- g_assert_not_reached();
- }
-}
-
static inline bool regime_is_user(CPUARMState *env, ARMMMUIdx mmu_idx)
{
switch (mmu_idx) {
diff --git a/target/arm/mmuidx-internal.h b/target/arm/mmuidx-internal.h
index 29bba4ecb5..d8d64a14d6 100644
--- a/target/arm/mmuidx-internal.h
+++ b/target/arm/mmuidx-internal.h
@@ -13,6 +13,8 @@
FIELD(MMUIDXINFO, EL, 0, 2)
FIELD(MMUIDXINFO, ELVALID, 2, 1)
+FIELD(MMUIDXINFO, REL, 3, 2)
+FIELD(MMUIDXINFO, RELVALID, 5, 1)
extern const uint32_t arm_mmuidx_table[ARM_MMU_IDX_M + 8];
@@ -26,4 +28,15 @@ static inline int arm_mmu_idx_to_el(ARMMMUIdx idx)
return FIELD_EX32(arm_mmuidx_table[idx], MMUIDXINFO, EL);
}
+/*
+ * Return the exception level for the address translation regime
+ * associated with this mmu index.
+ */
+static inline uint32_t regime_el(ARMMMUIdx idx)
+{
+ tcg_debug_assert(arm_mmuidx_is_valid(idx));
+ tcg_debug_assert(FIELD_EX32(arm_mmuidx_table[idx], MMUIDXINFO, RELVALID));
+ return FIELD_EX32(arm_mmuidx_table[idx], MMUIDXINFO, REL);
+}
+
#endif /* TARGET_ARM_MMUIDX_INTERNAL_H */
diff --git a/target/arm/mmuidx.c b/target/arm/mmuidx.c
index 309b1d68df..6dfefa56c2 100644
--- a/target/arm/mmuidx.c
+++ b/target/arm/mmuidx.c
@@ -8,34 +8,42 @@
#define EL(X) ((X << R_MMUIDXINFO_EL_SHIFT) | R_MMUIDXINFO_ELVALID_MASK)
+#define REL(X) ((X << R_MMUIDXINFO_REL_SHIFT) | R_MMUIDXINFO_RELVALID_MASK)
const uint32_t arm_mmuidx_table[ARM_MMU_IDX_M + 8] = {
/*
* A-profile.
*/
- [ARMMMUIdx_E10_0] = EL(0),
- [ARMMMUIdx_E10_1] = EL(1),
- [ARMMMUIdx_E10_1_PAN] = EL(1),
+ [ARMMMUIdx_E10_0] = EL(0) | REL(1),
+ [ARMMMUIdx_E10_1] = EL(1) | REL(1),
+ [ARMMMUIdx_E10_1_PAN] = EL(1) | REL(1),
- [ARMMMUIdx_E20_0] = EL(0),
- [ARMMMUIdx_E20_2] = EL(2),
- [ARMMMUIdx_E20_2_PAN] = EL(2),
+ [ARMMMUIdx_E20_0] = EL(0) | REL(2),
+ [ARMMMUIdx_E20_2] = EL(2) | REL(2),
+ [ARMMMUIdx_E20_2_PAN] = EL(2) | REL(2),
- [ARMMMUIdx_E2] = EL(2),
+ [ARMMMUIdx_E2] = EL(2) | REL(2),
- [ARMMMUIdx_E3] = EL(3),
- [ARMMMUIdx_E30_0] = EL(0),
- [ARMMMUIdx_E30_3_PAN] = EL(3),
+ [ARMMMUIdx_E3] = EL(3) | REL(3),
+ [ARMMMUIdx_E30_0] = EL(0) | REL(3),
+ [ARMMMUIdx_E30_3_PAN] = EL(3) | REL(3),
+
+ [ARMMMUIdx_Stage2_S] = REL(2),
+ [ARMMMUIdx_Stage2] = REL(2),
+
+ [ARMMMUIdx_Stage1_E0] = REL(1),
+ [ARMMMUIdx_Stage1_E1] = REL(1),
+ [ARMMMUIdx_Stage1_E1_PAN] = REL(1),
/*
* M-profile.
*/
- [ARMMMUIdx_MUser] = EL(0),
- [ARMMMUIdx_MPriv] = EL(1),
- [ARMMMUIdx_MUserNegPri] = EL(0),
- [ARMMMUIdx_MPrivNegPri] = EL(1),
- [ARMMMUIdx_MSUser] = EL(0),
- [ARMMMUIdx_MSPriv] = EL(1),
- [ARMMMUIdx_MSUserNegPri] = EL(0),
- [ARMMMUIdx_MSPrivNegPri] = EL(1),
+ [ARMMMUIdx_MUser] = EL(0) | REL(1),
+ [ARMMMUIdx_MPriv] = EL(1) | REL(1),
+ [ARMMMUIdx_MUserNegPri] = EL(0) | REL(1),
+ [ARMMMUIdx_MPrivNegPri] = EL(1) | REL(1),
+ [ARMMMUIdx_MSUser] = EL(0) | REL(1),
+ [ARMMMUIdx_MSPriv] = EL(1) | REL(1),
+ [ARMMMUIdx_MSUserNegPri] = EL(0) | REL(1),
+ [ARMMMUIdx_MSPrivNegPri] = EL(1) | REL(1),
};
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 23/76] target/arm: Convert regime_has_2_ranges from switch to table
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (21 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 22/76] target/arm: Convert regime_el from switch to table Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 24/76] target/arm: Remove unused env argument from regime_is_pan Richard Henderson
` (53 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/internals.h | 28 ----------------------------
target/arm/mmuidx-internal.h | 17 +++++++++++++++++
target/arm/mmuidx.c | 19 ++++++++++---------
3 files changed, 27 insertions(+), 37 deletions(-)
diff --git a/target/arm/internals.h b/target/arm/internals.h
index a886282957..a25ef90c4c 100644
--- a/target/arm/internals.h
+++ b/target/arm/internals.h
@@ -1027,34 +1027,6 @@ static inline void arm_call_el_change_hook(ARMCPU *cpu)
}
}
-/*
- * Return true if this address translation regime has two ranges.
- * Note that this will not return the correct answer for AArch32
- * Secure PL1&0 (i.e. mmu indexes E3, E30_0, E30_3_PAN), but it is
- * never called from a context where EL3 can be AArch32. (The
- * correct return value for ARMMMUIdx_E3 would be different for
- * that case, so we can't just make the function return the
- * correct value anyway; we would need an extra "bool e3_is_aarch32"
- * argument which all the current callsites would pass as 'false'.)
- */
-static inline bool regime_has_2_ranges(ARMMMUIdx mmu_idx)
-{
- switch (mmu_idx) {
- case ARMMMUIdx_Stage1_E0:
- case ARMMMUIdx_Stage1_E1:
- case ARMMMUIdx_Stage1_E1_PAN:
- case ARMMMUIdx_E10_0:
- case ARMMMUIdx_E10_1:
- case ARMMMUIdx_E10_1_PAN:
- case ARMMMUIdx_E20_0:
- case ARMMMUIdx_E20_2:
- case ARMMMUIdx_E20_2_PAN:
- return true;
- default:
- return false;
- }
-}
-
static inline bool regime_is_pan(CPUARMState *env, ARMMMUIdx mmu_idx)
{
switch (mmu_idx) {
diff --git a/target/arm/mmuidx-internal.h b/target/arm/mmuidx-internal.h
index d8d64a14d6..f03a2ab94c 100644
--- a/target/arm/mmuidx-internal.h
+++ b/target/arm/mmuidx-internal.h
@@ -15,6 +15,7 @@ FIELD(MMUIDXINFO, EL, 0, 2)
FIELD(MMUIDXINFO, ELVALID, 2, 1)
FIELD(MMUIDXINFO, REL, 3, 2)
FIELD(MMUIDXINFO, RELVALID, 5, 1)
+FIELD(MMUIDXINFO, 2RANGES, 6, 1)
extern const uint32_t arm_mmuidx_table[ARM_MMU_IDX_M + 8];
@@ -39,4 +40,20 @@ static inline uint32_t regime_el(ARMMMUIdx idx)
return FIELD_EX32(arm_mmuidx_table[idx], MMUIDXINFO, REL);
}
+/*
+ * Return true if this address translation regime has two ranges.
+ * Note that this will not return the correct answer for AArch32
+ * Secure PL1&0 (i.e. mmu indexes E3, E30_0, E30_3_PAN), but it is
+ * never called from a context where EL3 can be AArch32. (The
+ * correct return value for ARMMMUIdx_E3 would be different for
+ * that case, so we can't just make the function return the
+ * correct value anyway; we would need an extra "bool e3_is_aarch32"
+ * argument which all the current callsites would pass as 'false'.)
+ */
+static inline bool regime_has_2_ranges(ARMMMUIdx idx)
+{
+ tcg_debug_assert(arm_mmuidx_is_valid(idx));
+ return FIELD_EX32(arm_mmuidx_table[idx], MMUIDXINFO, 2RANGES);
+}
+
#endif /* TARGET_ARM_MMUIDX_INTERNAL_H */
diff --git a/target/arm/mmuidx.c b/target/arm/mmuidx.c
index 6dfefa56c2..f880d21606 100644
--- a/target/arm/mmuidx.c
+++ b/target/arm/mmuidx.c
@@ -9,18 +9,19 @@
#define EL(X) ((X << R_MMUIDXINFO_EL_SHIFT) | R_MMUIDXINFO_ELVALID_MASK)
#define REL(X) ((X << R_MMUIDXINFO_REL_SHIFT) | R_MMUIDXINFO_RELVALID_MASK)
+#define R2 R_MMUIDXINFO_2RANGES_MASK
const uint32_t arm_mmuidx_table[ARM_MMU_IDX_M + 8] = {
/*
* A-profile.
*/
- [ARMMMUIdx_E10_0] = EL(0) | REL(1),
- [ARMMMUIdx_E10_1] = EL(1) | REL(1),
- [ARMMMUIdx_E10_1_PAN] = EL(1) | REL(1),
+ [ARMMMUIdx_E10_0] = EL(0) | REL(1) | R2,
+ [ARMMMUIdx_E10_1] = EL(1) | REL(1) | R2,
+ [ARMMMUIdx_E10_1_PAN] = EL(1) | REL(1) | R2,
- [ARMMMUIdx_E20_0] = EL(0) | REL(2),
- [ARMMMUIdx_E20_2] = EL(2) | REL(2),
- [ARMMMUIdx_E20_2_PAN] = EL(2) | REL(2),
+ [ARMMMUIdx_E20_0] = EL(0) | REL(2) | R2,
+ [ARMMMUIdx_E20_2] = EL(2) | REL(2) | R2,
+ [ARMMMUIdx_E20_2_PAN] = EL(2) | REL(2) | R2,
[ARMMMUIdx_E2] = EL(2) | REL(2),
@@ -31,9 +32,9 @@ const uint32_t arm_mmuidx_table[ARM_MMU_IDX_M + 8] = {
[ARMMMUIdx_Stage2_S] = REL(2),
[ARMMMUIdx_Stage2] = REL(2),
- [ARMMMUIdx_Stage1_E0] = REL(1),
- [ARMMMUIdx_Stage1_E1] = REL(1),
- [ARMMMUIdx_Stage1_E1_PAN] = REL(1),
+ [ARMMMUIdx_Stage1_E0] = REL(1) | R2,
+ [ARMMMUIdx_Stage1_E1] = REL(1) | R2,
+ [ARMMMUIdx_Stage1_E1_PAN] = REL(1) | R2,
/*
* M-profile.
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 24/76] target/arm: Remove unused env argument from regime_is_pan
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (22 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 23/76] target/arm: Convert regime_has_2_ranges " Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 25/76] target/arm: Convert regime_is_pan from switch to table Richard Henderson
` (52 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Philippe Mathieu-Daudé, Pierrick Bouvier
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/internals.h | 2 +-
target/arm/ptw.c | 6 +++---
2 files changed, 4 insertions(+), 4 deletions(-)
diff --git a/target/arm/internals.h b/target/arm/internals.h
index a25ef90c4c..b32c31779c 100644
--- a/target/arm/internals.h
+++ b/target/arm/internals.h
@@ -1027,7 +1027,7 @@ static inline void arm_call_el_change_hook(ARMCPU *cpu)
}
}
-static inline bool regime_is_pan(CPUARMState *env, ARMMMUIdx mmu_idx)
+static inline bool regime_is_pan(ARMMMUIdx mmu_idx)
{
switch (mmu_idx) {
case ARMMMUIdx_Stage1_E1_PAN:
diff --git a/target/arm/ptw.c b/target/arm/ptw.c
index 622b5cf520..db51e55cd2 100644
--- a/target/arm/ptw.c
+++ b/target/arm/ptw.c
@@ -1393,10 +1393,10 @@ static int get_S1prot(CPUARMState *env, ARMMMUIdx mmu_idx, bool is_aa64,
* We make the IMPDEF choices that SCR_EL3.SIF and Realm EL2&0
* do not affect EPAN.
*/
- if (user_rw && regime_is_pan(env, mmu_idx)) {
+ if (user_rw && regime_is_pan(mmu_idx)) {
prot_rw = 0;
} else if (cpu_isar_feature(aa64_pan3, cpu) && is_aa64 &&
- regime_is_pan(env, mmu_idx) &&
+ regime_is_pan(mmu_idx) &&
(regime_sctlr(env, mmu_idx) & SCTLR_EPAN) && !xn) {
prot_rw = 0;
}
@@ -1535,7 +1535,7 @@ static int get_S1prot_indirect(CPUARMState *env, S1Translate *ptw,
p_perm &= ~(PAGE_RWX | PAGE_GCS);
u_perm &= ~(PAGE_RWX | PAGE_GCS);
}
- if ((u_perm & (PAGE_RWX | PAGE_GCS)) && regime_is_pan(env, mmu_idx)) {
+ if ((u_perm & (PAGE_RWX | PAGE_GCS)) && regime_is_pan(mmu_idx)) {
p_perm &= ~(PAGE_READ | PAGE_WRITE);
}
perm = regime_is_user(env, mmu_idx) ? u_perm : p_perm;
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 25/76] target/arm: Convert regime_is_pan from switch to table
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (23 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 24/76] target/arm: Remove unused env argument from regime_is_pan Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 26/76] target/arm: Remove unused env argument from regime_is_user Richard Henderson
` (51 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/internals.h | 13 -------------
target/arm/mmuidx-internal.h | 8 ++++++++
target/arm/mmuidx.c | 9 +++++----
3 files changed, 13 insertions(+), 17 deletions(-)
diff --git a/target/arm/internals.h b/target/arm/internals.h
index b32c31779c..b180499ecf 100644
--- a/target/arm/internals.h
+++ b/target/arm/internals.h
@@ -1027,19 +1027,6 @@ static inline void arm_call_el_change_hook(ARMCPU *cpu)
}
}
-static inline bool regime_is_pan(ARMMMUIdx mmu_idx)
-{
- switch (mmu_idx) {
- case ARMMMUIdx_Stage1_E1_PAN:
- case ARMMMUIdx_E10_1_PAN:
- case ARMMMUIdx_E20_2_PAN:
- case ARMMMUIdx_E30_3_PAN:
- return true;
- default:
- return false;
- }
-}
-
static inline bool regime_is_stage2(ARMMMUIdx mmu_idx)
{
return mmu_idx == ARMMMUIdx_Stage2 || mmu_idx == ARMMMUIdx_Stage2_S;
diff --git a/target/arm/mmuidx-internal.h b/target/arm/mmuidx-internal.h
index f03a2ab94c..41baf1a003 100644
--- a/target/arm/mmuidx-internal.h
+++ b/target/arm/mmuidx-internal.h
@@ -16,6 +16,7 @@ FIELD(MMUIDXINFO, ELVALID, 2, 1)
FIELD(MMUIDXINFO, REL, 3, 2)
FIELD(MMUIDXINFO, RELVALID, 5, 1)
FIELD(MMUIDXINFO, 2RANGES, 6, 1)
+FIELD(MMUIDXINFO, PAN, 7, 1)
extern const uint32_t arm_mmuidx_table[ARM_MMU_IDX_M + 8];
@@ -56,4 +57,11 @@ static inline bool regime_has_2_ranges(ARMMMUIdx idx)
return FIELD_EX32(arm_mmuidx_table[idx], MMUIDXINFO, 2RANGES);
}
+/* Return true if Privileged Access Never is enabled for this mmu index. */
+static inline bool regime_is_pan(ARMMMUIdx idx)
+{
+ tcg_debug_assert(arm_mmuidx_is_valid(idx));
+ return FIELD_EX32(arm_mmuidx_table[idx], MMUIDXINFO, PAN);
+}
+
#endif /* TARGET_ARM_MMUIDX_INTERNAL_H */
diff --git a/target/arm/mmuidx.c b/target/arm/mmuidx.c
index f880d21606..98db02b8e5 100644
--- a/target/arm/mmuidx.c
+++ b/target/arm/mmuidx.c
@@ -10,6 +10,7 @@
#define EL(X) ((X << R_MMUIDXINFO_EL_SHIFT) | R_MMUIDXINFO_ELVALID_MASK)
#define REL(X) ((X << R_MMUIDXINFO_REL_SHIFT) | R_MMUIDXINFO_RELVALID_MASK)
#define R2 R_MMUIDXINFO_2RANGES_MASK
+#define PAN R_MMUIDXINFO_PAN_MASK
const uint32_t arm_mmuidx_table[ARM_MMU_IDX_M + 8] = {
/*
@@ -17,24 +18,24 @@ const uint32_t arm_mmuidx_table[ARM_MMU_IDX_M + 8] = {
*/
[ARMMMUIdx_E10_0] = EL(0) | REL(1) | R2,
[ARMMMUIdx_E10_1] = EL(1) | REL(1) | R2,
- [ARMMMUIdx_E10_1_PAN] = EL(1) | REL(1) | R2,
+ [ARMMMUIdx_E10_1_PAN] = EL(1) | REL(1) | R2 | PAN,
[ARMMMUIdx_E20_0] = EL(0) | REL(2) | R2,
[ARMMMUIdx_E20_2] = EL(2) | REL(2) | R2,
- [ARMMMUIdx_E20_2_PAN] = EL(2) | REL(2) | R2,
+ [ARMMMUIdx_E20_2_PAN] = EL(2) | REL(2) | R2 | PAN,
[ARMMMUIdx_E2] = EL(2) | REL(2),
[ARMMMUIdx_E3] = EL(3) | REL(3),
[ARMMMUIdx_E30_0] = EL(0) | REL(3),
- [ARMMMUIdx_E30_3_PAN] = EL(3) | REL(3),
+ [ARMMMUIdx_E30_3_PAN] = EL(3) | REL(3) | PAN,
[ARMMMUIdx_Stage2_S] = REL(2),
[ARMMMUIdx_Stage2] = REL(2),
[ARMMMUIdx_Stage1_E0] = REL(1) | R2,
[ARMMMUIdx_Stage1_E1] = REL(1) | R2,
- [ARMMMUIdx_Stage1_E1_PAN] = REL(1) | R2,
+ [ARMMMUIdx_Stage1_E1_PAN] = REL(1) | R2 | PAN,
/*
* M-profile.
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 26/76] target/arm: Remove unused env argument from regime_is_user
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (24 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 25/76] target/arm: Convert regime_is_pan from switch to table Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 27/76] target/arm: Convert regime_is_user from switch to table Richard Henderson
` (50 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Philippe Mathieu-Daudé, Pierrick Bouvier
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/internals.h | 2 +-
target/arm/helper.c | 2 +-
target/arm/ptw.c | 16 ++++++++--------
3 files changed, 10 insertions(+), 10 deletions(-)
diff --git a/target/arm/internals.h b/target/arm/internals.h
index b180499ecf..ad5584eb0b 100644
--- a/target/arm/internals.h
+++ b/target/arm/internals.h
@@ -1032,7 +1032,7 @@ static inline bool regime_is_stage2(ARMMMUIdx mmu_idx)
return mmu_idx == ARMMMUIdx_Stage2 || mmu_idx == ARMMMUIdx_Stage2_S;
}
-static inline bool regime_is_user(CPUARMState *env, ARMMMUIdx mmu_idx)
+static inline bool regime_is_user(ARMMMUIdx mmu_idx)
{
switch (mmu_idx) {
case ARMMMUIdx_E10_0:
diff --git a/target/arm/helper.c b/target/arm/helper.c
index 83a675cc22..c0e808a9e8 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -9436,7 +9436,7 @@ ARMVAParameters aa64_va_parameters(CPUARMState *env, uint64_t va,
ds = extract64(tcr, 59, 1);
if (e0pd && cpu_isar_feature(aa64_e0pd, cpu) &&
- regime_is_user(env, mmu_idx)) {
+ regime_is_user(mmu_idx)) {
epd = true;
}
diff --git a/target/arm/ptw.c b/target/arm/ptw.c
index db51e55cd2..b72c1fc348 100644
--- a/target/arm/ptw.c
+++ b/target/arm/ptw.c
@@ -964,7 +964,7 @@ static int ap_to_rw_prot(CPUARMState *env, ARMMMUIdx mmu_idx,
int ap, int domain_prot)
{
return ap_to_rw_prot_is_user(env, mmu_idx, ap, domain_prot,
- regime_is_user(env, mmu_idx));
+ regime_is_user(mmu_idx));
}
/*
@@ -990,7 +990,7 @@ static int simple_ap_to_rw_prot_is_user(int ap, bool is_user)
static int simple_ap_to_rw_prot(CPUARMState *env, ARMMMUIdx mmu_idx, int ap)
{
- return simple_ap_to_rw_prot_is_user(ap, regime_is_user(env, mmu_idx));
+ return simple_ap_to_rw_prot_is_user(ap, regime_is_user(mmu_idx));
}
static bool get_phys_addr_v5(CPUARMState *env, S1Translate *ptw,
@@ -1376,7 +1376,7 @@ static int get_S1prot(CPUARMState *env, ARMMMUIdx mmu_idx, bool is_aa64,
ARMSecuritySpace in_pa, ARMSecuritySpace out_pa)
{
ARMCPU *cpu = env_archcpu(env);
- bool is_user = regime_is_user(env, mmu_idx);
+ bool is_user = regime_is_user(mmu_idx);
bool have_wxn;
int wxn = 0;
@@ -1538,7 +1538,7 @@ static int get_S1prot_indirect(CPUARMState *env, S1Translate *ptw,
if ((u_perm & (PAGE_RWX | PAGE_GCS)) && regime_is_pan(mmu_idx)) {
p_perm &= ~(PAGE_READ | PAGE_WRITE);
}
- perm = regime_is_user(env, mmu_idx) ? u_perm : p_perm;
+ perm = regime_is_user(mmu_idx) ? u_perm : p_perm;
}
if (in_pa != out_pa) {
@@ -2389,7 +2389,7 @@ static bool get_phys_addr_pmsav5(CPUARMState *env,
uint32_t mask;
uint32_t base;
ARMMMUIdx mmu_idx = ptw->in_mmu_idx;
- bool is_user = regime_is_user(env, mmu_idx);
+ bool is_user = regime_is_user(mmu_idx);
if (regime_translation_disabled(env, mmu_idx, ptw->in_space)) {
/* MPU disabled. */
@@ -2556,7 +2556,7 @@ static bool get_phys_addr_pmsav7(CPUARMState *env,
ARMCPU *cpu = env_archcpu(env);
int n;
ARMMMUIdx mmu_idx = ptw->in_mmu_idx;
- bool is_user = regime_is_user(env, mmu_idx);
+ bool is_user = regime_is_user(mmu_idx);
bool secure = arm_space_is_secure(ptw->in_space);
result->f.phys_addr = address;
@@ -2776,7 +2776,7 @@ bool pmsav8_mpu_lookup(CPUARMState *env, uint32_t address,
* memory system to use a subpage.
*/
ARMCPU *cpu = env_archcpu(env);
- bool is_user = regime_is_user(env, mmu_idx);
+ bool is_user = regime_is_user(mmu_idx);
int n;
int matchregion = -1;
bool hit = false;
@@ -3656,7 +3656,7 @@ static bool get_phys_addr_nogpc(CPUARMState *env, S1Translate *ptw,
break;
}
- result->f.attrs.user = regime_is_user(env, mmu_idx);
+ result->f.attrs.user = regime_is_user(mmu_idx);
/*
* Fast Context Switch Extension. This doesn't exist at all in v8.
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 27/76] target/arm: Convert regime_is_user from switch to table
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (25 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 26/76] target/arm: Remove unused env argument from regime_is_user Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 28/76] target/arm: Convert arm_mmu_idx_is_stage1_of_2 " Richard Henderson
` (49 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Philippe Mathieu-Daudé, Pierrick Bouvier
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/internals.h | 17 -----------------
target/arm/mmuidx-internal.h | 12 ++++++++++++
target/arm/mmuidx.c | 6 ++++--
3 files changed, 16 insertions(+), 19 deletions(-)
diff --git a/target/arm/internals.h b/target/arm/internals.h
index ad5584eb0b..afdc5a92ae 100644
--- a/target/arm/internals.h
+++ b/target/arm/internals.h
@@ -1032,23 +1032,6 @@ static inline bool regime_is_stage2(ARMMMUIdx mmu_idx)
return mmu_idx == ARMMMUIdx_Stage2 || mmu_idx == ARMMMUIdx_Stage2_S;
}
-static inline bool regime_is_user(ARMMMUIdx mmu_idx)
-{
- switch (mmu_idx) {
- case ARMMMUIdx_E10_0:
- case ARMMMUIdx_E20_0:
- case ARMMMUIdx_E30_0:
- case ARMMMUIdx_Stage1_E0:
- case ARMMMUIdx_MUser:
- case ARMMMUIdx_MSUser:
- case ARMMMUIdx_MUserNegPri:
- case ARMMMUIdx_MSUserNegPri:
- return true;
- default:
- return false;
- }
-}
-
/* Return the SCTLR value which controls this address translation regime */
static inline uint64_t regime_sctlr(CPUARMState *env, ARMMMUIdx mmu_idx)
{
diff --git a/target/arm/mmuidx-internal.h b/target/arm/mmuidx-internal.h
index 41baf1a003..3e51c0f579 100644
--- a/target/arm/mmuidx-internal.h
+++ b/target/arm/mmuidx-internal.h
@@ -17,6 +17,7 @@ FIELD(MMUIDXINFO, REL, 3, 2)
FIELD(MMUIDXINFO, RELVALID, 5, 1)
FIELD(MMUIDXINFO, 2RANGES, 6, 1)
FIELD(MMUIDXINFO, PAN, 7, 1)
+FIELD(MMUIDXINFO, USER, 8, 1)
extern const uint32_t arm_mmuidx_table[ARM_MMU_IDX_M + 8];
@@ -64,4 +65,15 @@ static inline bool regime_is_pan(ARMMMUIdx idx)
return FIELD_EX32(arm_mmuidx_table[idx], MMUIDXINFO, PAN);
}
+/*
+ * Return true if the exception level associated with this mmu index is 0.
+ * Differs from arm_mmu_idx_to_el(idx) == 0 in that this allows querying
+ * Stage1 and Stage2 mmu indexes.
+ */
+static inline bool regime_is_user(ARMMMUIdx idx)
+{
+ tcg_debug_assert(arm_mmuidx_is_valid(idx));
+ return FIELD_EX32(arm_mmuidx_table[idx], MMUIDXINFO, USER);
+}
+
#endif /* TARGET_ARM_MMUIDX_INTERNAL_H */
diff --git a/target/arm/mmuidx.c b/target/arm/mmuidx.c
index 98db02b8e5..1c1e062bfe 100644
--- a/target/arm/mmuidx.c
+++ b/target/arm/mmuidx.c
@@ -7,10 +7,12 @@
#include "mmuidx-internal.h"
-#define EL(X) ((X << R_MMUIDXINFO_EL_SHIFT) | R_MMUIDXINFO_ELVALID_MASK)
+#define EL(X) ((X << R_MMUIDXINFO_EL_SHIFT) | R_MMUIDXINFO_ELVALID_MASK | \
+ ((X == 0) << R_MMUIDXINFO_USER_SHIFT))
#define REL(X) ((X << R_MMUIDXINFO_REL_SHIFT) | R_MMUIDXINFO_RELVALID_MASK)
#define R2 R_MMUIDXINFO_2RANGES_MASK
#define PAN R_MMUIDXINFO_PAN_MASK
+#define USER R_MMUIDXINFO_USER_MASK
const uint32_t arm_mmuidx_table[ARM_MMU_IDX_M + 8] = {
/*
@@ -33,7 +35,7 @@ const uint32_t arm_mmuidx_table[ARM_MMU_IDX_M + 8] = {
[ARMMMUIdx_Stage2_S] = REL(2),
[ARMMMUIdx_Stage2] = REL(2),
- [ARMMMUIdx_Stage1_E0] = REL(1) | R2,
+ [ARMMMUIdx_Stage1_E0] = REL(1) | R2 | USER,
[ARMMMUIdx_Stage1_E1] = REL(1) | R2,
[ARMMMUIdx_Stage1_E1_PAN] = REL(1) | R2 | PAN,
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 28/76] target/arm: Convert arm_mmu_idx_is_stage1_of_2 from switch to table
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (26 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 27/76] target/arm: Convert regime_is_user from switch to table Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 29/76] target/arm: Convert regime_is_stage2 " Richard Henderson
` (48 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Philippe Mathieu-Daudé, Pierrick Bouvier
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/internals.h | 19 -------------------
target/arm/mmuidx-internal.h | 8 ++++++++
target/arm/mmuidx.c | 7 ++++---
3 files changed, 12 insertions(+), 22 deletions(-)
diff --git a/target/arm/internals.h b/target/arm/internals.h
index afdc5a92ae..771f7748de 100644
--- a/target/arm/internals.h
+++ b/target/arm/internals.h
@@ -1280,25 +1280,6 @@ ARMMMUIdx stage_1_mmu_idx(ARMMMUIdx mmu_idx);
ARMMMUIdx arm_stage1_mmu_idx(CPUARMState *env);
#endif
-/**
- * arm_mmu_idx_is_stage1_of_2:
- * @mmu_idx: The ARMMMUIdx to test
- *
- * Return true if @mmu_idx is a NOTLB mmu_idx that is the
- * first stage of a two stage regime.
- */
-static inline bool arm_mmu_idx_is_stage1_of_2(ARMMMUIdx mmu_idx)
-{
- switch (mmu_idx) {
- case ARMMMUIdx_Stage1_E0:
- case ARMMMUIdx_Stage1_E1:
- case ARMMMUIdx_Stage1_E1_PAN:
- return true;
- default:
- return false;
- }
-}
-
static inline uint32_t aarch32_cpsr_valid_mask(uint64_t features,
const ARMISARegisters *id)
{
diff --git a/target/arm/mmuidx-internal.h b/target/arm/mmuidx-internal.h
index 3e51c0f579..55fba4aae2 100644
--- a/target/arm/mmuidx-internal.h
+++ b/target/arm/mmuidx-internal.h
@@ -18,6 +18,7 @@ FIELD(MMUIDXINFO, RELVALID, 5, 1)
FIELD(MMUIDXINFO, 2RANGES, 6, 1)
FIELD(MMUIDXINFO, PAN, 7, 1)
FIELD(MMUIDXINFO, USER, 8, 1)
+FIELD(MMUIDXINFO, STAGE1, 9, 1)
extern const uint32_t arm_mmuidx_table[ARM_MMU_IDX_M + 8];
@@ -76,4 +77,11 @@ static inline bool regime_is_user(ARMMMUIdx idx)
return FIELD_EX32(arm_mmuidx_table[idx], MMUIDXINFO, USER);
}
+/* Return true if this mmu index is stage 1 of a 2-stage translation. */
+static inline bool arm_mmu_idx_is_stage1_of_2(ARMMMUIdx idx)
+{
+ tcg_debug_assert(arm_mmuidx_is_valid(idx));
+ return FIELD_EX32(arm_mmuidx_table[idx], MMUIDXINFO, STAGE1);
+}
+
#endif /* TARGET_ARM_MMUIDX_INTERNAL_H */
diff --git a/target/arm/mmuidx.c b/target/arm/mmuidx.c
index 1c1e062bfe..c5b43a5932 100644
--- a/target/arm/mmuidx.c
+++ b/target/arm/mmuidx.c
@@ -13,6 +13,7 @@
#define R2 R_MMUIDXINFO_2RANGES_MASK
#define PAN R_MMUIDXINFO_PAN_MASK
#define USER R_MMUIDXINFO_USER_MASK
+#define S1 R_MMUIDXINFO_STAGE1_MASK
const uint32_t arm_mmuidx_table[ARM_MMU_IDX_M + 8] = {
/*
@@ -35,9 +36,9 @@ const uint32_t arm_mmuidx_table[ARM_MMU_IDX_M + 8] = {
[ARMMMUIdx_Stage2_S] = REL(2),
[ARMMMUIdx_Stage2] = REL(2),
- [ARMMMUIdx_Stage1_E0] = REL(1) | R2 | USER,
- [ARMMMUIdx_Stage1_E1] = REL(1) | R2,
- [ARMMMUIdx_Stage1_E1_PAN] = REL(1) | R2 | PAN,
+ [ARMMMUIdx_Stage1_E0] = REL(1) | R2 | S1 | USER,
+ [ARMMMUIdx_Stage1_E1] = REL(1) | R2 | S1,
+ [ARMMMUIdx_Stage1_E1_PAN] = REL(1) | R2 | S1 | PAN,
/*
* M-profile.
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 29/76] target/arm: Convert regime_is_stage2 to table
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (27 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 28/76] target/arm: Convert arm_mmu_idx_is_stage1_of_2 " Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 30/76] target/arm: Introduce mmu indexes for GCS Richard Henderson
` (47 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Philippe Mathieu-Daudé, Pierrick Bouvier
This wasn't using a switch, but two comparisons.
Convert it to arm_mmuidx_table for consistency.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/internals.h | 5 -----
target/arm/mmuidx-internal.h | 8 ++++++++
target/arm/mmuidx.c | 5 +++--
3 files changed, 11 insertions(+), 7 deletions(-)
diff --git a/target/arm/internals.h b/target/arm/internals.h
index 771f7748de..dc6bf3548e 100644
--- a/target/arm/internals.h
+++ b/target/arm/internals.h
@@ -1027,11 +1027,6 @@ static inline void arm_call_el_change_hook(ARMCPU *cpu)
}
}
-static inline bool regime_is_stage2(ARMMMUIdx mmu_idx)
-{
- return mmu_idx == ARMMMUIdx_Stage2 || mmu_idx == ARMMMUIdx_Stage2_S;
-}
-
/* Return the SCTLR value which controls this address translation regime */
static inline uint64_t regime_sctlr(CPUARMState *env, ARMMMUIdx mmu_idx)
{
diff --git a/target/arm/mmuidx-internal.h b/target/arm/mmuidx-internal.h
index 55fba4aae2..1d948aa6f4 100644
--- a/target/arm/mmuidx-internal.h
+++ b/target/arm/mmuidx-internal.h
@@ -19,6 +19,7 @@ FIELD(MMUIDXINFO, 2RANGES, 6, 1)
FIELD(MMUIDXINFO, PAN, 7, 1)
FIELD(MMUIDXINFO, USER, 8, 1)
FIELD(MMUIDXINFO, STAGE1, 9, 1)
+FIELD(MMUIDXINFO, STAGE2, 10, 1)
extern const uint32_t arm_mmuidx_table[ARM_MMU_IDX_M + 8];
@@ -84,4 +85,11 @@ static inline bool arm_mmu_idx_is_stage1_of_2(ARMMMUIdx idx)
return FIELD_EX32(arm_mmuidx_table[idx], MMUIDXINFO, STAGE1);
}
+/* Return true if this mmu index is stage 2 of a 2-stage translation. */
+static inline bool regime_is_stage2(ARMMMUIdx idx)
+{
+ tcg_debug_assert(arm_mmuidx_is_valid(idx));
+ return FIELD_EX32(arm_mmuidx_table[idx], MMUIDXINFO, STAGE2);
+}
+
#endif /* TARGET_ARM_MMUIDX_INTERNAL_H */
diff --git a/target/arm/mmuidx.c b/target/arm/mmuidx.c
index c5b43a5932..61a682e655 100644
--- a/target/arm/mmuidx.c
+++ b/target/arm/mmuidx.c
@@ -14,6 +14,7 @@
#define PAN R_MMUIDXINFO_PAN_MASK
#define USER R_MMUIDXINFO_USER_MASK
#define S1 R_MMUIDXINFO_STAGE1_MASK
+#define S2 R_MMUIDXINFO_STAGE2_MASK
const uint32_t arm_mmuidx_table[ARM_MMU_IDX_M + 8] = {
/*
@@ -33,8 +34,8 @@ const uint32_t arm_mmuidx_table[ARM_MMU_IDX_M + 8] = {
[ARMMMUIdx_E30_0] = EL(0) | REL(3),
[ARMMMUIdx_E30_3_PAN] = EL(3) | REL(3) | PAN,
- [ARMMMUIdx_Stage2_S] = REL(2),
- [ARMMMUIdx_Stage2] = REL(2),
+ [ARMMMUIdx_Stage2_S] = REL(2) | S2,
+ [ARMMMUIdx_Stage2] = REL(2) | S2,
[ARMMMUIdx_Stage1_E0] = REL(1) | R2 | S1 | USER,
[ARMMMUIdx_Stage1_E1] = REL(1) | R2 | S1,
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 30/76] target/arm: Introduce mmu indexes for GCS
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (28 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 29/76] target/arm: Convert regime_is_stage2 " Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 31/76] target/arm: Introduce regime_to_gcs Richard Henderson
` (46 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
include/hw/core/cpu.h | 5 ++-
target/arm/mmuidx-internal.h | 8 ++++
target/arm/mmuidx.h | 71 ++++++++++++++++++++++++------------
target/arm/helper.c | 15 ++++++--
target/arm/mmuidx.c | 9 +++++
target/arm/ptw.c | 20 ++++++++++
target/arm/tcg/tlb-insns.c | 47 ++++++++++++++++--------
7 files changed, 132 insertions(+), 43 deletions(-)
diff --git a/include/hw/core/cpu.h b/include/hw/core/cpu.h
index 83d6e9c24e..aad4ace27c 100644
--- a/include/hw/core/cpu.h
+++ b/include/hw/core/cpu.h
@@ -198,9 +198,10 @@ struct CPUClass {
};
/*
- * Fix the number of mmu modes to 16.
+ * Fix the number of mmu modes across all targets.
+ * Current maximum is target/arm/.
*/
-#define NB_MMU_MODES 16
+#define NB_MMU_MODES 22
typedef uint32_t MMUIdxMap;
/* Use a fully associative victim tlb of 8 entries. */
diff --git a/target/arm/mmuidx-internal.h b/target/arm/mmuidx-internal.h
index 1d948aa6f4..f494ec348d 100644
--- a/target/arm/mmuidx-internal.h
+++ b/target/arm/mmuidx-internal.h
@@ -20,6 +20,7 @@ FIELD(MMUIDXINFO, PAN, 7, 1)
FIELD(MMUIDXINFO, USER, 8, 1)
FIELD(MMUIDXINFO, STAGE1, 9, 1)
FIELD(MMUIDXINFO, STAGE2, 10, 1)
+FIELD(MMUIDXINFO, GCS, 11, 1)
extern const uint32_t arm_mmuidx_table[ARM_MMU_IDX_M + 8];
@@ -92,4 +93,11 @@ static inline bool regime_is_stage2(ARMMMUIdx idx)
return FIELD_EX32(arm_mmuidx_table[idx], MMUIDXINFO, STAGE2);
}
+/* Return true if this mmu index implies AccessType_GCS. */
+static inline bool regime_is_gcs(ARMMMUIdx idx)
+{
+ tcg_debug_assert(arm_mmuidx_is_valid(idx));
+ return FIELD_EX32(arm_mmuidx_table[idx], MMUIDXINFO, GCS);
+}
+
#endif /* TARGET_ARM_MMUIDX_INTERNAL_H */
diff --git a/target/arm/mmuidx.h b/target/arm/mmuidx.h
index 5b9b4bc84f..8d8d27337e 100644
--- a/target/arm/mmuidx.h
+++ b/target/arm/mmuidx.h
@@ -58,24 +58,31 @@
* already heavyweight.
* 8. we cannot fold together Stage 2 Secure and Stage 2 NonSecure,
* because both are in use simultaneously for Secure EL2.
+ * 9. we need separate indexes for handling AccessType_GCS.
*
* This gives us the following list of cases:
*
* EL0 EL1&0 stage 1+2 (aka NS PL0 PL1&0 stage 1+2)
+ * EL0 EL1&0 stage 1+2 +GCS
* EL1 EL1&0 stage 1+2 (aka NS PL1 PL1&0 stage 1+2)
* EL1 EL1&0 stage 1+2 +PAN (aka NS PL1 P1&0 stage 1+2 +PAN)
+ * EL1 EL1&0 stage 1+2 +GCS
* EL0 EL2&0
+ * EL0 EL2&0 +GCS
* EL2 EL2&0
* EL2 EL2&0 +PAN
+ * EL2 EL2&0 +GCS
* EL2 (aka NS PL2)
+ * EL2 +GCS
* EL3 (aka AArch32 S PL1 PL1&0)
+ * EL3 +GCS
* AArch32 S PL0 PL1&0 (we call this EL30_0)
* AArch32 S PL1 PL1&0 +PAN (we call this EL30_3_PAN)
* Stage2 Secure
* Stage2 NonSecure
* plus one TLB per Physical address space: S, NS, Realm, Root
*
- * for a total of 16 different mmu_idx.
+ * for a total of 22 different mmu_idx.
*
* R profile CPUs have an MPU, but can use the same set of MMU indexes
* as A profile. They only need to distinguish EL0 and EL1 (and
@@ -114,9 +121,9 @@
* For M profile we arrange them to have a bit for priv, a bit for negpri
* and a bit for secure.
*/
-#define ARM_MMU_IDX_A 0x10 /* A profile */
-#define ARM_MMU_IDX_NOTLB 0x20 /* does not have a TLB */
-#define ARM_MMU_IDX_M 0x40 /* M profile */
+#define ARM_MMU_IDX_A 0x20 /* A profile */
+#define ARM_MMU_IDX_NOTLB 0x40 /* does not have a TLB */
+#define ARM_MMU_IDX_M 0x80 /* M profile */
/* Meanings of the bits for M profile mmu idx values */
#define ARM_MMU_IDX_M_PRIV 0x1
@@ -125,22 +132,32 @@
#define ARM_MMU_IDX_TYPE_MASK \
(ARM_MMU_IDX_A | ARM_MMU_IDX_M | ARM_MMU_IDX_NOTLB)
-#define ARM_MMU_IDX_COREIDX_MASK 0xf
+#define ARM_MMU_IDX_COREIDX_MASK 0x1f
typedef enum ARMMMUIdx {
/*
* A-profile.
*/
- ARMMMUIdx_E10_0 = 0 | ARM_MMU_IDX_A,
- ARMMMUIdx_E20_0 = 1 | ARM_MMU_IDX_A,
- ARMMMUIdx_E10_1 = 2 | ARM_MMU_IDX_A,
- ARMMMUIdx_E20_2 = 3 | ARM_MMU_IDX_A,
- ARMMMUIdx_E10_1_PAN = 4 | ARM_MMU_IDX_A,
- ARMMMUIdx_E20_2_PAN = 5 | ARM_MMU_IDX_A,
- ARMMMUIdx_E2 = 6 | ARM_MMU_IDX_A,
- ARMMMUIdx_E3 = 7 | ARM_MMU_IDX_A,
- ARMMMUIdx_E30_0 = 8 | ARM_MMU_IDX_A,
- ARMMMUIdx_E30_3_PAN = 9 | ARM_MMU_IDX_A,
+
+ ARMMMUIdx_E10_0 = 0 | ARM_MMU_IDX_A,
+ ARMMMUIdx_E10_0_GCS = 1 | ARM_MMU_IDX_A,
+ ARMMMUIdx_E10_1 = 2 | ARM_MMU_IDX_A,
+ ARMMMUIdx_E10_1_PAN = 3 | ARM_MMU_IDX_A,
+ ARMMMUIdx_E10_1_GCS = 4 | ARM_MMU_IDX_A,
+
+ ARMMMUIdx_E20_0 = 5 | ARM_MMU_IDX_A,
+ ARMMMUIdx_E20_0_GCS = 6 | ARM_MMU_IDX_A,
+ ARMMMUIdx_E20_2 = 7 | ARM_MMU_IDX_A,
+ ARMMMUIdx_E20_2_PAN = 8 | ARM_MMU_IDX_A,
+ ARMMMUIdx_E20_2_GCS = 9 | ARM_MMU_IDX_A,
+
+ ARMMMUIdx_E2 = 10 | ARM_MMU_IDX_A,
+ ARMMMUIdx_E2_GCS = 11 | ARM_MMU_IDX_A,
+
+ ARMMMUIdx_E3 = 12 | ARM_MMU_IDX_A,
+ ARMMMUIdx_E3_GCS = 13 | ARM_MMU_IDX_A,
+ ARMMMUIdx_E30_0 = 14 | ARM_MMU_IDX_A,
+ ARMMMUIdx_E30_3_PAN = 15 | ARM_MMU_IDX_A,
/*
* Used for second stage of an S12 page table walk, or for descriptor
@@ -148,14 +165,14 @@ typedef enum ARMMMUIdx {
* are in use simultaneously for SecureEL2: the security state for
* the S2 ptw is selected by the NS bit from the S1 ptw.
*/
- ARMMMUIdx_Stage2_S = 10 | ARM_MMU_IDX_A,
- ARMMMUIdx_Stage2 = 11 | ARM_MMU_IDX_A,
+ ARMMMUIdx_Stage2_S = 16 | ARM_MMU_IDX_A,
+ ARMMMUIdx_Stage2 = 17 | ARM_MMU_IDX_A,
/* TLBs with 1-1 mapping to the physical address spaces. */
- ARMMMUIdx_Phys_S = 12 | ARM_MMU_IDX_A,
- ARMMMUIdx_Phys_NS = 13 | ARM_MMU_IDX_A,
- ARMMMUIdx_Phys_Root = 14 | ARM_MMU_IDX_A,
- ARMMMUIdx_Phys_Realm = 15 | ARM_MMU_IDX_A,
+ ARMMMUIdx_Phys_S = 18 | ARM_MMU_IDX_A,
+ ARMMMUIdx_Phys_NS = 19 | ARM_MMU_IDX_A,
+ ARMMMUIdx_Phys_Root = 20 | ARM_MMU_IDX_A,
+ ARMMMUIdx_Phys_Realm = 21 | ARM_MMU_IDX_A,
/*
* These are not allocated TLBs and are used only for AT system
@@ -164,6 +181,8 @@ typedef enum ARMMMUIdx {
ARMMMUIdx_Stage1_E0 = 0 | ARM_MMU_IDX_NOTLB,
ARMMMUIdx_Stage1_E1 = 1 | ARM_MMU_IDX_NOTLB,
ARMMMUIdx_Stage1_E1_PAN = 2 | ARM_MMU_IDX_NOTLB,
+ ARMMMUIdx_Stage1_E0_GCS = 3 | ARM_MMU_IDX_NOTLB,
+ ARMMMUIdx_Stage1_E1_GCS = 4 | ARM_MMU_IDX_NOTLB,
/*
* M-profile.
@@ -187,13 +206,19 @@ typedef enum ARMMMUIdx {
typedef enum ARMMMUIdxBit {
TO_CORE_BIT(E10_0),
- TO_CORE_BIT(E20_0),
+ TO_CORE_BIT(E10_0_GCS),
TO_CORE_BIT(E10_1),
TO_CORE_BIT(E10_1_PAN),
- TO_CORE_BIT(E2),
+ TO_CORE_BIT(E10_1_GCS),
+ TO_CORE_BIT(E20_0),
+ TO_CORE_BIT(E20_0_GCS),
TO_CORE_BIT(E20_2),
TO_CORE_BIT(E20_2_PAN),
+ TO_CORE_BIT(E20_2_GCS),
+ TO_CORE_BIT(E2),
+ TO_CORE_BIT(E2_GCS),
TO_CORE_BIT(E3),
+ TO_CORE_BIT(E3_GCS),
TO_CORE_BIT(E30_0),
TO_CORE_BIT(E30_3_PAN),
TO_CORE_BIT(Stage2),
diff --git a/target/arm/helper.c b/target/arm/helper.c
index c0e808a9e8..65f38727d6 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -420,7 +420,9 @@ int alle1_tlbmask(CPUARMState *env)
*/
return (ARMMMUIdxBit_E10_1 |
ARMMMUIdxBit_E10_1_PAN |
+ ARMMMUIdxBit_E10_1_GCS |
ARMMMUIdxBit_E10_0 |
+ ARMMMUIdxBit_E10_0_GCS |
ARMMMUIdxBit_Stage2 |
ARMMMUIdxBit_Stage2_S);
}
@@ -808,12 +810,17 @@ static void scr_write(CPUARMState *env, const ARMCPRegInfo *ri, uint64_t value)
*/
if (changed & (SCR_NS | SCR_NSE)) {
tlb_flush_by_mmuidx(env_cpu(env), (ARMMMUIdxBit_E10_0 |
+ ARMMMUIdxBit_E10_0_GCS |
ARMMMUIdxBit_E20_0 |
+ ARMMMUIdxBit_E20_0_GCS |
ARMMMUIdxBit_E10_1 |
- ARMMMUIdxBit_E20_2 |
ARMMMUIdxBit_E10_1_PAN |
+ ARMMMUIdxBit_E10_1_GCS |
+ ARMMMUIdxBit_E20_2 |
ARMMMUIdxBit_E20_2_PAN |
- ARMMMUIdxBit_E2));
+ ARMMMUIdxBit_E20_2_GCS |
+ ARMMMUIdxBit_E2 |
+ ARMMMUIdxBit_E2_GCS));
}
}
@@ -2787,7 +2794,9 @@ static void vmsa_tcr_ttbr_el2_write(CPUARMState *env, const ARMCPRegInfo *ri,
(arm_hcr_el2_eff(env) & HCR_E2H)) {
uint16_t mask = ARMMMUIdxBit_E20_2 |
ARMMMUIdxBit_E20_2_PAN |
- ARMMMUIdxBit_E20_0;
+ ARMMMUIdxBit_E20_2_GCS |
+ ARMMMUIdxBit_E20_0 |
+ ARMMMUIdxBit_E20_0_GCS;
tlb_flush_by_mmuidx(env_cpu(env), mask);
}
raw_write(env, ri, value);
diff --git a/target/arm/mmuidx.c b/target/arm/mmuidx.c
index 61a682e655..42b003db9c 100644
--- a/target/arm/mmuidx.c
+++ b/target/arm/mmuidx.c
@@ -15,22 +15,29 @@
#define USER R_MMUIDXINFO_USER_MASK
#define S1 R_MMUIDXINFO_STAGE1_MASK
#define S2 R_MMUIDXINFO_STAGE2_MASK
+#define GCS R_MMUIDXINFO_GCS_MASK
const uint32_t arm_mmuidx_table[ARM_MMU_IDX_M + 8] = {
/*
* A-profile.
*/
[ARMMMUIdx_E10_0] = EL(0) | REL(1) | R2,
+ [ARMMMUIdx_E10_0_GCS] = EL(0) | REL(1) | R2 | GCS,
[ARMMMUIdx_E10_1] = EL(1) | REL(1) | R2,
[ARMMMUIdx_E10_1_PAN] = EL(1) | REL(1) | R2 | PAN,
+ [ARMMMUIdx_E10_1_GCS] = EL(1) | REL(1) | R2 | GCS,
[ARMMMUIdx_E20_0] = EL(0) | REL(2) | R2,
+ [ARMMMUIdx_E20_0_GCS] = EL(0) | REL(2) | R2 | GCS,
[ARMMMUIdx_E20_2] = EL(2) | REL(2) | R2,
[ARMMMUIdx_E20_2_PAN] = EL(2) | REL(2) | R2 | PAN,
+ [ARMMMUIdx_E20_2_GCS] = EL(2) | REL(2) | R2 | GCS,
[ARMMMUIdx_E2] = EL(2) | REL(2),
+ [ARMMMUIdx_E2_GCS] = EL(2) | REL(2) | GCS,
[ARMMMUIdx_E3] = EL(3) | REL(3),
+ [ARMMMUIdx_E3_GCS] = EL(3) | REL(3) | GCS,
[ARMMMUIdx_E30_0] = EL(0) | REL(3),
[ARMMMUIdx_E30_3_PAN] = EL(3) | REL(3) | PAN,
@@ -38,8 +45,10 @@ const uint32_t arm_mmuidx_table[ARM_MMU_IDX_M + 8] = {
[ARMMMUIdx_Stage2] = REL(2) | S2,
[ARMMMUIdx_Stage1_E0] = REL(1) | R2 | S1 | USER,
+ [ARMMMUIdx_Stage1_E0_GCS] = REL(1) | R2 | S1 | USER | GCS,
[ARMMMUIdx_Stage1_E1] = REL(1) | R2 | S1,
[ARMMMUIdx_Stage1_E1_PAN] = REL(1) | R2 | S1 | PAN,
+ [ARMMMUIdx_Stage1_E1_GCS] = REL(1) | R2 | S1 | GCS,
/*
* M-profile.
diff --git a/target/arm/ptw.c b/target/arm/ptw.c
index b72c1fc348..16ca2b6187 100644
--- a/target/arm/ptw.c
+++ b/target/arm/ptw.c
@@ -166,6 +166,10 @@ ARMMMUIdx stage_1_mmu_idx(ARMMMUIdx mmu_idx)
return ARMMMUIdx_Stage1_E1;
case ARMMMUIdx_E10_1_PAN:
return ARMMMUIdx_Stage1_E1_PAN;
+ case ARMMMUIdx_E10_0_GCS:
+ return ARMMMUIdx_Stage1_E0_GCS;
+ case ARMMMUIdx_E10_1_GCS:
+ return ARMMMUIdx_Stage1_E1_GCS;
default:
return mmu_idx;
}
@@ -273,8 +277,10 @@ static bool regime_translation_disabled(CPUARMState *env, ARMMMUIdx mmu_idx,
return (hcr_el2 & (HCR_DC | HCR_VM)) == 0;
case ARMMMUIdx_E10_0:
+ case ARMMMUIdx_E10_0_GCS:
case ARMMMUIdx_E10_1:
case ARMMMUIdx_E10_1_PAN:
+ case ARMMMUIdx_E10_1_GCS:
/* TGE means that EL0/1 act as if SCTLR_EL1.M is zero */
hcr_el2 = arm_hcr_el2_eff_secstate(env, space);
if (hcr_el2 & HCR_TGE) {
@@ -283,8 +289,10 @@ static bool regime_translation_disabled(CPUARMState *env, ARMMMUIdx mmu_idx,
break;
case ARMMMUIdx_Stage1_E0:
+ case ARMMMUIdx_Stage1_E0_GCS:
case ARMMMUIdx_Stage1_E1:
case ARMMMUIdx_Stage1_E1_PAN:
+ case ARMMMUIdx_Stage1_E1_GCS:
/* HCR.DC means SCTLR_EL1.M behaves as 0 */
hcr_el2 = arm_hcr_el2_eff_secstate(env, space);
if (hcr_el2 & HCR_DC) {
@@ -293,10 +301,14 @@ static bool regime_translation_disabled(CPUARMState *env, ARMMMUIdx mmu_idx,
break;
case ARMMMUIdx_E20_0:
+ case ARMMMUIdx_E20_0_GCS:
case ARMMMUIdx_E20_2:
case ARMMMUIdx_E20_2_PAN:
+ case ARMMMUIdx_E20_2_GCS:
case ARMMMUIdx_E2:
+ case ARMMMUIdx_E2_GCS:
case ARMMMUIdx_E3:
+ case ARMMMUIdx_E3_GCS:
case ARMMMUIdx_E30_0:
case ARMMMUIdx_E30_3_PAN:
break;
@@ -3766,15 +3778,22 @@ arm_mmu_idx_to_security_space(CPUARMState *env, ARMMMUIdx mmu_idx)
switch (mmu_idx) {
case ARMMMUIdx_E10_0:
+ case ARMMMUIdx_E10_0_GCS:
case ARMMMUIdx_E10_1:
case ARMMMUIdx_E10_1_PAN:
+ case ARMMMUIdx_E10_1_GCS:
case ARMMMUIdx_E20_0:
+ case ARMMMUIdx_E20_0_GCS:
case ARMMMUIdx_E20_2:
case ARMMMUIdx_E20_2_PAN:
+ case ARMMMUIdx_E20_2_GCS:
case ARMMMUIdx_Stage1_E0:
+ case ARMMMUIdx_Stage1_E0_GCS:
case ARMMMUIdx_Stage1_E1:
case ARMMMUIdx_Stage1_E1_PAN:
+ case ARMMMUIdx_Stage1_E1_GCS:
case ARMMMUIdx_E2:
+ case ARMMMUIdx_E2_GCS:
ss = arm_security_space_below_el3(env);
break;
case ARMMMUIdx_Stage2:
@@ -3803,6 +3822,7 @@ arm_mmu_idx_to_security_space(CPUARMState *env, ARMMMUIdx mmu_idx)
ss = ARMSS_Secure;
break;
case ARMMMUIdx_E3:
+ case ARMMMUIdx_E3_GCS:
case ARMMMUIdx_E30_0:
case ARMMMUIdx_E30_3_PAN:
if (arm_feature(env, ARM_FEATURE_AARCH64) &&
diff --git a/target/arm/tcg/tlb-insns.c b/target/arm/tcg/tlb-insns.c
index 95c26c6d46..1a0a332583 100644
--- a/target/arm/tcg/tlb-insns.c
+++ b/target/arm/tcg/tlb-insns.c
@@ -149,7 +149,8 @@ static void tlbimva_hyp_write(CPUARMState *env, const ARMCPRegInfo *ri,
CPUState *cs = env_cpu(env);
uint64_t pageaddr = value & ~MAKE_64BIT_MASK(0, 12);
- tlb_flush_page_by_mmuidx(cs, pageaddr, ARMMMUIdxBit_E2);
+ tlb_flush_page_by_mmuidx(cs, pageaddr,
+ ARMMMUIdxBit_E2 | ARMMMUIdxBit_E2_GCS);
}
static void tlbimva_hyp_is_write(CPUARMState *env, const ARMCPRegInfo *ri,
@@ -159,7 +160,8 @@ static void tlbimva_hyp_is_write(CPUARMState *env, const ARMCPRegInfo *ri,
uint64_t pageaddr = value & ~MAKE_64BIT_MASK(0, 12);
tlb_flush_page_by_mmuidx_all_cpus_synced(cs, pageaddr,
- ARMMMUIdxBit_E2);
+ ARMMMUIdxBit_E2 |
+ ARMMMUIdxBit_E2_GCS);
}
static void tlbiipas2_hyp_write(CPUARMState *env, const ARMCPRegInfo *ri,
@@ -202,7 +204,7 @@ static void tlbiall_hyp_write(CPUARMState *env, const ARMCPRegInfo *ri,
{
CPUState *cs = env_cpu(env);
- tlb_flush_by_mmuidx(cs, ARMMMUIdxBit_E2);
+ tlb_flush_by_mmuidx(cs, ARMMMUIdxBit_E2 | ARMMMUIdxBit_E2_GCS);
}
static void tlbiall_hyp_is_write(CPUARMState *env, const ARMCPRegInfo *ri,
@@ -210,7 +212,8 @@ static void tlbiall_hyp_is_write(CPUARMState *env, const ARMCPRegInfo *ri,
{
CPUState *cs = env_cpu(env);
- tlb_flush_by_mmuidx_all_cpus_synced(cs, ARMMMUIdxBit_E2);
+ tlb_flush_by_mmuidx_all_cpus_synced(cs, ARMMMUIdxBit_E2 |
+ ARMMMUIdxBit_E2_GCS);
}
/*
@@ -228,12 +231,16 @@ static int vae1_tlbmask(CPUARMState *env)
if ((hcr & (HCR_E2H | HCR_TGE)) == (HCR_E2H | HCR_TGE)) {
mask = ARMMMUIdxBit_E20_2 |
ARMMMUIdxBit_E20_2_PAN |
- ARMMMUIdxBit_E20_0;
+ ARMMMUIdxBit_E20_2_GCS |
+ ARMMMUIdxBit_E20_0 |
+ ARMMMUIdxBit_E20_0_GCS;
} else {
/* This is AArch64 only, so we don't need to touch the EL30_x TLBs */
mask = ARMMMUIdxBit_E10_1 |
ARMMMUIdxBit_E10_1_PAN |
- ARMMMUIdxBit_E10_0;
+ ARMMMUIdxBit_E10_1_GCS |
+ ARMMMUIdxBit_E10_0 |
+ ARMMMUIdxBit_E10_0_GCS;
}
return mask;
}
@@ -246,13 +253,20 @@ static int vae2_tlbmask(CPUARMState *env)
if (hcr & HCR_E2H) {
mask = ARMMMUIdxBit_E20_2 |
ARMMMUIdxBit_E20_2_PAN |
- ARMMMUIdxBit_E20_0;
+ ARMMMUIdxBit_E20_2_GCS |
+ ARMMMUIdxBit_E20_0 |
+ ARMMMUIdxBit_E20_0_GCS;
} else {
- mask = ARMMMUIdxBit_E2;
+ mask = ARMMMUIdxBit_E2 | ARMMMUIdxBit_E2_GCS;
}
return mask;
}
+static int vae3_tlbmask(void)
+{
+ return ARMMMUIdxBit_E3 | ARMMMUIdxBit_E3_GCS;
+}
+
/* Return 56 if TBI is enabled, 64 otherwise. */
static int tlbbits_for_regime(CPUARMState *env, ARMMMUIdx mmu_idx,
uint64_t addr)
@@ -325,9 +339,12 @@ static void tlbi_aa64_vmalle1_write(CPUARMState *env, const ARMCPRegInfo *ri,
static int e2_tlbmask(CPUARMState *env)
{
return (ARMMMUIdxBit_E20_0 |
+ ARMMMUIdxBit_E20_0_GCS |
ARMMMUIdxBit_E20_2 |
ARMMMUIdxBit_E20_2_PAN |
- ARMMMUIdxBit_E2);
+ ARMMMUIdxBit_E20_2_GCS |
+ ARMMMUIdxBit_E2 |
+ ARMMMUIdxBit_E2_GCS);
}
static void tlbi_aa64_alle1_write(CPUARMState *env, const ARMCPRegInfo *ri,
@@ -354,7 +371,7 @@ static void tlbi_aa64_alle3_write(CPUARMState *env, const ARMCPRegInfo *ri,
ARMCPU *cpu = env_archcpu(env);
CPUState *cs = CPU(cpu);
- tlb_flush_by_mmuidx(cs, ARMMMUIdxBit_E3);
+ tlb_flush_by_mmuidx(cs, vae3_tlbmask());
}
static void tlbi_aa64_alle1is_write(CPUARMState *env, const ARMCPRegInfo *ri,
@@ -380,7 +397,7 @@ static void tlbi_aa64_alle3is_write(CPUARMState *env, const ARMCPRegInfo *ri,
{
CPUState *cs = env_cpu(env);
- tlb_flush_by_mmuidx_all_cpus_synced(cs, ARMMMUIdxBit_E3);
+ tlb_flush_by_mmuidx_all_cpus_synced(cs, vae3_tlbmask());
}
static void tlbi_aa64_vae2_write(CPUARMState *env, const ARMCPRegInfo *ri,
@@ -411,7 +428,7 @@ static void tlbi_aa64_vae3_write(CPUARMState *env, const ARMCPRegInfo *ri,
CPUState *cs = CPU(cpu);
uint64_t pageaddr = sextract64(value << 12, 0, 56);
- tlb_flush_page_by_mmuidx(cs, pageaddr, ARMMMUIdxBit_E3);
+ tlb_flush_page_by_mmuidx(cs, pageaddr, vae3_tlbmask());
}
static void tlbi_aa64_vae1is_write(CPUARMState *env, const ARMCPRegInfo *ri,
@@ -465,7 +482,7 @@ static void tlbi_aa64_vae3is_write(CPUARMState *env, const ARMCPRegInfo *ri,
int bits = tlbbits_for_regime(env, ARMMMUIdx_E3, pageaddr);
tlb_flush_page_bits_by_mmuidx_all_cpus_synced(cs, pageaddr,
- ARMMMUIdxBit_E3, bits);
+ vae3_tlbmask(), bits);
}
static int ipas2e1_tlbmask(CPUARMState *env, int64_t value)
@@ -963,7 +980,7 @@ static void tlbi_aa64_rvae3_write(CPUARMState *env,
* flush-last-level-only.
*/
- do_rvae_write(env, value, ARMMMUIdxBit_E3, tlb_force_broadcast(env));
+ do_rvae_write(env, value, vae3_tlbmask(), tlb_force_broadcast(env));
}
static void tlbi_aa64_rvae3is_write(CPUARMState *env,
@@ -977,7 +994,7 @@ static void tlbi_aa64_rvae3is_write(CPUARMState *env,
* flush-last-level-only or inner/outer specific flushes.
*/
- do_rvae_write(env, value, ARMMMUIdxBit_E3, true);
+ do_rvae_write(env, value, vae3_tlbmask(), true);
}
static void tlbi_aa64_ripas2e1_write(CPUARMState *env, const ARMCPRegInfo *ri,
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 31/76] target/arm: Introduce regime_to_gcs
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (29 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 30/76] target/arm: Introduce mmu indexes for GCS Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 32/76] target/arm: Support page protections for GCS mmu indexes Richard Henderson
` (45 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Add a lookup from any a64 mmu index to the gcs mmu index
within the same translation regime.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/mmuidx-internal.h | 10 ++++++++++
target/arm/mmuidx.c | 24 +++++++++++++-----------
2 files changed, 23 insertions(+), 11 deletions(-)
diff --git a/target/arm/mmuidx-internal.h b/target/arm/mmuidx-internal.h
index f494ec348d..962b053852 100644
--- a/target/arm/mmuidx-internal.h
+++ b/target/arm/mmuidx-internal.h
@@ -21,6 +21,7 @@ FIELD(MMUIDXINFO, USER, 8, 1)
FIELD(MMUIDXINFO, STAGE1, 9, 1)
FIELD(MMUIDXINFO, STAGE2, 10, 1)
FIELD(MMUIDXINFO, GCS, 11, 1)
+FIELD(MMUIDXINFO, TG, 12, 5)
extern const uint32_t arm_mmuidx_table[ARM_MMU_IDX_M + 8];
@@ -100,4 +101,13 @@ static inline bool regime_is_gcs(ARMMMUIdx idx)
return FIELD_EX32(arm_mmuidx_table[idx], MMUIDXINFO, GCS);
}
+/* Return the GCS MMUIdx for a given regime. */
+static inline ARMMMUIdx regime_to_gcs(ARMMMUIdx idx)
+{
+ tcg_debug_assert(arm_mmuidx_is_valid(idx));
+ uint32_t core = FIELD_EX32(arm_mmuidx_table[idx], MMUIDXINFO, TG);
+ tcg_debug_assert(core != 0); /* core 0 is E10_0, not a GCS index */
+ return core | ARM_MMU_IDX_A;
+}
+
#endif /* TARGET_ARM_MMUIDX_INTERNAL_H */
diff --git a/target/arm/mmuidx.c b/target/arm/mmuidx.c
index 42b003db9c..a4663c8d87 100644
--- a/target/arm/mmuidx.c
+++ b/target/arm/mmuidx.c
@@ -16,27 +16,29 @@
#define S1 R_MMUIDXINFO_STAGE1_MASK
#define S2 R_MMUIDXINFO_STAGE2_MASK
#define GCS R_MMUIDXINFO_GCS_MASK
+#define TG(X) \
+ ((ARMMMUIdx_##X##_GCS & ARM_MMU_IDX_COREIDX_MASK) << R_MMUIDXINFO_TG_SHIFT)
const uint32_t arm_mmuidx_table[ARM_MMU_IDX_M + 8] = {
/*
* A-profile.
*/
- [ARMMMUIdx_E10_0] = EL(0) | REL(1) | R2,
+ [ARMMMUIdx_E10_0] = EL(0) | REL(1) | R2 | TG(E10_0),
[ARMMMUIdx_E10_0_GCS] = EL(0) | REL(1) | R2 | GCS,
- [ARMMMUIdx_E10_1] = EL(1) | REL(1) | R2,
- [ARMMMUIdx_E10_1_PAN] = EL(1) | REL(1) | R2 | PAN,
+ [ARMMMUIdx_E10_1] = EL(1) | REL(1) | R2 | TG(E10_1),
+ [ARMMMUIdx_E10_1_PAN] = EL(1) | REL(1) | R2 | TG(E10_1) | PAN,
[ARMMMUIdx_E10_1_GCS] = EL(1) | REL(1) | R2 | GCS,
- [ARMMMUIdx_E20_0] = EL(0) | REL(2) | R2,
+ [ARMMMUIdx_E20_0] = EL(0) | REL(2) | R2 | TG(E20_0),
[ARMMMUIdx_E20_0_GCS] = EL(0) | REL(2) | R2 | GCS,
- [ARMMMUIdx_E20_2] = EL(2) | REL(2) | R2,
- [ARMMMUIdx_E20_2_PAN] = EL(2) | REL(2) | R2 | PAN,
+ [ARMMMUIdx_E20_2] = EL(2) | REL(2) | R2 | TG(E20_2),
+ [ARMMMUIdx_E20_2_PAN] = EL(2) | REL(2) | R2 | TG(E20_2) | PAN,
[ARMMMUIdx_E20_2_GCS] = EL(2) | REL(2) | R2 | GCS,
- [ARMMMUIdx_E2] = EL(2) | REL(2),
+ [ARMMMUIdx_E2] = EL(2) | REL(2) | TG(E2),
[ARMMMUIdx_E2_GCS] = EL(2) | REL(2) | GCS,
- [ARMMMUIdx_E3] = EL(3) | REL(3),
+ [ARMMMUIdx_E3] = EL(3) | REL(3) | TG(E3),
[ARMMMUIdx_E3_GCS] = EL(3) | REL(3) | GCS,
[ARMMMUIdx_E30_0] = EL(0) | REL(3),
[ARMMMUIdx_E30_3_PAN] = EL(3) | REL(3) | PAN,
@@ -44,10 +46,10 @@ const uint32_t arm_mmuidx_table[ARM_MMU_IDX_M + 8] = {
[ARMMMUIdx_Stage2_S] = REL(2) | S2,
[ARMMMUIdx_Stage2] = REL(2) | S2,
- [ARMMMUIdx_Stage1_E0] = REL(1) | R2 | S1 | USER,
+ [ARMMMUIdx_Stage1_E0] = REL(1) | R2 | S1 | USER | TG(Stage1_E0),
[ARMMMUIdx_Stage1_E0_GCS] = REL(1) | R2 | S1 | USER | GCS,
- [ARMMMUIdx_Stage1_E1] = REL(1) | R2 | S1,
- [ARMMMUIdx_Stage1_E1_PAN] = REL(1) | R2 | S1 | PAN,
+ [ARMMMUIdx_Stage1_E1] = REL(1) | R2 | S1 | TG(Stage1_E1),
+ [ARMMMUIdx_Stage1_E1_PAN] = REL(1) | R2 | S1 | TG(Stage1_E1) | PAN,
[ARMMMUIdx_Stage1_E1_GCS] = REL(1) | R2 | S1 | GCS,
/*
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 32/76] target/arm: Support page protections for GCS mmu indexes
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (30 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 31/76] target/arm: Introduce regime_to_gcs Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 33/76] target/arm: Implement gcs bit for data abort Richard Henderson
` (44 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Take read and write from the s1perms.gcs bit computed
by the Arm pseudocode.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/ptw.c | 17 ++++++++++++++---
1 file changed, 14 insertions(+), 3 deletions(-)
diff --git a/target/arm/ptw.c b/target/arm/ptw.c
index 16ca2b6187..bfa0aeb4c0 100644
--- a/target/arm/ptw.c
+++ b/target/arm/ptw.c
@@ -1583,12 +1583,16 @@ static int get_S1prot_indirect(CPUARMState *env, S1Translate *ptw,
}
}
- if (perm & PAGE_WXN) {
+ if (regime_is_gcs(mmu_idx)) {
+ /*
+ * Note that the one s1perms.gcs bit controls both read and write
+ * access via AccessType_GCS. See AArch64.S1CheckPermissions.
+ */
+ perm = (perm & PAGE_GCS ? PAGE_READ | PAGE_WRITE : 0);
+ } else if (perm & PAGE_WXN) {
perm &= ~PAGE_EXEC;
}
- /* TODO: FEAT_GCS */
-
return perm & PAGE_RWX;
}
@@ -2241,6 +2245,13 @@ static bool get_phys_addr_lpae(CPUARMState *env, S1Translate *ptw,
*/
prot = get_S1prot_indirect(env, ptw, mmu_idx, pi, po,
result->f.attrs.space, out_space);
+ } else if (regime_is_gcs(mmu_idx)) {
+ /*
+ * While one must use indirect permissions to successfully
+ * use GCS instructions, AArch64.S1DirectBasePermissions
+ * faithfully supplies s1perms.gcs = 0, Just In Case.
+ */
+ prot = 0;
} else {
int xn = extract64(attrs, 54, 1);
int pxn = extract64(attrs, 53, 1);
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 33/76] target/arm: Implement gcs bit for data abort
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (31 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 32/76] target/arm: Support page protections for GCS mmu indexes Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 34/76] target/arm: Add GCS cpregs Richard Henderson
` (43 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/tcg/tlb_helper.c | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/target/arm/tcg/tlb_helper.c b/target/arm/tcg/tlb_helper.c
index ae2acd6727..f1983a5732 100644
--- a/target/arm/tcg/tlb_helper.c
+++ b/target/arm/tcg/tlb_helper.c
@@ -28,7 +28,7 @@ static inline uint64_t merge_syn_data_abort(uint32_t template_syn,
ARMMMUFaultInfo *fi,
unsigned int target_el,
bool same_el, bool is_write,
- int fsc)
+ int fsc, bool gcs)
{
uint64_t syn;
@@ -78,6 +78,7 @@ static inline uint64_t merge_syn_data_abort(uint32_t template_syn,
/* Form ISS2 at the top of the syndrome. */
syn |= (uint64_t)fi->dirtybit << 37;
+ syn |= (uint64_t)gcs << 40;
return syn;
}
@@ -252,9 +253,10 @@ void arm_deliver_fault(ARMCPU *cpu, vaddr addr,
syn = syn_insn_abort(same_el, fi->ea, fi->s1ptw, fsc);
exc = EXCP_PREFETCH_ABORT;
} else {
+ bool gcs = regime_is_gcs(core_to_arm_mmu_idx(env, mmu_idx));
syn = merge_syn_data_abort(env->exception.syndrome, fi, target_el,
same_el, access_type == MMU_DATA_STORE,
- fsc);
+ fsc, gcs);
if (access_type == MMU_DATA_STORE
&& arm_feature(env, ARM_FEATURE_V6)) {
fsr |= (1 << 11);
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 34/76] target/arm: Add GCS cpregs
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (32 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 33/76] target/arm: Implement gcs bit for data abort Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 35/76] target/arm: Add GCS enable and trap levels to DisasContext Richard Henderson
` (42 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Add isar_feature_aa64_gcs.
Enable SCR_GCSEN in scr_write.
Enable HCRX_GCSEN in hcrx_write.
Default HCRX_GCSEN on if EL2 disabled.
Add the GCSCR* and GCSPR* registers.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/cpregs.h | 2 +
target/arm/cpu-features.h | 5 +++
target/arm/cpu.h | 12 +++++
target/arm/internals.h | 3 ++
target/arm/cpregs-gcs.c | 95 +++++++++++++++++++++++++++++++++++++++
target/arm/cpu.c | 3 ++
target/arm/helper.c | 10 +++++
target/arm/meson.build | 2 +
8 files changed, 132 insertions(+)
create mode 100644 target/arm/cpregs-gcs.c
diff --git a/target/arm/cpregs.h b/target/arm/cpregs.h
index f48c4df30f..bd2121a336 100644
--- a/target/arm/cpregs.h
+++ b/target/arm/cpregs.h
@@ -779,6 +779,8 @@ typedef enum FGTBit {
DO_BIT(HFGRTR, VBAR_EL1),
DO_BIT(HFGRTR, ICC_IGRPENN_EL1),
DO_BIT(HFGRTR, ERRIDR_EL1),
+ DO_REV_BIT(HFGRTR, NGCS_EL0),
+ DO_REV_BIT(HFGRTR, NGCS_EL1),
DO_REV_BIT(HFGRTR, NSMPRI_EL1),
DO_REV_BIT(HFGRTR, NTPIDR2_EL0),
DO_REV_BIT(HFGRTR, NPIRE0_EL1),
diff --git a/target/arm/cpu-features.h b/target/arm/cpu-features.h
index 30226814bb..4a35cf6b69 100644
--- a/target/arm/cpu-features.h
+++ b/target/arm/cpu-features.h
@@ -729,6 +729,11 @@ static inline bool isar_feature_aa64_nmi(const ARMISARegisters *id)
return FIELD_EX64_IDREG(id, ID_AA64PFR1, NMI) != 0;
}
+static inline bool isar_feature_aa64_gcs(const ARMISARegisters *id)
+{
+ return FIELD_EX64_IDREG(id, ID_AA64PFR1, GCS) != 0;
+}
+
static inline bool isar_feature_aa64_tgran4_lpa2(const ARMISARegisters *id)
{
return FIELD_SEX64_IDREG(id, ID_AA64MMFR0, TGRAN4) >= 1;
diff --git a/target/arm/cpu.h b/target/arm/cpu.h
index d5a45e9b43..14a08d2f99 100644
--- a/target/arm/cpu.h
+++ b/target/arm/cpu.h
@@ -580,6 +580,9 @@ typedef struct CPUArchState {
/* NV2 register */
uint64_t vncr_el2;
+
+ uint64_t gcscr_el[4]; /* GCSCRE0_EL1, GCSCR_EL[123] */
+ uint64_t gcspr_el[4]; /* GCSPR_EL[0123] */
} cp15;
struct {
@@ -1717,6 +1720,7 @@ static inline void xpsr_write(CPUARMState *env, uint32_t val, uint32_t mask)
#define SCR_ENAS0 (1ULL << 36)
#define SCR_ADEN (1ULL << 37)
#define SCR_HXEN (1ULL << 38)
+#define SCR_GCSEN (1ULL << 39)
#define SCR_TRNDR (1ULL << 40)
#define SCR_ENTP2 (1ULL << 41)
#define SCR_TCR2EN (1ULL << 43)
@@ -1725,6 +1729,14 @@ static inline void xpsr_write(CPUARMState *env, uint32_t val, uint32_t mask)
#define SCR_GPF (1ULL << 48)
#define SCR_NSE (1ULL << 62)
+/* GCSCR_ELx fields */
+#define GCSCR_PCRSEL (1ULL << 0)
+#define GCSCR_RVCHKEN (1ULL << 5)
+#define GCSCR_EXLOCKEN (1ULL << 6)
+#define GCSCR_PUSHMEN (1ULL << 8)
+#define GCSCR_STREN (1ULL << 9)
+#define GCSCRE0_NTR (1ULL << 10)
+
/* Return the current FPSCR value. */
uint32_t vfp_get_fpscr(CPUARMState *env);
void vfp_set_fpscr(CPUARMState *env, uint32_t val);
diff --git a/target/arm/internals.h b/target/arm/internals.h
index dc6bf3548e..f72e95a5ba 100644
--- a/target/arm/internals.h
+++ b/target/arm/internals.h
@@ -251,6 +251,7 @@ FIELD(VSTCR, SA, 30, 1)
#define HCRX_MSCEN (1ULL << 11)
#define HCRX_TCR2EN (1ULL << 14)
#define HCRX_SCTLR2EN (1ULL << 15)
+#define HCRX_GCSEN (1ULL << 22)
#define HPFAR_NS (1ULL << 63)
@@ -1778,6 +1779,8 @@ void define_tlb_insn_regs(ARMCPU *cpu);
void define_at_insn_regs(ARMCPU *cpu);
/* Add the cpreg definitions for PM cpregs */
void define_pm_cpregs(ARMCPU *cpu);
+/* Add the cpreg definitions for GCS cpregs */
+void define_gcs_cpregs(ARMCPU *cpu);
/* Effective value of MDCR_EL2 */
static inline uint64_t arm_mdcr_el2_eff(CPUARMState *env)
diff --git a/target/arm/cpregs-gcs.c b/target/arm/cpregs-gcs.c
new file mode 100644
index 0000000000..1ff041811d
--- /dev/null
+++ b/target/arm/cpregs-gcs.c
@@ -0,0 +1,95 @@
+/*
+ * QEMU ARM CP Register GCS regiters and instructions
+ * SPDX-License-Identifier: GPL-2.0-or-later
+ */
+
+#include "qemu/osdep.h"
+#include "qemu/timer.h"
+#include "exec/icount.h"
+#include "hw/irq.h"
+#include "cpu.h"
+#include "cpu-features.h"
+#include "cpregs.h"
+#include "internals.h"
+
+
+static CPAccessResult access_gcs(CPUARMState *env, const ARMCPRegInfo *ri,
+ bool isread)
+{
+ if (arm_current_el(env) < 3
+ && arm_feature(env, ARM_FEATURE_EL3)
+ && !(env->cp15.scr_el3 & SCR_GCSEN)) {
+ return CP_ACCESS_TRAP_EL3;
+ }
+ return CP_ACCESS_OK;
+}
+
+static CPAccessResult access_gcs_el0(CPUARMState *env, const ARMCPRegInfo *ri,
+ bool isread)
+{
+ if (arm_current_el(env) == 0 && !(env->cp15.gcscr_el[0] & GCSCRE0_NTR)) {
+ return CP_ACCESS_TRAP_EL1;
+ }
+ return access_gcs(env, ri, isread);
+}
+
+static void gcspr_write(CPUARMState *env, const ARMCPRegInfo *ri,
+ uint64_t value)
+{
+ /*
+ * Bits [2:0] are RES0, so we might as well clear them now,
+ * rather than upon each usage a-la GetCurrentGCSPointer.
+ */
+ raw_write(env, ri, value & ~7);
+}
+
+static const ARMCPRegInfo gcs_reginfo[] = {
+ { .name = "GCSCRE0_EL1", .state = ARM_CP_STATE_AA64,
+ .opc0 = 3, .opc1 = 0, .crn = 2, .crm = 5, .opc2 = 2,
+ .access = PL1_RW, .accessfn = access_gcs, .fgt = FGT_NGCS_EL0,
+ .fieldoffset = offsetof(CPUARMState, cp15.gcscr_el[0]) },
+ { .name = "GCSCR_EL1", .state = ARM_CP_STATE_AA64,
+ .opc0 = 3, .opc1 = 0, .crn = 2, .crm = 5, .opc2 = 0,
+ .access = PL1_RW, .accessfn = access_gcs, .fgt = FGT_NGCS_EL1,
+ .nv2_redirect_offset = 0x8d0 | NV2_REDIR_NV1,
+ .vhe_redir_to_el2 = ENCODE_AA64_CP_REG(3, 4, 2, 5, 0),
+ .vhe_redir_to_el01 = ENCODE_AA64_CP_REG(3, 5, 2, 5, 0),
+ .fieldoffset = offsetof(CPUARMState, cp15.gcscr_el[1]) },
+ { .name = "GCSCR_EL2", .state = ARM_CP_STATE_AA64,
+ .opc0 = 3, .opc1 = 4, .crn = 2, .crm = 5, .opc2 = 0,
+ .access = PL2_RW, .accessfn = access_gcs,
+ .fieldoffset = offsetof(CPUARMState, cp15.gcscr_el[2]) },
+ { .name = "GCSCR_EL3", .state = ARM_CP_STATE_AA64,
+ .opc0 = 3, .opc1 = 6, .crn = 2, .crm = 5, .opc2 = 0,
+ .access = PL3_RW,
+ .fieldoffset = offsetof(CPUARMState, cp15.gcscr_el[3]) },
+
+ { .name = "GCSPR_EL0", .state = ARM_CP_STATE_AA64,
+ .opc0 = 3, .opc1 = 3, .crn = 2, .crm = 5, .opc2 = 1,
+ .access = PL0_R | PL1_W, .accessfn = access_gcs_el0,
+ .fgt = FGT_NGCS_EL0, .writefn = gcspr_write,
+ .fieldoffset = offsetof(CPUARMState, cp15.gcspr_el[0]) },
+ { .name = "GCSPR_EL1", .state = ARM_CP_STATE_AA64,
+ .opc0 = 3, .opc1 = 0, .crn = 2, .crm = 5, .opc2 = 1,
+ .access = PL1_RW, .accessfn = access_gcs,
+ .fgt = FGT_NGCS_EL1, .writefn = gcspr_write,
+ .nv2_redirect_offset = 0x8c0 | NV2_REDIR_NV1,
+ .vhe_redir_to_el2 = ENCODE_AA64_CP_REG(3, 4, 2, 5, 1),
+ .vhe_redir_to_el01 = ENCODE_AA64_CP_REG(3, 5, 2, 5, 1),
+ .fieldoffset = offsetof(CPUARMState, cp15.gcspr_el[1]) },
+ { .name = "GCSPR_EL2", .state = ARM_CP_STATE_AA64,
+ .opc0 = 3, .opc1 = 4, .crn = 2, .crm = 5, .opc2 = 1,
+ .access = PL2_RW, .accessfn = access_gcs, .writefn = gcspr_write,
+ .fieldoffset = offsetof(CPUARMState, cp15.gcspr_el[2]) },
+ { .name = "GCSPR_EL3", .state = ARM_CP_STATE_AA64,
+ .opc0 = 3, .opc1 = 6, .crn = 2, .crm = 5, .opc2 = 1,
+ .access = PL3_RW, .writefn = gcspr_write,
+ .fieldoffset = offsetof(CPUARMState, cp15.gcspr_el[2]) },
+};
+
+void define_gcs_cpregs(ARMCPU *cpu)
+{
+ if (cpu_isar_feature(aa64_gcs, cpu)) {
+ define_arm_cp_regs(cpu, gcs_reginfo);
+ }
+}
diff --git a/target/arm/cpu.c b/target/arm/cpu.c
index d4bc2e6a92..aab7250399 100644
--- a/target/arm/cpu.c
+++ b/target/arm/cpu.c
@@ -629,6 +629,9 @@ void arm_emulate_firmware_reset(CPUState *cpustate, int target_el)
if (cpu_isar_feature(aa64_fgt, cpu)) {
env->cp15.scr_el3 |= SCR_FGTEN;
}
+ if (cpu_isar_feature(aa64_gcs, cpu)) {
+ env->cp15.scr_el3 |= SCR_GCSEN;
+ }
if (cpu_isar_feature(aa64_tcr2, cpu)) {
env->cp15.scr_el3 |= SCR_TCR2EN;
}
diff --git a/target/arm/helper.c b/target/arm/helper.c
index 65f38727d6..3b9e7a822e 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -766,6 +766,9 @@ static void scr_write(CPUARMState *env, const ARMCPRegInfo *ri, uint64_t value)
if (cpu_isar_feature(aa64_ecv, cpu)) {
valid_mask |= SCR_ECVEN;
}
+ if (cpu_isar_feature(aa64_gcs, cpu)) {
+ valid_mask |= SCR_GCSEN;
+ }
if (cpu_isar_feature(aa64_tcr2, cpu)) {
valid_mask |= SCR_TCR2EN;
}
@@ -3948,6 +3951,9 @@ static void hcrx_write(CPUARMState *env, const ARMCPRegInfo *ri,
if (cpu_isar_feature(aa64_sctlr2, cpu)) {
valid_mask |= HCRX_SCTLR2EN;
}
+ if (cpu_isar_feature(aa64_gcs, cpu)) {
+ valid_mask |= HCRX_GCSEN;
+ }
/* Clear RES0 bits. */
env->cp15.hcrx_el2 = value & valid_mask;
@@ -4018,6 +4024,9 @@ uint64_t arm_hcrx_el2_eff(CPUARMState *env)
if (cpu_isar_feature(aa64_sctlr2, cpu)) {
hcrx |= HCRX_SCTLR2EN;
}
+ if (cpu_isar_feature(aa64_gcs, cpu)) {
+ hcrx |= HCRX_GCSEN;
+ }
return hcrx;
}
if (arm_feature(env, ARM_FEATURE_EL3) && !(env->cp15.scr_el3 & SCR_HXEN)) {
@@ -7247,6 +7256,7 @@ void register_cp_regs_for_features(ARMCPU *cpu)
}
define_pm_cpregs(cpu);
+ define_gcs_cpregs(cpu);
}
/*
diff --git a/target/arm/meson.build b/target/arm/meson.build
index 91630a1f72..8c82304fde 100644
--- a/target/arm/meson.build
+++ b/target/arm/meson.build
@@ -27,6 +27,7 @@ arm_user_ss.add(when: 'TARGET_AARCH64', if_false: files(
'cpu32-stubs.c',
))
arm_user_ss.add(files(
+ 'cpregs-gcs.c',
'cpregs-pmu.c',
'debug_helper.c',
'helper.c',
@@ -42,6 +43,7 @@ arm_common_system_ss.add(files(
'arch_dump.c',
'arm-powerctl.c',
'cortex-regs.c',
+ 'cpregs-gcs.c',
'cpregs-pmu.c',
'debug_helper.c',
'helper.c',
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 35/76] target/arm: Add GCS enable and trap levels to DisasContext
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (33 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 34/76] target/arm: Add GCS cpregs Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 36/76] target/arm: Implement FEAT_CHK Richard Henderson
` (41 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Pipe GCSEnabled, GCSReturnValueCheckEnabled, and CheckGCSSTREnabled
through hflags to the translator.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/cpu.h | 3 +++
target/arm/tcg/translate.h | 6 ++++++
target/arm/tcg/hflags.c | 38 ++++++++++++++++++++++++++++++++++
target/arm/tcg/translate-a64.c | 3 +++
4 files changed, 50 insertions(+)
diff --git a/target/arm/cpu.h b/target/arm/cpu.h
index 14a08d2f99..6a214557ac 100644
--- a/target/arm/cpu.h
+++ b/target/arm/cpu.h
@@ -2879,6 +2879,9 @@ FIELD(TBFLAG_A64, NV2_MEM_BE, 36, 1)
FIELD(TBFLAG_A64, AH, 37, 1) /* FPCR.AH */
FIELD(TBFLAG_A64, NEP, 38, 1) /* FPCR.NEP */
FIELD(TBFLAG_A64, ZT0EXC_EL, 39, 2)
+FIELD(TBFLAG_A64, GCS_EN, 41, 1)
+FIELD(TBFLAG_A64, GCS_RVCEN, 42, 1)
+FIELD(TBFLAG_A64, GCSSTR_EL, 43, 2)
/*
* Helpers for using the above. Note that only the A64 accessors use
diff --git a/target/arm/tcg/translate.h b/target/arm/tcg/translate.h
index f1a6e5e2b6..761edded52 100644
--- a/target/arm/tcg/translate.h
+++ b/target/arm/tcg/translate.h
@@ -166,6 +166,12 @@ typedef struct DisasContext {
bool fpcr_ah;
/* True if FPCR.NEP is 1 (FEAT_AFP scalar upper-element result handling) */
bool fpcr_nep;
+ /* True if GCSEnabled. */
+ bool gcs_en;
+ /* True if GCSReturnValueCheckEnabled. */
+ bool gcs_rvcen;
+ /* GCSSTR exception EL or 0 if enabled */
+ uint8_t gcsstr_el;
/*
* >= 0, a copy of PSTATE.BTYPE, which will be 0 without v8.5-BTI.
* < 0, set by the current instruction.
diff --git a/target/arm/tcg/hflags.c b/target/arm/tcg/hflags.c
index 17f83f13a4..5c9b9bec3b 100644
--- a/target/arm/tcg/hflags.c
+++ b/target/arm/tcg/hflags.c
@@ -451,6 +451,44 @@ static CPUARMTBFlags rebuild_hflags_a64(CPUARMState *env, int el, int fp_el,
DP_TBFLAG_A64(flags, TCMA, aa64_va_parameter_tcma(tcr, mmu_idx));
}
+ if (cpu_isar_feature(aa64_gcs, env_archcpu(env))) {
+ /* C.f. GCSEnabled */
+ if (env->cp15.gcscr_el[el] & GCSCR_PCRSEL) {
+ switch (el) {
+ default:
+ if (!el_is_in_host(env, el)
+ && !(arm_hcrx_el2_eff(env) & HCRX_GCSEN)) {
+ break;
+ }
+ /* fall through */
+ case 2:
+ if (arm_feature(env, ARM_FEATURE_EL3)
+ && !(env->cp15.scr_el3 & SCR_GCSEN)) {
+ break;
+ }
+ /* fall through */
+ case 3:
+ DP_TBFLAG_A64(flags, GCS_EN, 1);
+ break;
+ }
+ }
+
+ /* C.f. GCSReturnValueCheckEnabled */
+ if (env->cp15.gcscr_el[el] & GCSCR_RVCHKEN) {
+ DP_TBFLAG_A64(flags, GCS_RVCEN, 1);
+ }
+
+ /* C.f. CheckGCSSTREnabled */
+ if (!(env->cp15.gcscr_el[el] & GCSCR_STREN)) {
+ DP_TBFLAG_A64(flags, GCSSTR_EL, el ? el : 1);
+ } else if (el == 1
+ && EX_TBFLAG_ANY(flags, FGT_ACTIVE)
+ && !FIELD_EX64(env->cp15.fgt_exec[FGTREG_HFGITR],
+ HFGITR_EL2, NGCSSTR_EL1)) {
+ DP_TBFLAG_A64(flags, GCSSTR_EL, 2);
+ }
+ }
+
if (env->vfp.fpcr & FPCR_AH) {
DP_TBFLAG_A64(flags, AH, 1);
}
diff --git a/target/arm/tcg/translate-a64.c b/target/arm/tcg/translate-a64.c
index 6728e362b6..d5c41fbb3a 100644
--- a/target/arm/tcg/translate-a64.c
+++ b/target/arm/tcg/translate-a64.c
@@ -10344,6 +10344,9 @@ static void aarch64_tr_init_disas_context(DisasContextBase *dcbase,
dc->nv2_mem_be = EX_TBFLAG_A64(tb_flags, NV2_MEM_BE);
dc->fpcr_ah = EX_TBFLAG_A64(tb_flags, AH);
dc->fpcr_nep = EX_TBFLAG_A64(tb_flags, NEP);
+ dc->gcs_en = EX_TBFLAG_A64(tb_flags, GCS_EN);
+ dc->gcs_rvcen = EX_TBFLAG_A64(tb_flags, GCS_RVCEN);
+ dc->gcsstr_el = EX_TBFLAG_A64(tb_flags, GCSSTR_EL);
dc->vec_len = 0;
dc->vec_stride = 0;
dc->cp_regs = arm_cpu->cp_regs;
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 36/76] target/arm: Implement FEAT_CHK
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (34 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 35/76] target/arm: Add GCS enable and trap levels to DisasContext Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 37/76] target/arm: Make helper_exception_return system-only Richard Henderson
` (40 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Gustavo Romero, Pierrick Bouvier
This feature contains only the CHKFEAT instruction. It has
no ID enable, being back-allocated into the hint nop space.
Reviewed-by: Gustavo Romero <gustavo.romero@linaro.org>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/tcg/translate-a64.c | 14 ++++++++++++++
docs/system/arm/emulation.rst | 1 +
target/arm/tcg/a64.decode | 1 +
3 files changed, 16 insertions(+)
diff --git a/target/arm/tcg/translate-a64.c b/target/arm/tcg/translate-a64.c
index d5c41fbb3a..9dfcf4f993 100644
--- a/target/arm/tcg/translate-a64.c
+++ b/target/arm/tcg/translate-a64.c
@@ -2124,6 +2124,20 @@ static bool trans_AUTIBSP(DisasContext *s, arg_AUTIBSP *a)
return true;
}
+static bool trans_CHKFEAT(DisasContext *s, arg_CHKFEAT *a)
+{
+ uint64_t feat_en = 0;
+
+ if (s->gcs_en) {
+ feat_en |= 1 << 0;
+ }
+ if (feat_en) {
+ TCGv_i64 x16 = cpu_reg(s, 16);
+ tcg_gen_andi_i64(x16, x16, ~feat_en);
+ }
+ return true;
+}
+
static bool trans_CLREX(DisasContext *s, arg_CLREX *a)
{
tcg_gen_movi_i64(cpu_exclusive_addr, -1);
diff --git a/docs/system/arm/emulation.rst b/docs/system/arm/emulation.rst
index 94a6192fa9..4b07af9eac 100644
--- a/docs/system/arm/emulation.rst
+++ b/docs/system/arm/emulation.rst
@@ -28,6 +28,7 @@ the following architecture extensions:
- FEAT_BF16 (AArch64 BFloat16 instructions)
- FEAT_BTI (Branch Target Identification)
- FEAT_CCIDX (Extended cache index)
+- FEAT_CHK (Check Feature Status)
- FEAT_CMOW (Control for cache maintenance permission)
- FEAT_CRC32 (CRC32 instructions)
- FEAT_Crypto (Cryptographic Extension)
diff --git a/target/arm/tcg/a64.decode b/target/arm/tcg/a64.decode
index 55ff6c504f..8283a9c83d 100644
--- a/target/arm/tcg/a64.decode
+++ b/target/arm/tcg/a64.decode
@@ -256,6 +256,7 @@ ERETA 1101011 0100 11111 00001 m:1 11111 11111 &reta # ERETAA, ERETAB
AUTIASP 1101 0101 0000 0011 0010 0011 101 11111
AUTIBZ 1101 0101 0000 0011 0010 0011 110 11111
AUTIBSP 1101 0101 0000 0011 0010 0011 111 11111
+ CHKFEAT 1101 0101 0000 0011 0010 0101 000 11111
]
# The canonical NOP has CRm == op2 == 0, but all of the space
# that isn't specifically allocated to an instruction must NOP
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 37/76] target/arm: Make helper_exception_return system-only
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (35 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 36/76] target/arm: Implement FEAT_CHK Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 38/76] target/arm: Export cpsr_{read_for, write_from}_spsr_elx Richard Henderson
` (39 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/tcg/helper-a64.h | 5 ++++-
target/arm/tcg/helper-a64.c | 2 ++
target/arm/tcg/translate-a64.c | 8 ++++++++
3 files changed, 14 insertions(+), 1 deletion(-)
diff --git a/target/arm/tcg/helper-a64.h b/target/arm/tcg/helper-a64.h
index 85023465b7..b6008b5a3a 100644
--- a/target/arm/tcg/helper-a64.h
+++ b/target/arm/tcg/helper-a64.h
@@ -80,7 +80,6 @@ DEF_HELPER_3(vfp_ah_maxh, f16, f16, f16, fpst)
DEF_HELPER_3(vfp_ah_maxs, f32, f32, f32, fpst)
DEF_HELPER_3(vfp_ah_maxd, f64, f64, f64, fpst)
-DEF_HELPER_2(exception_return, void, env, i64)
DEF_HELPER_FLAGS_2(dc_zva, TCG_CALL_NO_WG, void, env, i64)
DEF_HELPER_FLAGS_3(pacia, TCG_CALL_NO_WG, i64, env, i64, i64)
@@ -145,3 +144,7 @@ DEF_HELPER_FLAGS_5(gvec_fmulx_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, fpst, i32
DEF_HELPER_FLAGS_5(gvec_fmulx_idx_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, fpst, i32)
DEF_HELPER_FLAGS_5(gvec_fmulx_idx_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, fpst, i32)
DEF_HELPER_FLAGS_5(gvec_fmulx_idx_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, fpst, i32)
+
+#ifndef CONFIG_USER_ONLY
+DEF_HELPER_2(exception_return, void, env, i64)
+#endif
diff --git a/target/arm/tcg/helper-a64.c b/target/arm/tcg/helper-a64.c
index 71c6c44ee8..6d77fd0113 100644
--- a/target/arm/tcg/helper-a64.c
+++ b/target/arm/tcg/helper-a64.c
@@ -576,6 +576,7 @@ uint32_t HELPER(advsimd_rinth)(uint32_t x, float_status *fp_status)
return ret;
}
+#ifndef CONFIG_USER_ONLY
static int el_from_spsr(uint32_t spsr)
{
/* Return the exception level that this SPSR is requesting a return to,
@@ -787,6 +788,7 @@ illegal_return:
qemu_log_mask(LOG_GUEST_ERROR, "Illegal exception return at EL%d: "
"resuming execution at 0x%" PRIx64 "\n", cur_el, env->pc);
}
+#endif /* !CONFIG_USER_ONLY */
void HELPER(dc_zva)(CPUARMState *env, uint64_t vaddr_in)
{
diff --git a/target/arm/tcg/translate-a64.c b/target/arm/tcg/translate-a64.c
index 9dfcf4f993..ce5d597c93 100644
--- a/target/arm/tcg/translate-a64.c
+++ b/target/arm/tcg/translate-a64.c
@@ -1870,6 +1870,9 @@ static bool trans_BLRA(DisasContext *s, arg_bra *a)
static bool trans_ERET(DisasContext *s, arg_ERET *a)
{
+#ifdef CONFIG_USER_ONLY
+ return false;
+#else
TCGv_i64 dst;
if (s->current_el == 0) {
@@ -1889,10 +1892,14 @@ static bool trans_ERET(DisasContext *s, arg_ERET *a)
/* Must exit loop to check un-masked IRQs */
s->base.is_jmp = DISAS_EXIT;
return true;
+#endif
}
static bool trans_ERETA(DisasContext *s, arg_reta *a)
{
+#ifdef CONFIG_USER_ONLY
+ return false;
+#else
TCGv_i64 dst;
if (!dc_isar_feature(aa64_pauth, s)) {
@@ -1918,6 +1925,7 @@ static bool trans_ERETA(DisasContext *s, arg_reta *a)
/* Must exit loop to check un-masked IRQs */
s->base.is_jmp = DISAS_EXIT;
return true;
+#endif
}
static bool trans_NOP(DisasContext *s, arg_NOP *a)
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 38/76] target/arm: Export cpsr_{read_for, write_from}_spsr_elx
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (36 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 37/76] target/arm: Make helper_exception_return system-only Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 39/76] target/arm: Expand pstate to 64 bits Richard Henderson
` (38 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm
Move cpsr_write_from_spsr_elx from tcg/helper-a64.c to
helper.c, so that it's present with --disable-tcg.
Declare both in internals.h.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/internals.h | 2 ++
target/arm/helper.c | 20 +++++++++++++++++++-
target/arm/tcg/helper-a64.c | 20 --------------------
3 files changed, 21 insertions(+), 21 deletions(-)
diff --git a/target/arm/internals.h b/target/arm/internals.h
index f72e95a5ba..3766a4ccfe 100644
--- a/target/arm/internals.h
+++ b/target/arm/internals.h
@@ -1892,6 +1892,8 @@ void vfp_clear_float_status_exc_flags(CPUARMState *env);
*/
void vfp_set_fpcr_to_host(CPUARMState *env, uint32_t val, uint32_t mask);
bool arm_pan_enabled(CPUARMState *env);
+uint32_t cpsr_read_for_spsr_elx(CPUARMState *env);
+void cpsr_write_from_spsr_elx(CPUARMState *env, uint32_t val);
/* Compare uint64_t for qsort and bsearch. */
int compare_u64(const void *a, const void *b);
diff --git a/target/arm/helper.c b/target/arm/helper.c
index 3b9e7a822e..3f32eb7a9d 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -8882,7 +8882,7 @@ static int aarch64_regnum(CPUARMState *env, int aarch32_reg)
}
}
-static uint32_t cpsr_read_for_spsr_elx(CPUARMState *env)
+uint32_t cpsr_read_for_spsr_elx(CPUARMState *env)
{
uint32_t ret = cpsr_read(env);
@@ -8897,6 +8897,24 @@ static uint32_t cpsr_read_for_spsr_elx(CPUARMState *env)
return ret;
}
+void cpsr_write_from_spsr_elx(CPUARMState *env, uint32_t val)
+{
+ uint32_t mask;
+
+ /* Save SPSR_ELx.SS into PSTATE. */
+ env->pstate = (env->pstate & ~PSTATE_SS) | (val & PSTATE_SS);
+ val &= ~PSTATE_SS;
+
+ /* Move DIT to the correct location for CPSR */
+ if (val & PSTATE_DIT) {
+ val &= ~PSTATE_DIT;
+ val |= CPSR_DIT;
+ }
+
+ mask = aarch32_cpsr_valid_mask(env->features, &env_archcpu(env)->isar);
+ cpsr_write(env, val, mask, CPSRWriteRaw);
+}
+
static bool syndrome_is_sync_extabt(uint32_t syndrome)
{
/* Return true if this syndrome value is a synchronous external abort */
diff --git a/target/arm/tcg/helper-a64.c b/target/arm/tcg/helper-a64.c
index 6d77fd0113..eaea7859d3 100644
--- a/target/arm/tcg/helper-a64.c
+++ b/target/arm/tcg/helper-a64.c
@@ -615,26 +615,6 @@ static int el_from_spsr(uint32_t spsr)
}
}
-static void cpsr_write_from_spsr_elx(CPUARMState *env,
- uint32_t val)
-{
- uint32_t mask;
-
- /* Save SPSR_ELx.SS into PSTATE. */
- env->pstate = (env->pstate & ~PSTATE_SS) | (val & PSTATE_SS);
- val &= ~PSTATE_SS;
-
- /* Move DIT to the correct location for CPSR */
- if (val & PSTATE_DIT) {
- val &= ~PSTATE_DIT;
- val |= CPSR_DIT;
- }
-
- mask = aarch32_cpsr_valid_mask(env->features, \
- &env_archcpu(env)->isar);
- cpsr_write(env, val, mask, CPSRWriteRaw);
-}
-
void HELPER(exception_return)(CPUARMState *env, uint64_t new_pc)
{
ARMCPU *cpu = env_archcpu(env);
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 39/76] target/arm: Expand pstate to 64 bits
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (37 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 38/76] target/arm: Export cpsr_{read_for, write_from}_spsr_elx Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 40/76] target/arm: Add syndrome data for EC_GCS Richard Henderson
` (37 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
The ARM now defines 36 bits in SPSR_ELx in aarch64 mode, so
it's time to bite the bullet and extend PSTATE to match.
Most changes are straightforward, adjusting printf formats,
changing local variable types. More complex is migration,
where to maintain backward compatibility a new pstate64
record is introduced, and only when one of the extensions
that sets bits 32-35 are active.
The fate of gdbstub is left undecided for the moment.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/cpu.h | 8 ++--
target/arm/tcg/translate.h | 20 ++++-----
target/arm/cpu.c | 6 +--
target/arm/gdbstub64.c | 2 +
target/arm/helper.c | 11 ++---
target/arm/machine.c | 81 +++++++++++++++++++++++++++++++++++++
target/arm/tcg/helper-a64.c | 2 +-
7 files changed, 107 insertions(+), 23 deletions(-)
diff --git a/target/arm/cpu.h b/target/arm/cpu.h
index 6a214557ac..c79393d9c2 100644
--- a/target/arm/cpu.h
+++ b/target/arm/cpu.h
@@ -268,7 +268,7 @@ typedef struct CPUArchState {
uint64_t xregs[32];
uint64_t pc;
/* PSTATE isn't an architectural register for ARMv8. However, it is
- * convenient for us to assemble the underlying state into a 32 bit format
+ * convenient for us to assemble the underlying state into a 64 bit format
* identical to the architectural format used for the SPSR. (This is also
* what the Linux kernel's 'pstate' field in signal handlers and KVM's
* 'pstate' register are.) Of the PSTATE bits:
@@ -280,7 +280,7 @@ typedef struct CPUArchState {
* SM and ZA are kept in env->svcr
* all other bits are stored in their correct places in env->pstate
*/
- uint32_t pstate;
+ uint64_t pstate;
bool aarch64; /* True if CPU is in aarch64 state; inverse of PSTATE.nRW */
bool thumb; /* True if CPU is in thumb mode; cpsr[5] */
@@ -1538,7 +1538,7 @@ static inline unsigned int aarch64_pstate_mode(unsigned int el, bool handler)
* interprocessing, so we don't attempt to sync with the cpsr state used by
* the 32 bit decoder.
*/
-static inline uint32_t pstate_read(CPUARMState *env)
+static inline uint64_t pstate_read(CPUARMState *env)
{
int ZF;
@@ -1548,7 +1548,7 @@ static inline uint32_t pstate_read(CPUARMState *env)
| env->pstate | env->daif | (env->btype << 10);
}
-static inline void pstate_write(CPUARMState *env, uint32_t val)
+static inline void pstate_write(CPUARMState *env, uint64_t val)
{
env->ZF = (~val) & PSTATE_Z;
env->NF = val;
diff --git a/target/arm/tcg/translate.h b/target/arm/tcg/translate.h
index 761edded52..943dfd45fe 100644
--- a/target/arm/tcg/translate.h
+++ b/target/arm/tcg/translate.h
@@ -378,27 +378,27 @@ static inline TCGv_i32 get_ahp_flag(void)
}
/* Set bits within PSTATE. */
-static inline void set_pstate_bits(uint32_t bits)
+static inline void set_pstate_bits(uint64_t bits)
{
- TCGv_i32 p = tcg_temp_new_i32();
+ TCGv_i64 p = tcg_temp_new_i64();
tcg_debug_assert(!(bits & CACHED_PSTATE_BITS));
- tcg_gen_ld_i32(p, tcg_env, offsetof(CPUARMState, pstate));
- tcg_gen_ori_i32(p, p, bits);
- tcg_gen_st_i32(p, tcg_env, offsetof(CPUARMState, pstate));
+ tcg_gen_ld_i64(p, tcg_env, offsetof(CPUARMState, pstate));
+ tcg_gen_ori_i64(p, p, bits);
+ tcg_gen_st_i64(p, tcg_env, offsetof(CPUARMState, pstate));
}
/* Clear bits within PSTATE. */
-static inline void clear_pstate_bits(uint32_t bits)
+static inline void clear_pstate_bits(uint64_t bits)
{
- TCGv_i32 p = tcg_temp_new_i32();
+ TCGv_i64 p = tcg_temp_new_i64();
tcg_debug_assert(!(bits & CACHED_PSTATE_BITS));
- tcg_gen_ld_i32(p, tcg_env, offsetof(CPUARMState, pstate));
- tcg_gen_andi_i32(p, p, ~bits);
- tcg_gen_st_i32(p, tcg_env, offsetof(CPUARMState, pstate));
+ tcg_gen_ld_i64(p, tcg_env, offsetof(CPUARMState, pstate));
+ tcg_gen_andi_i64(p, p, ~bits);
+ tcg_gen_st_i64(p, tcg_env, offsetof(CPUARMState, pstate));
}
/* If the singlestep state is Active-not-pending, advance to Active-pending. */
diff --git a/target/arm/cpu.c b/target/arm/cpu.c
index aab7250399..731b4d5725 100644
--- a/target/arm/cpu.c
+++ b/target/arm/cpu.c
@@ -1190,7 +1190,7 @@ static void aarch64_cpu_dump_state(CPUState *cs, FILE *f, int flags)
{
ARMCPU *cpu = ARM_CPU(cs);
CPUARMState *env = &cpu->env;
- uint32_t psr = pstate_read(env);
+ uint64_t psr = pstate_read(env);
int i, j;
int el = arm_current_el(env);
uint64_t hcr = arm_hcr_el2_eff(env);
@@ -1212,7 +1212,7 @@ static void aarch64_cpu_dump_state(CPUState *cs, FILE *f, int flags)
} else {
ns_status = "";
}
- qemu_fprintf(f, "PSTATE=%08x %c%c%c%c %sEL%d%c",
+ qemu_fprintf(f, "PSTATE=%016" PRIx64 " %c%c%c%c %sEL%d%c",
psr,
psr & PSTATE_N ? 'N' : '-',
psr & PSTATE_Z ? 'Z' : '-',
@@ -1229,7 +1229,7 @@ static void aarch64_cpu_dump_state(CPUState *cs, FILE *f, int flags)
(FIELD_EX64(env->svcr, SVCR, SM) ? 'S' : '-'));
}
if (cpu_isar_feature(aa64_bti, cpu)) {
- qemu_fprintf(f, " BTYPE=%d", (psr & PSTATE_BTYPE) >> 10);
+ qemu_fprintf(f, " BTYPE=%d", (int)(psr & PSTATE_BTYPE) >> 10);
}
qemu_fprintf(f, "%s%s%s",
(hcr & HCR_NV) ? " NV" : "",
diff --git a/target/arm/gdbstub64.c b/target/arm/gdbstub64.c
index 3bccde2bf2..65d6bbe65f 100644
--- a/target/arm/gdbstub64.c
+++ b/target/arm/gdbstub64.c
@@ -47,6 +47,7 @@ int aarch64_cpu_gdb_read_register(CPUState *cs, GByteArray *mem_buf, int n)
case 32:
return gdb_get_reg64(mem_buf, env->pc);
case 33:
+ /* pstate is now a 64-bit value; can we simply adjust the xml? */
return gdb_get_reg32(mem_buf, pstate_read(env));
}
/* Unknown register. */
@@ -75,6 +76,7 @@ int aarch64_cpu_gdb_write_register(CPUState *cs, uint8_t *mem_buf, int n)
return 8;
case 33:
/* CPSR */
+ /* pstate is now a 64-bit value; can we simply adjust the xml? */
pstate_write(env, tmp);
return 4;
}
diff --git a/target/arm/helper.c b/target/arm/helper.c
index 3f32eb7a9d..37cfd2064c 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -8947,8 +8947,8 @@ static void arm_cpu_do_interrupt_aarch64(CPUState *cs)
CPUARMState *env = &cpu->env;
unsigned int new_el = env->exception.target_el;
vaddr addr = env->cp15.vbar_el[new_el];
- unsigned int new_mode = aarch64_pstate_mode(new_el, true);
- unsigned int old_mode;
+ uint64_t new_mode = aarch64_pstate_mode(new_el, true);
+ uint64_t old_mode;
unsigned int cur_el = arm_current_el(env);
int rt;
@@ -9096,7 +9096,7 @@ static void arm_cpu_do_interrupt_aarch64(CPUState *cs)
* If NV2 is disabled, change SPSR when NV,NV1 == 1,0 (I_ZJRNN)
* If NV2 is enabled, change SPSR when NV is 1 (I_DBTLM)
*/
- old_mode = deposit32(old_mode, 2, 2, 2);
+ old_mode = deposit64(old_mode, 2, 2, 2);
}
}
} else {
@@ -9109,7 +9109,7 @@ static void arm_cpu_do_interrupt_aarch64(CPUState *cs)
}
env->banked_spsr[aarch64_banked_spsr_index(new_el)] = old_mode;
- qemu_log_mask(CPU_LOG_INT, "...with SPSR 0x%x\n", old_mode);
+ qemu_log_mask(CPU_LOG_INT, "...with SPSR 0x%" PRIx64 "\n", old_mode);
qemu_log_mask(CPU_LOG_INT, "...with ELR 0x%" PRIx64 "\n",
env->elr_el[new_el]);
@@ -9163,7 +9163,8 @@ static void arm_cpu_do_interrupt_aarch64(CPUState *cs)
env->pc = addr;
- qemu_log_mask(CPU_LOG_INT, "...to EL%d PC 0x%" PRIx64 " PSTATE 0x%x\n",
+ qemu_log_mask(CPU_LOG_INT, "...to EL%d PC 0x%" PRIx64
+ " PSTATE 0x%" PRIx64 "\n",
new_el, env->pc, pstate_read(env));
}
diff --git a/target/arm/machine.c b/target/arm/machine.c
index ce20b46f50..44a0cf844b 100644
--- a/target/arm/machine.c
+++ b/target/arm/machine.c
@@ -816,6 +816,80 @@ static const VMStateInfo vmstate_cpsr = {
.put = put_cpsr,
};
+static int get_pstate64(QEMUFile *f, void *opaque, size_t size,
+ const VMStateField *field)
+{
+ ARMCPU *cpu = opaque;
+ CPUARMState *env = &cpu->env;
+ uint64_t val = qemu_get_be64(f);
+
+ env->aarch64 = ((val & PSTATE_nRW) == 0);
+ if (is_a64(env)) {
+ pstate_write(env, val);
+ } else {
+ cpsr_write_from_spsr_elx(env, val);
+ }
+ return 0;
+}
+
+static int put_pstate64(QEMUFile *f, void *opaque, size_t size,
+ const VMStateField *field, JSONWriter *vmdesc)
+{
+ ARMCPU *cpu = opaque;
+ CPUARMState *env = &cpu->env;
+ uint64_t val;
+
+ if (is_a64(env)) {
+ val = pstate_read(env);
+ } else {
+ val = cpsr_read_for_spsr_elx(env);
+ }
+ qemu_put_be64(f, val);
+ return 0;
+}
+
+static bool pstate64_needed(void *opaque)
+{
+ ARMCPU *cpu = opaque;
+ CPUARMState *env = &cpu->env;
+ uint64_t val;
+
+ if (arm_feature(env, ARM_FEATURE_M)) {
+ return false;
+ }
+ if (is_a64(env)) {
+ val = pstate_read(env);
+ } else {
+ val = cpsr_read_for_spsr_elx(env);
+ if (val & PSTATE_SS) {
+ return true;
+ }
+ }
+ return val > UINT32_MAX;
+}
+
+static const VMStateDescription vmstate_pstate64 = {
+ .name = "cpu/pstate64",
+ .version_id = 1,
+ .minimum_version_id = 1,
+ .needed = pstate64_needed,
+ .fields = (const VMStateField[]) {
+ {
+ .name = "pstate64",
+ .version_id = 0,
+ .size = sizeof(uint64_t),
+ .info = &(const VMStateInfo) {
+ .name = "pstate64",
+ .get = get_pstate64,
+ .put = put_pstate64,
+ },
+ .flags = VMS_SINGLE,
+ .offset = 0,
+ },
+ VMSTATE_END_OF_LIST()
+ },
+};
+
static int get_power(QEMUFile *f, void *opaque, size_t size,
const VMStateField *field)
{
@@ -1052,6 +1126,12 @@ const VMStateDescription vmstate_arm_cpu = {
VMSTATE_UINT32_ARRAY(env.regs, ARMCPU, 16),
VMSTATE_UINT64_ARRAY(env.xregs, ARMCPU, 32),
VMSTATE_UINT64(env.pc, ARMCPU),
+ /*
+ * If any bits are set in the upper 32 bits of cpsr/pstate,
+ * or if the cpu is in aa32 mode and PSTATE.SS is set, then
+ * the cpu/pstate64 subsection will override this with the
+ * full 64 bit state.
+ */
{
.name = "cpsr",
.version_id = 0,
@@ -1128,6 +1208,7 @@ const VMStateDescription vmstate_arm_cpu = {
&vmstate_irq_line_state,
&vmstate_wfxt_timer,
&vmstate_syndrome64,
+ &vmstate_pstate64,
NULL
}
};
diff --git a/target/arm/tcg/helper-a64.c b/target/arm/tcg/helper-a64.c
index eaea7859d3..08b7db7c46 100644
--- a/target/arm/tcg/helper-a64.c
+++ b/target/arm/tcg/helper-a64.c
@@ -620,7 +620,7 @@ void HELPER(exception_return)(CPUARMState *env, uint64_t new_pc)
ARMCPU *cpu = env_archcpu(env);
int cur_el = arm_current_el(env);
unsigned int spsr_idx = aarch64_banked_spsr_index(cur_el);
- uint32_t spsr = env->banked_spsr[spsr_idx];
+ uint64_t spsr = env->banked_spsr[spsr_idx];
int new_el;
bool return_to_aa64 = (spsr & PSTATE_nRW) == 0;
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 40/76] target/arm: Add syndrome data for EC_GCS
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (38 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 39/76] target/arm: Expand pstate to 64 bits Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 41/76] target/arm: Add arm_hcr_el2_nvx_eff Richard Henderson
` (36 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/syndrome.h | 35 +++++++++++++++++++++++++++++++++++
1 file changed, 35 insertions(+)
diff --git a/target/arm/syndrome.h b/target/arm/syndrome.h
index c48d3b8587..bff61f052c 100644
--- a/target/arm/syndrome.h
+++ b/target/arm/syndrome.h
@@ -63,6 +63,7 @@ enum arm_exception_class {
EC_MOP = 0x27,
EC_AA32_FPTRAP = 0x28,
EC_AA64_FPTRAP = 0x2c,
+ EC_GCS = 0x2d,
EC_SERROR = 0x2f,
EC_BREAKPOINT = 0x30,
EC_BREAKPOINT_SAME_EL = 0x31,
@@ -83,6 +84,23 @@ typedef enum {
SME_ET_InaccessibleZT0,
} SMEExceptionType;
+typedef enum {
+ GCS_ET_DataCheck,
+ GCS_ET_EXLOCK,
+ GCS_ET_GCSSTR_GCSSTTR,
+} GCSExceptionType;
+
+typedef enum {
+ GCS_IT_RET_nPauth = 0,
+ GCS_IT_GCSPOPM = 1,
+ GCS_IT_RET_PauthA = 2,
+ GCS_IT_RET_PauthB = 3,
+ GCS_IT_GCSSS1 = 4,
+ GCS_IT_GCSSS2 = 5,
+ GCS_IT_GCSPOPCX = 8,
+ GCS_IT_GCSPOPX = 9,
+} GCSInstructionType;
+
#define ARM_EL_EC_LENGTH 6
#define ARM_EL_EC_SHIFT 26
#define ARM_EL_IL_SHIFT 25
@@ -351,6 +369,23 @@ static inline uint32_t syn_pcalignment(void)
return (EC_PCALIGNMENT << ARM_EL_EC_SHIFT) | ARM_EL_IL;
}
+static inline uint32_t syn_gcs_data_check(GCSInstructionType it, int rn)
+{
+ return ((EC_GCS << ARM_EL_EC_SHIFT) | ARM_EL_IL |
+ (GCS_ET_DataCheck << 20) | (rn << 5) | it);
+}
+
+static inline uint32_t syn_gcs_exlock(void)
+{
+ return (EC_GCS << ARM_EL_EC_SHIFT) | ARM_EL_IL | (GCS_ET_EXLOCK << 20);
+}
+
+static inline uint32_t syn_gcs_gcsstr(int ra, int rn)
+{
+ return ((EC_GCS << ARM_EL_EC_SHIFT) | ARM_EL_IL |
+ (GCS_ET_GCSSTR_GCSSTTR << 20) | (ra << 10) | (rn << 5));
+}
+
static inline uint32_t syn_serror(uint32_t extra)
{
return (EC_SERROR << ARM_EL_EC_SHIFT) | ARM_EL_IL | extra;
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 41/76] target/arm: Add arm_hcr_el2_nvx_eff
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (39 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 40/76] target/arm: Add syndrome data for EC_GCS Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 42/76] target/arm: Use arm_hcr_el2_nvx_eff in access_nv1 Richard Henderson
` (35 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm
Implement the pseudocode function EffectiveHCR_EL2_NVx.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/cpu.h | 1 +
target/arm/helper.c | 10 ++++++++++
2 files changed, 11 insertions(+)
diff --git a/target/arm/cpu.h b/target/arm/cpu.h
index c79393d9c2..b36436ee2b 100644
--- a/target/arm/cpu.h
+++ b/target/arm/cpu.h
@@ -2639,6 +2639,7 @@ static inline bool arm_is_el2_enabled(CPUARMState *env)
*/
uint64_t arm_hcr_el2_eff_secstate(CPUARMState *env, ARMSecuritySpace space);
uint64_t arm_hcr_el2_eff(CPUARMState *env);
+uint64_t arm_hcr_el2_nvx_eff(CPUARMState *env);
uint64_t arm_hcrx_el2_eff(CPUARMState *env);
/*
diff --git a/target/arm/helper.c b/target/arm/helper.c
index 37cfd2064c..e3ea5254c5 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -3899,6 +3899,16 @@ uint64_t arm_hcr_el2_eff(CPUARMState *env)
return arm_hcr_el2_eff_secstate(env, arm_security_space_below_el3(env));
}
+uint64_t arm_hcr_el2_nvx_eff(CPUARMState *env)
+{
+ uint64_t hcr = arm_hcr_el2_eff(env);
+
+ if (!(hcr & HCR_NV)) {
+ return 0; /* CONSTRAINED UNPREDICTABLE wrt NV1 */
+ }
+ return hcr & (HCR_NV2 | HCR_NV1 | HCR_NV);
+}
+
/*
* Corresponds to ARM pseudocode function ELIsInHost().
*/
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 42/76] target/arm: Use arm_hcr_el2_nvx_eff in access_nv1
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (40 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 41/76] target/arm: Add arm_hcr_el2_nvx_eff Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 43/76] target/arm: Split out access_nv1_with_nvx Richard Henderson
` (34 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/helper.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/target/arm/helper.c b/target/arm/helper.c
index e3ea5254c5..8b5ae0c543 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -3423,7 +3423,7 @@ static CPAccessResult access_nv1(CPUARMState *env, const ARMCPRegInfo *ri,
bool isread)
{
if (arm_current_el(env) == 1) {
- uint64_t hcr_nv = arm_hcr_el2_eff(env) & (HCR_NV | HCR_NV1 | HCR_NV2);
+ uint64_t hcr_nv = arm_hcr_el2_nvx_eff(env);
if (hcr_nv == (HCR_NV | HCR_NV1)) {
return CP_ACCESS_TRAP_EL2;
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 43/76] target/arm: Split out access_nv1_with_nvx
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (41 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 42/76] target/arm: Use arm_hcr_el2_nvx_eff in access_nv1 Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 44/76] target/arm: Implement EXLOCKException for ELR_ELx and SPSR_ELx Richard Henderson
` (33 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/helper.c | 11 ++++++-----
1 file changed, 6 insertions(+), 5 deletions(-)
diff --git a/target/arm/helper.c b/target/arm/helper.c
index 8b5ae0c543..5519484186 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -3419,15 +3419,16 @@ static void mdcr_el2_write(CPUARMState *env, const ARMCPRegInfo *ri,
}
}
+static CPAccessResult access_nv1_with_nvx(uint64_t hcr_nv)
+{
+ return hcr_nv == (HCR_NV | HCR_NV1) ? CP_ACCESS_TRAP_EL2 : CP_ACCESS_OK;
+}
+
static CPAccessResult access_nv1(CPUARMState *env, const ARMCPRegInfo *ri,
bool isread)
{
if (arm_current_el(env) == 1) {
- uint64_t hcr_nv = arm_hcr_el2_nvx_eff(env);
-
- if (hcr_nv == (HCR_NV | HCR_NV1)) {
- return CP_ACCESS_TRAP_EL2;
- }
+ return access_nv1_with_nvx(arm_hcr_el2_nvx_eff(env));
}
return CP_ACCESS_OK;
}
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 44/76] target/arm: Implement EXLOCKException for ELR_ELx and SPSR_ELx
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (42 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 43/76] target/arm: Split out access_nv1_with_nvx Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 45/76] target/arm: Split {full,core}_a64_user_mem_index Richard Henderson
` (32 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
If PSTATE.EXLOCK is set, and the GCS EXLOCK enable bit is set,
and nested virt is in the appropriate state, then we need to
raise an EXLOCK exception.
Since PSTATE.EXLOCK cannot be set without GCS being present
and enabled, no explicit check for GCS is required.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/cpregs.h | 8 +++++
target/arm/cpu.h | 1 +
target/arm/helper.c | 67 ++++++++++++++++++++++++++++++++++----
target/arm/tcg/op_helper.c | 7 ++++
4 files changed, 77 insertions(+), 6 deletions(-)
diff --git a/target/arm/cpregs.h b/target/arm/cpregs.h
index bd2121a336..a79f00351c 100644
--- a/target/arm/cpregs.h
+++ b/target/arm/cpregs.h
@@ -351,6 +351,14 @@ typedef enum CPAccessResult {
* specified target EL.
*/
CP_ACCESS_UNDEFINED = (2 << 2),
+
+ /*
+ * Access fails with EXLOCK, a GCS exception syndrome.
+ * These traps are always to the current execution EL,
+ * which is the same as the usual target EL because
+ * they cannot occur from EL0.
+ */
+ CP_ACCESS_EXLOCK = (3 << 2),
} CPAccessResult;
/* Indexes into fgt_read[] */
diff --git a/target/arm/cpu.h b/target/arm/cpu.h
index b36436ee2b..97cdcd8cdc 100644
--- a/target/arm/cpu.h
+++ b/target/arm/cpu.h
@@ -1502,6 +1502,7 @@ void pmu_init(ARMCPU *cpu);
#define PSTATE_C (1U << 29)
#define PSTATE_Z (1U << 30)
#define PSTATE_N (1U << 31)
+#define PSTATE_EXLOCK (1ULL << 34)
#define PSTATE_NZCV (PSTATE_N | PSTATE_Z | PSTATE_C | PSTATE_V)
#define PSTATE_DAIF (PSTATE_D | PSTATE_A | PSTATE_I | PSTATE_F)
#define CACHED_PSTATE_BITS (PSTATE_NZCV | PSTATE_DAIF | PSTATE_BTYPE)
diff --git a/target/arm/helper.c b/target/arm/helper.c
index 5519484186..e90398acc9 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -3433,6 +3433,61 @@ static CPAccessResult access_nv1(CPUARMState *env, const ARMCPRegInfo *ri,
return CP_ACCESS_OK;
}
+static CPAccessResult access_nv1_or_exlock_el1(CPUARMState *env,
+ const ARMCPRegInfo *ri,
+ bool isread)
+{
+ if (arm_current_el(env) == 1) {
+ uint64_t nvx = arm_hcr_el2_nvx_eff(env);
+
+ if (!isread &&
+ (env->pstate & PSTATE_EXLOCK) &&
+ (env->cp15.gcscr_el[1] & GCSCR_EXLOCKEN) &&
+ !(nvx & HCR_NV1)) {
+ return CP_ACCESS_EXLOCK;
+ }
+ return access_nv1_with_nvx(nvx);
+ }
+
+ /*
+ * At EL2, since VHE redirection is done at translation time,
+ * el_is_in_host is always false here, so EXLOCK does not apply.
+ */
+ return CP_ACCESS_OK;
+}
+
+static CPAccessResult access_exlock_el2(CPUARMState *env,
+ const ARMCPRegInfo *ri, bool isread)
+{
+ int el = arm_current_el(env);
+
+ if (el == 3) {
+ return CP_ACCESS_OK;
+ }
+
+ /*
+ * Access to the EL2 register from EL1 means NV is set, and
+ * EXLOCK has priority over an NV1 trap to EL2.
+ */
+ if (!isread &&
+ (env->pstate & PSTATE_EXLOCK) &&
+ (env->cp15.gcscr_el[el] & GCSCR_EXLOCKEN)) {
+ return CP_ACCESS_EXLOCK;
+ }
+ return CP_ACCESS_OK;
+}
+
+static CPAccessResult access_exlock_el3(CPUARMState *env,
+ const ARMCPRegInfo *ri, bool isread)
+{
+ if (!isread &&
+ (env->pstate & PSTATE_EXLOCK) &&
+ (env->cp15.gcscr_el[3] & GCSCR_EXLOCKEN)) {
+ return CP_ACCESS_EXLOCK;
+ }
+ return CP_ACCESS_OK;
+}
+
#ifdef CONFIG_USER_ONLY
/*
* `IC IVAU` is handled to improve compatibility with JITs that dual-map their
@@ -3604,7 +3659,7 @@ static const ARMCPRegInfo v8_cp_reginfo[] = {
{ .name = "ELR_EL1", .state = ARM_CP_STATE_AA64,
.type = ARM_CP_ALIAS,
.opc0 = 3, .opc1 = 0, .crn = 4, .crm = 0, .opc2 = 1,
- .access = PL1_RW, .accessfn = access_nv1,
+ .access = PL1_RW, .accessfn = access_nv1_or_exlock_el1,
.nv2_redirect_offset = 0x230 | NV2_REDIR_NV1,
.vhe_redir_to_el2 = ENCODE_AA64_CP_REG(3, 4, 4, 0, 1),
.vhe_redir_to_el01 = ENCODE_AA64_CP_REG(3, 5, 4, 0, 1),
@@ -3612,7 +3667,7 @@ static const ARMCPRegInfo v8_cp_reginfo[] = {
{ .name = "SPSR_EL1", .state = ARM_CP_STATE_AA64,
.type = ARM_CP_ALIAS,
.opc0 = 3, .opc1 = 0, .crn = 4, .crm = 0, .opc2 = 0,
- .access = PL1_RW, .accessfn = access_nv1,
+ .access = PL1_RW, .accessfn = access_nv1_or_exlock_el1,
.nv2_redirect_offset = 0x160 | NV2_REDIR_NV1,
.vhe_redir_to_el2 = ENCODE_AA64_CP_REG(3, 4, 4, 0, 0),
.vhe_redir_to_el01 = ENCODE_AA64_CP_REG(3, 5, 4, 0, 0),
@@ -4095,7 +4150,7 @@ static const ARMCPRegInfo el2_cp_reginfo[] = {
{ .name = "ELR_EL2", .state = ARM_CP_STATE_AA64,
.type = ARM_CP_ALIAS | ARM_CP_NV2_REDIRECT,
.opc0 = 3, .opc1 = 4, .crn = 4, .crm = 0, .opc2 = 1,
- .access = PL2_RW,
+ .access = PL2_RW, .accessfn = access_exlock_el2,
.fieldoffset = offsetof(CPUARMState, elr_el[2]) },
{ .name = "ESR_EL2", .state = ARM_CP_STATE_BOTH,
.type = ARM_CP_NV2_REDIRECT,
@@ -4113,7 +4168,7 @@ static const ARMCPRegInfo el2_cp_reginfo[] = {
{ .name = "SPSR_EL2", .state = ARM_CP_STATE_AA64,
.type = ARM_CP_ALIAS | ARM_CP_NV2_REDIRECT,
.opc0 = 3, .opc1 = 4, .crn = 4, .crm = 0, .opc2 = 0,
- .access = PL2_RW,
+ .access = PL2_RW, .accessfn = access_exlock_el2,
.fieldoffset = offsetof(CPUARMState, banked_spsr[BANK_HYP]) },
{ .name = "VBAR_EL2", .state = ARM_CP_STATE_BOTH,
.opc0 = 3, .opc1 = 4, .crn = 12, .crm = 0, .opc2 = 0,
@@ -4395,7 +4450,7 @@ static const ARMCPRegInfo el3_cp_reginfo[] = {
{ .name = "ELR_EL3", .state = ARM_CP_STATE_AA64,
.type = ARM_CP_ALIAS,
.opc0 = 3, .opc1 = 6, .crn = 4, .crm = 0, .opc2 = 1,
- .access = PL3_RW,
+ .access = PL3_RW, .accessfn = access_exlock_el3,
.fieldoffset = offsetof(CPUARMState, elr_el[3]) },
{ .name = "ESR_EL3", .state = ARM_CP_STATE_AA64,
.opc0 = 3, .opc1 = 6, .crn = 5, .crm = 2, .opc2 = 0,
@@ -4406,7 +4461,7 @@ static const ARMCPRegInfo el3_cp_reginfo[] = {
{ .name = "SPSR_EL3", .state = ARM_CP_STATE_AA64,
.type = ARM_CP_ALIAS,
.opc0 = 3, .opc1 = 6, .crn = 4, .crm = 0, .opc2 = 0,
- .access = PL3_RW,
+ .access = PL3_RW, .accessfn = access_exlock_el3,
.fieldoffset = offsetof(CPUARMState, banked_spsr[BANK_MON]) },
{ .name = "VBAR_EL3", .state = ARM_CP_STATE_AA64,
.opc0 = 3, .opc1 = 6, .crn = 12, .crm = 0, .opc2 = 0,
diff --git a/target/arm/tcg/op_helper.c b/target/arm/tcg/op_helper.c
index dd3700dc6f..4fbd219555 100644
--- a/target/arm/tcg/op_helper.c
+++ b/target/arm/tcg/op_helper.c
@@ -881,6 +881,13 @@ const void *HELPER(access_check_cp_reg)(CPUARMState *env, uint32_t key,
}
syndrome = syn_uncategorized();
break;
+ case CP_ACCESS_EXLOCK:
+ /*
+ * CP_ACCESS_EXLOCK is always directed to the current EL,
+ * which is going to be the same as the usual target EL.
+ */
+ syndrome = syn_gcs_exlock();
+ break;
default:
g_assert_not_reached();
}
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 45/76] target/arm: Split {full,core}_a64_user_mem_index
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (43 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 44/76] target/arm: Implement EXLOCKException for ELR_ELx and SPSR_ELx Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 46/76] target/arm: Introduce delay_exception{_el} Richard Henderson
` (31 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Separate get_a64_user_mem_index into two separate functions, one which
returns the full ARMMMUIdx and one which returns the core mmu_idx.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/tcg/translate-a64.c | 24 +++++++++++++++---------
1 file changed, 15 insertions(+), 9 deletions(-)
diff --git a/target/arm/tcg/translate-a64.c b/target/arm/tcg/translate-a64.c
index ce5d597c93..f0331830b5 100644
--- a/target/arm/tcg/translate-a64.c
+++ b/target/arm/tcg/translate-a64.c
@@ -93,7 +93,7 @@ void a64_translate_init(void)
}
/*
- * Return the core mmu_idx to use for A64 load/store insns which
+ * Return the full arm mmu_idx to use for A64 load/store insns which
* have a "unprivileged load/store" variant. Those insns access
* EL0 if executed from an EL which has control over EL0 (usually
* EL1) but behave like normal loads and stores if executed from
@@ -103,7 +103,7 @@ void a64_translate_init(void)
* normal encoding (in which case we will return the same
* thing as get_mem_index().
*/
-static int get_a64_user_mem_index(DisasContext *s, bool unpriv)
+static ARMMMUIdx full_a64_user_mem_index(DisasContext *s, bool unpriv)
{
/*
* If AccType_UNPRIV is not used, the insn uses AccType_NORMAL,
@@ -130,7 +130,13 @@ static int get_a64_user_mem_index(DisasContext *s, bool unpriv)
g_assert_not_reached();
}
}
- return arm_to_core_mmu_idx(useridx);
+ return useridx;
+}
+
+/* Return the core mmu_idx per above. */
+static int core_a64_user_mem_index(DisasContext *s, bool unpriv)
+{
+ return arm_to_core_mmu_idx(full_a64_user_mem_index(s, unpriv));
}
static void set_btype_raw(int val)
@@ -3577,7 +3583,7 @@ static void op_addr_ldst_imm_pre(DisasContext *s, arg_ldst_imm *a,
if (!a->p) {
tcg_gen_addi_i64(*dirty_addr, *dirty_addr, offset);
}
- memidx = get_a64_user_mem_index(s, a->unpriv);
+ memidx = core_a64_user_mem_index(s, a->unpriv);
*clean_addr = gen_mte_check1_mmuidx(s, *dirty_addr, is_store,
a->w || a->rn != 31,
mop, a->unpriv, memidx);
@@ -3598,7 +3604,7 @@ static bool trans_STR_i(DisasContext *s, arg_ldst_imm *a)
{
bool iss_sf, iss_valid = !a->w;
TCGv_i64 clean_addr, dirty_addr, tcg_rt;
- int memidx = get_a64_user_mem_index(s, a->unpriv);
+ int memidx = core_a64_user_mem_index(s, a->unpriv);
MemOp mop = finalize_memop(s, a->sz + a->sign * MO_SIGN);
op_addr_ldst_imm_pre(s, a, &clean_addr, &dirty_addr, a->imm, true, mop);
@@ -3616,7 +3622,7 @@ static bool trans_LDR_i(DisasContext *s, arg_ldst_imm *a)
{
bool iss_sf, iss_valid = !a->w;
TCGv_i64 clean_addr, dirty_addr, tcg_rt;
- int memidx = get_a64_user_mem_index(s, a->unpriv);
+ int memidx = core_a64_user_mem_index(s, a->unpriv);
MemOp mop = finalize_memop(s, a->sz + a->sign * MO_SIGN);
op_addr_ldst_imm_pre(s, a, &clean_addr, &dirty_addr, a->imm, false, mop);
@@ -4514,7 +4520,7 @@ static bool do_SET(DisasContext *s, arg_set *a, bool is_epilogue,
return false;
}
- memidx = get_a64_user_mem_index(s, a->unpriv);
+ memidx = core_a64_user_mem_index(s, a->unpriv);
/*
* We pass option_a == true, matching our implementation;
@@ -4568,8 +4574,8 @@ static bool do_CPY(DisasContext *s, arg_cpy *a, bool is_epilogue, CpyFn fn)
return false;
}
- rmemidx = get_a64_user_mem_index(s, runpriv);
- wmemidx = get_a64_user_mem_index(s, wunpriv);
+ rmemidx = core_a64_user_mem_index(s, runpriv);
+ wmemidx = core_a64_user_mem_index(s, wunpriv);
/*
* We pass option_a == true, matching our implementation;
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 46/76] target/arm: Introduce delay_exception{_el}
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (44 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 45/76] target/arm: Split {full,core}_a64_user_mem_index Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 47/76] target/arm: Emit HSTR trap exception out of line Richard Henderson
` (30 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Add infrastructure to raise an exception out of line.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/tcg/translate.h | 20 +++++++++++++
target/arm/tcg/translate-a64.c | 2 ++
target/arm/tcg/translate.c | 53 ++++++++++++++++++++++++++++++++++
3 files changed, 75 insertions(+)
diff --git a/target/arm/tcg/translate.h b/target/arm/tcg/translate.h
index 943dfd45fe..9a85ea74db 100644
--- a/target/arm/tcg/translate.h
+++ b/target/arm/tcg/translate.h
@@ -21,9 +21,25 @@ typedef struct DisasLabel {
target_ulong pc_save;
} DisasLabel;
+/*
+ * Emit an exception call out of line.
+ */
+typedef struct DisasDelayException {
+ struct DisasDelayException *next;
+ TCGLabel *lab;
+ target_long pc_curr;
+ target_long pc_save;
+ int condexec_mask;
+ int condexec_cond;
+ uint32_t excp;
+ uint32_t syn;
+ uint32_t target_el;
+} DisasDelayException;
+
typedef struct DisasContext {
DisasContextBase base;
const ARMISARegisters *isar;
+ DisasDelayException *delay_excp_list;
/* The address of the current instruction being translated. */
target_ulong pc_curr;
@@ -365,6 +381,10 @@ 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, target_long pc_diff,
int excp, uint32_t syn);
+TCGLabel *delay_exception_el(DisasContext *s, int excp,
+ uint32_t syn, uint32_t target_el);
+TCGLabel *delay_exception(DisasContext *s, int excp, uint32_t syn);
+void emit_delayed_exceptions(DisasContext *s);
/* Return state of Alternate Half-precision flag, caller frees result */
static inline TCGv_i32 get_ahp_flag(void)
diff --git a/target/arm/tcg/translate-a64.c b/target/arm/tcg/translate-a64.c
index f0331830b5..d607a0afd6 100644
--- a/target/arm/tcg/translate-a64.c
+++ b/target/arm/tcg/translate-a64.c
@@ -10601,6 +10601,8 @@ static void aarch64_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu)
break;
}
}
+
+ emit_delayed_exceptions(dc);
}
const TranslatorOps aarch64_translator_ops = {
diff --git a/target/arm/tcg/translate.c b/target/arm/tcg/translate.c
index e62dcc5d85..78d26aac04 100644
--- a/target/arm/tcg/translate.c
+++ b/target/arm/tcg/translate.c
@@ -1088,6 +1088,57 @@ void gen_exception_insn(DisasContext *s, target_long pc_diff,
s->base.is_jmp = DISAS_NORETURN;
}
+TCGLabel *delay_exception_el(DisasContext *s, int excp,
+ uint32_t syn, uint32_t target_el)
+{
+ /* Use tcg_malloc for automatic release on longjmp out of translation. */
+ DisasDelayException *e = tcg_malloc(sizeof(DisasDelayException));
+
+ memset(e, 0, sizeof(*e));
+
+ /* Save enough of the current state to satisfy gen_exception_insn. */
+ e->pc_curr = s->pc_curr;
+ e->pc_save = s->pc_save;
+ if (!s->aarch64) {
+ e->condexec_cond = s->condexec_cond;
+ e->condexec_mask = s->condexec_mask;
+ }
+
+ e->excp = excp;
+ e->syn = syn;
+ e->target_el = target_el;
+
+ e->next = s->delay_excp_list;
+ s->delay_excp_list = e;
+
+ e->lab = gen_new_label();
+ return e->lab;
+}
+
+TCGLabel *delay_exception(DisasContext *s, int excp, uint32_t syn)
+{
+ return delay_exception_el(s, excp, syn, 0);
+}
+
+void emit_delayed_exceptions(DisasContext *s)
+{
+ for (DisasDelayException *e = s->delay_excp_list; e ; e = e->next) {
+ gen_set_label(e->lab);
+
+ /* Restore the insn state to satisfy gen_exception_insn. */
+ s->pc_curr = e->pc_curr;
+ s->pc_save = e->pc_save;
+ s->condexec_cond = e->condexec_cond;
+ s->condexec_mask = e->condexec_mask;
+
+ if (e->target_el) {
+ gen_exception_insn_el(s, 0, e->excp, e->syn, e->target_el);
+ } else {
+ gen_exception_insn(s, 0, e->excp, e->syn);
+ }
+ }
+}
+
static void gen_exception_bkpt_insn(DisasContext *s, uint32_t syn)
{
gen_set_condexec(s);
@@ -6791,6 +6842,8 @@ static void arm_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu)
gen_goto_tb(dc, 1, curr_insn_len(dc));
}
}
+
+ emit_delayed_exceptions(dc);
}
static const TranslatorOps arm_translator_ops = {
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 47/76] target/arm: Emit HSTR trap exception out of line
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (45 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 46/76] target/arm: Introduce delay_exception{_el} Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 48/76] target/arm: Emit v7m LTPSIZE " Richard Henderson
` (29 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Use delay_exception_el to move the exception out of line.
Use TCG_COND_TSTNE instead of separate AND+NE.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/tcg/translate.c | 18 ++++--------------
1 file changed, 4 insertions(+), 14 deletions(-)
diff --git a/target/arm/tcg/translate.c b/target/arm/tcg/translate.c
index 78d26aac04..57bc8ea4c5 100644
--- a/target/arm/tcg/translate.c
+++ b/target/arm/tcg/translate.c
@@ -1774,21 +1774,11 @@ static void do_coproc_insn(DisasContext *s, int cpnum, int is64,
if (maskbit != 4 && maskbit != 14) {
/* T4 and T14 are RES0 so never cause traps */
- TCGv_i32 t;
- DisasLabel over = gen_disas_label(s);
+ TCGLabel *fail = delay_exception_el(s, EXCP_UDEF, syndrome, 2);
+ TCGv_i32 t =
+ load_cpu_offset(offsetoflow32(CPUARMState, cp15.hstr_el2));
- t = load_cpu_offset(offsetoflow32(CPUARMState, cp15.hstr_el2));
- tcg_gen_andi_i32(t, t, 1u << maskbit);
- tcg_gen_brcondi_i32(TCG_COND_EQ, t, 0, over.label);
-
- gen_exception_insn_el(s, 0, EXCP_UDEF, syndrome, 2);
- /*
- * gen_exception_insn() will set is_jmp to DISAS_NORETURN,
- * but since we're conditionally branching over it, we want
- * to assume continue-to-next-instruction.
- */
- s->base.is_jmp = DISAS_NEXT;
- set_disas_label(s, over);
+ tcg_gen_brcondi_i32(TCG_COND_TSTNE, t, 1u << maskbit, fail);
}
}
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 48/76] target/arm: Emit v7m LTPSIZE exception out of line
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (46 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 47/76] target/arm: Emit HSTR trap exception out of line Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 49/76] target/arm: Implement GCSSTR, GCSSTTR Richard Henderson
` (28 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/tcg/translate.c | 7 +++----
1 file changed, 3 insertions(+), 4 deletions(-)
diff --git a/target/arm/tcg/translate.c b/target/arm/tcg/translate.c
index 57bc8ea4c5..3df0bbcb7f 100644
--- a/target/arm/tcg/translate.c
+++ b/target/arm/tcg/translate.c
@@ -5598,11 +5598,10 @@ 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 */
- DisasLabel skipexc = gen_disas_label(s);
+ TCGLabel *fail = delay_exception(s, EXCP_INVSTATE, syn_uncategorized());
+
tmp = load_cpu_field(v7m.ltpsize);
- tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, 4, skipexc.label);
- gen_exception_insn(s, 0, EXCP_INVSTATE, syn_uncategorized());
- set_disas_label(s, skipexc);
+ tcg_gen_brcondi_i32(TCG_COND_NE, tmp, 4, fail);
}
if (a->f) {
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 49/76] target/arm: Implement GCSSTR, GCSSTTR
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (47 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 48/76] target/arm: Emit v7m LTPSIZE " Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 50/76] target/arm: Implement GCSB Richard Henderson
` (27 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Note that CreateAccDescGCS() does not enable tagchecked,
and Data Aborts from GCS instructions do not set iss.isv.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/tcg/translate-a64.c | 42 ++++++++++++++++++++++++++++++++++
target/arm/tcg/a64.decode | 3 +++
2 files changed, 45 insertions(+)
diff --git a/target/arm/tcg/translate-a64.c b/target/arm/tcg/translate-a64.c
index d607a0afd6..2849805f1c 100644
--- a/target/arm/tcg/translate-a64.c
+++ b/target/arm/tcg/translate-a64.c
@@ -139,6 +139,12 @@ static int core_a64_user_mem_index(DisasContext *s, bool unpriv)
return arm_to_core_mmu_idx(full_a64_user_mem_index(s, unpriv));
}
+/* For a given translation regime, return the core mmu_idx for gcs access. */
+static int core_gcs_mem_index(ARMMMUIdx armidx)
+{
+ return arm_to_core_mmu_idx(regime_to_gcs(armidx));
+}
+
static void set_btype_raw(int val)
{
tcg_gen_st_i32(tcg_constant_i32(val), tcg_env,
@@ -3989,6 +3995,42 @@ static bool trans_STLR_i(DisasContext *s, arg_ldapr_stlr_i *a)
return true;
}
+static bool trans_GCSSTR(DisasContext *s, arg_GCSSTR *a)
+{
+ ARMMMUIdx armidx;
+
+ if (!dc_isar_feature(aa64_gcs, s)) {
+ return false;
+ }
+
+ /*
+ * The pseudocode for GCSSTTR is
+ *
+ * effective_el = AArch64.IsUnprivAccessPriv() ? PSTATE.EL : EL0;
+ * if (effective_el == PSTATE.EL) CheckGCSSTREnabled();
+ *
+ * We have cached the result of IsUnprivAccessPriv in DisasContext,
+ * but since we need the result of full_a64_user_mem_index anyway,
+ * use the mmu_idx test as a proxy for the effective_el test.
+ */
+ armidx = full_a64_user_mem_index(s, a->unpriv);
+ if (armidx == s->mmu_idx && s->gcsstr_el != 0) {
+ gen_exception_insn_el(s, 0, EXCP_UDEF,
+ syn_gcs_gcsstr(a->rn, a->rt),
+ s->gcsstr_el);
+ return true;
+ }
+
+ if (a->rn == 31) {
+ gen_check_sp_alignment(s);
+ }
+ tcg_gen_qemu_st_i64(cpu_reg(s, a->rt),
+ clean_data_tbi(s, cpu_reg_sp(s, a->rn)),
+ core_gcs_mem_index(armidx),
+ finalize_memop(s, MO_64 | MO_ALIGN));
+ return true;
+}
+
static bool trans_LD_mult(DisasContext *s, arg_ldst_mult *a)
{
TCGv_i64 clean_addr, tcg_rn, tcg_ebytes;
diff --git a/target/arm/tcg/a64.decode b/target/arm/tcg/a64.decode
index 8283a9c83d..2ae73f443a 100644
--- a/target/arm/tcg/a64.decode
+++ b/target/arm/tcg/a64.decode
@@ -571,6 +571,9 @@ LDAPR_i 10 011001 10 0 ......... 00 ..... ..... @ldapr_stlr_i sign=1 ext
LDAPR_i 00 011001 11 0 ......... 00 ..... ..... @ldapr_stlr_i sign=1 ext=1 sz=0
LDAPR_i 01 011001 11 0 ......... 00 ..... ..... @ldapr_stlr_i sign=1 ext=1 sz=1
+# GCSSTR, GCSSTTR
+GCSSTR 11011001 000 11111 000 unpriv:1 11 rn:5 rt:5
+
# Load/store multiple structures
# The 4-bit opcode in [15:12] encodes repeat count and structure elements
&ldst_mult rm rn rt sz q p rpt selem
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 50/76] target/arm: Implement GCSB
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (48 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 49/76] target/arm: Implement GCSSTR, GCSSTTR Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:48 ` [PATCH v5 51/76] target/arm: Implement GCSPUSHM Richard Henderson
` (26 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/tcg/translate-a64.c | 8 ++++++++
target/arm/tcg/a64.decode | 1 +
2 files changed, 9 insertions(+)
diff --git a/target/arm/tcg/translate-a64.c b/target/arm/tcg/translate-a64.c
index 2849805f1c..7e45197d8a 100644
--- a/target/arm/tcg/translate-a64.c
+++ b/target/arm/tcg/translate-a64.c
@@ -2080,6 +2080,14 @@ static bool trans_ESB(DisasContext *s, arg_ESB *a)
return true;
}
+static bool trans_GCSB(DisasContext *s, arg_GCSB *a)
+{
+ if (dc_isar_feature(aa64_gcs, s)) {
+ tcg_gen_mb(TCG_BAR_SC | TCG_MO_ALL);
+ }
+ return true;
+}
+
static bool trans_PACIAZ(DisasContext *s, arg_PACIAZ *a)
{
if (s->pauth_active) {
diff --git a/target/arm/tcg/a64.decode b/target/arm/tcg/a64.decode
index 2ae73f443a..01b1b3e38b 100644
--- a/target/arm/tcg/a64.decode
+++ b/target/arm/tcg/a64.decode
@@ -248,6 +248,7 @@ ERETA 1101011 0100 11111 00001 m:1 11111 11111 &reta # ERETAA, ERETAB
AUTIA1716 1101 0101 0000 0011 0010 0001 100 11111
AUTIB1716 1101 0101 0000 0011 0010 0001 110 11111
ESB 1101 0101 0000 0011 0010 0010 000 11111
+ GCSB 1101 0101 0000 0011 0010 0010 011 11111
PACIAZ 1101 0101 0000 0011 0010 0011 000 11111
PACIASP 1101 0101 0000 0011 0010 0011 001 11111
PACIBZ 1101 0101 0000 0011 0010 0011 010 11111
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 51/76] target/arm: Implement GCSPUSHM
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (49 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 50/76] target/arm: Implement GCSB Richard Henderson
@ 2025-09-22 18:48 ` Richard Henderson
2025-09-22 18:49 ` [PATCH v5 52/76] target/arm: Implement GCSPOPM Richard Henderson
` (25 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:48 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/cpregs.h | 3 +++
target/arm/cpregs-gcs.c | 15 +++++++++++++++
target/arm/tcg/translate-a64.c | 29 +++++++++++++++++++++++++++++
3 files changed, 47 insertions(+)
diff --git a/target/arm/cpregs.h b/target/arm/cpregs.h
index a79f00351c..d22ae383cd 100644
--- a/target/arm/cpregs.h
+++ b/target/arm/cpregs.h
@@ -47,6 +47,8 @@ enum {
ARM_CP_DC_ZVA = 0x0005,
ARM_CP_DC_GVA = 0x0006,
ARM_CP_DC_GZVA = 0x0007,
+ /* Special: gcs instructions */
+ ARM_CP_GCSPUSHM = 0x0008,
/* Flag: reads produce resetvalue; writes ignored. */
ARM_CP_CONST = 1 << 4,
@@ -871,6 +873,7 @@ typedef enum FGTBit {
DO_BIT(HFGITR, DVPRCTX),
DO_BIT(HFGITR, CPPRCTX),
DO_BIT(HFGITR, DCCVAC),
+ DO_REV_BIT(HFGITR, NGCSPUSHM_EL1),
DO_BIT(HFGITR, ATS1E1A),
} FGTBit;
diff --git a/target/arm/cpregs-gcs.c b/target/arm/cpregs-gcs.c
index 1ff041811d..cc5a0b86e4 100644
--- a/target/arm/cpregs-gcs.c
+++ b/target/arm/cpregs-gcs.c
@@ -43,6 +43,16 @@ static void gcspr_write(CPUARMState *env, const ARMCPRegInfo *ri,
raw_write(env, ri, value & ~7);
}
+static CPAccessResult access_gcspushm(CPUARMState *env, const ARMCPRegInfo *ri,
+ bool isread)
+{
+ int el = arm_current_el(env);
+ if (!(env->cp15.gcscr_el[el] & GCSCR_PUSHMEN)) {
+ return CP_ACCESS_TRAP_BIT | (el ? el : 1);
+ }
+ return CP_ACCESS_OK;
+}
+
static const ARMCPRegInfo gcs_reginfo[] = {
{ .name = "GCSCRE0_EL1", .state = ARM_CP_STATE_AA64,
.opc0 = 3, .opc1 = 0, .crn = 2, .crm = 5, .opc2 = 2,
@@ -85,6 +95,11 @@ static const ARMCPRegInfo gcs_reginfo[] = {
.opc0 = 3, .opc1 = 6, .crn = 2, .crm = 5, .opc2 = 1,
.access = PL3_RW, .writefn = gcspr_write,
.fieldoffset = offsetof(CPUARMState, cp15.gcspr_el[2]) },
+
+ { .name = "GCSPUSHM", .state = ARM_CP_STATE_AA64,
+ .opc0 = 1, .opc1 = 3, .crn = 7, .crm = 7, .opc2 = 0,
+ .access = PL0_W, .accessfn = access_gcspushm,
+ .fgt = FGT_NGCSPUSHM_EL1, .type = ARM_CP_GCSPUSHM },
};
void define_gcs_cpregs(ARMCPU *cpu)
diff --git a/target/arm/tcg/translate-a64.c b/target/arm/tcg/translate-a64.c
index 7e45197d8a..e2af10c075 100644
--- a/target/arm/tcg/translate-a64.c
+++ b/target/arm/tcg/translate-a64.c
@@ -26,6 +26,7 @@
#include "cpregs.h"
static TCGv_i64 cpu_X[32];
+static TCGv_i64 cpu_gcspr[32];
static TCGv_i64 cpu_pc;
/* Load/store exclusive handling */
@@ -77,6 +78,10 @@ static int scale_by_log2_tag_granule(DisasContext *s, int x)
/* initialize TCG globals. */
void a64_translate_init(void)
{
+ static const char gcspr_names[4][12] = {
+ "gcspr_el0", "gcspr_el1", "gcspr_el2", "gcspr_el3"
+ };
+
int i;
cpu_pc = tcg_global_mem_new_i64(tcg_env,
@@ -90,6 +95,13 @@ void a64_translate_init(void)
cpu_exclusive_high = tcg_global_mem_new_i64(tcg_env,
offsetof(CPUARMState, exclusive_high), "exclusive_high");
+
+ for (i = 0; i < 4; i++) {
+ cpu_gcspr[i] =
+ tcg_global_mem_new_i64(tcg_env,
+ offsetof(CPUARMState, cp15.gcspr_el[i]),
+ gcspr_names[i]);
+ }
}
/*
@@ -420,6 +432,18 @@ static MemOp check_ordered_align(DisasContext *s, int rn, int imm,
return finalize_memop(s, mop);
}
+static void gen_add_gcs_record(DisasContext *s, TCGv_i64 value)
+{
+ TCGv_i64 addr = tcg_temp_new_i64();
+ TCGv_i64 gcspr = cpu_gcspr[s->current_el];
+ int mmuidx = core_gcs_mem_index(s->mmu_idx);
+ MemOp mop = finalize_memop(s, MO_64 | MO_ALIGN);
+
+ tcg_gen_addi_i64(addr, gcspr, -8);
+ tcg_gen_qemu_st_i64(value, clean_data_tbi(s, addr), mmuidx, mop);
+ tcg_gen_mov_i64(gcspr, addr);
+}
+
typedef struct DisasCompare64 {
TCGCond cond;
TCGv_i64 value;
@@ -2803,6 +2827,11 @@ static void handle_sys(DisasContext *s, bool isread,
}
}
return;
+ case ARM_CP_GCSPUSHM:
+ if (s->gcs_en) {
+ gen_add_gcs_record(s, cpu_reg(s, rt));
+ }
+ return;
default:
g_assert_not_reached();
}
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 52/76] target/arm: Implement GCSPOPM
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (50 preceding siblings ...)
2025-09-22 18:48 ` [PATCH v5 51/76] target/arm: Implement GCSPUSHM Richard Henderson
@ 2025-09-22 18:49 ` Richard Henderson
2025-09-22 18:49 ` [PATCH v5 53/76] target/arm: Implement GCSPUSHX Richard Henderson
` (24 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:49 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/cpregs.h | 1 +
target/arm/cpregs-gcs.c | 3 +++
target/arm/tcg/translate-a64.c | 24 ++++++++++++++++++++++++
3 files changed, 28 insertions(+)
diff --git a/target/arm/cpregs.h b/target/arm/cpregs.h
index d22ae383cd..81d8f0e32b 100644
--- a/target/arm/cpregs.h
+++ b/target/arm/cpregs.h
@@ -49,6 +49,7 @@ enum {
ARM_CP_DC_GZVA = 0x0007,
/* Special: gcs instructions */
ARM_CP_GCSPUSHM = 0x0008,
+ ARM_CP_GCSPOPM = 0x0009,
/* Flag: reads produce resetvalue; writes ignored. */
ARM_CP_CONST = 1 << 4,
diff --git a/target/arm/cpregs-gcs.c b/target/arm/cpregs-gcs.c
index cc5a0b86e4..15d383b2a4 100644
--- a/target/arm/cpregs-gcs.c
+++ b/target/arm/cpregs-gcs.c
@@ -100,6 +100,9 @@ static const ARMCPRegInfo gcs_reginfo[] = {
.opc0 = 1, .opc1 = 3, .crn = 7, .crm = 7, .opc2 = 0,
.access = PL0_W, .accessfn = access_gcspushm,
.fgt = FGT_NGCSPUSHM_EL1, .type = ARM_CP_GCSPUSHM },
+ { .name = "GCSPOPM", .state = ARM_CP_STATE_AA64,
+ .opc0 = 1, .opc1 = 3, .crn = 7, .crm = 7, .opc2 = 1,
+ .access = PL0_R, .type = ARM_CP_GCSPOPM },
};
void define_gcs_cpregs(ARMCPU *cpu)
diff --git a/target/arm/tcg/translate-a64.c b/target/arm/tcg/translate-a64.c
index e2af10c075..c52b02b003 100644
--- a/target/arm/tcg/translate-a64.c
+++ b/target/arm/tcg/translate-a64.c
@@ -2521,6 +2521,24 @@ static void gen_sysreg_undef(DisasContext *s, bool isread,
gen_exception_insn(s, 0, EXCP_UDEF, syndrome);
}
+static void gen_gcspopm(DisasContext *s, int rt)
+{
+ TCGv_i64 gcspr = cpu_gcspr[s->current_el];
+ int mmuidx = core_gcs_mem_index(s->mmu_idx);
+ MemOp mop = finalize_memop(s, MO_64 | MO_ALIGN);
+ TCGv_i64 value = tcg_temp_new_i64();
+ TCGLabel *fail_label =
+ delay_exception(s, EXCP_UDEF, syn_gcs_data_check(GCS_IT_GCSPOPM, rt));
+
+ /* The value at top-of-stack must have low 2 bits clear. */
+ tcg_gen_qemu_ld_i64(value, clean_data_tbi(s, gcspr), mmuidx, mop);
+ tcg_gen_brcondi_i64(TCG_COND_TSTNE, value, 3, fail_label);
+
+ /* Complete the pop and return the value. */
+ tcg_gen_addi_i64(gcspr, gcspr, 8);
+ tcg_gen_mov_i64(cpu_reg(s, rt), value);
+}
+
/*
* Look up @key, returning the cpreg, which must exist.
* Additionally, the new cpreg must also be accessible.
@@ -2832,6 +2850,12 @@ static void handle_sys(DisasContext *s, bool isread,
gen_add_gcs_record(s, cpu_reg(s, rt));
}
return;
+ case ARM_CP_GCSPOPM:
+ /* Note that X[rt] is unchanged if !GCSEnabled. */
+ if (s->gcs_en) {
+ gen_gcspopm(s, rt);
+ }
+ return;
default:
g_assert_not_reached();
}
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 53/76] target/arm: Implement GCSPUSHX
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (51 preceding siblings ...)
2025-09-22 18:49 ` [PATCH v5 52/76] target/arm: Implement GCSPOPM Richard Henderson
@ 2025-09-22 18:49 ` Richard Henderson
2025-09-22 18:49 ` [PATCH v5 54/76] target/arm: Implement GCSPOPX Richard Henderson
` (23 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:49 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/cpregs.h | 2 ++
target/arm/cpregs-gcs.c | 17 ++++++++++++++++
target/arm/tcg/translate-a64.c | 37 ++++++++++++++++++++++++++++++++++
3 files changed, 56 insertions(+)
diff --git a/target/arm/cpregs.h b/target/arm/cpregs.h
index 81d8f0e32b..909916b7fd 100644
--- a/target/arm/cpregs.h
+++ b/target/arm/cpregs.h
@@ -50,6 +50,7 @@ enum {
/* Special: gcs instructions */
ARM_CP_GCSPUSHM = 0x0008,
ARM_CP_GCSPOPM = 0x0009,
+ ARM_CP_GCSPUSHX = 0x000a,
/* Flag: reads produce resetvalue; writes ignored. */
ARM_CP_CONST = 1 << 4,
@@ -875,6 +876,7 @@ typedef enum FGTBit {
DO_BIT(HFGITR, CPPRCTX),
DO_BIT(HFGITR, DCCVAC),
DO_REV_BIT(HFGITR, NGCSPUSHM_EL1),
+ DO_REV_BIT(HFGITR, NGCSEPP),
DO_BIT(HFGITR, ATS1E1A),
} FGTBit;
diff --git a/target/arm/cpregs-gcs.c b/target/arm/cpregs-gcs.c
index 15d383b2a4..e6c7025d02 100644
--- a/target/arm/cpregs-gcs.c
+++ b/target/arm/cpregs-gcs.c
@@ -53,6 +53,19 @@ static CPAccessResult access_gcspushm(CPUARMState *env, const ARMCPRegInfo *ri,
return CP_ACCESS_OK;
}
+static CPAccessResult access_gcspushx(CPUARMState *env, const ARMCPRegInfo *ri,
+ bool isread)
+{
+ /* Trap if lock taken, and enabled. */
+ if (!(env->pstate & PSTATE_EXLOCK)) {
+ int el = arm_current_el(env);
+ if (env->cp15.gcscr_el[el] & GCSCR_EXLOCKEN) {
+ return CP_ACCESS_EXLOCK;
+ }
+ }
+ return CP_ACCESS_OK;
+}
+
static const ARMCPRegInfo gcs_reginfo[] = {
{ .name = "GCSCRE0_EL1", .state = ARM_CP_STATE_AA64,
.opc0 = 3, .opc1 = 0, .crn = 2, .crm = 5, .opc2 = 2,
@@ -103,6 +116,10 @@ static const ARMCPRegInfo gcs_reginfo[] = {
{ .name = "GCSPOPM", .state = ARM_CP_STATE_AA64,
.opc0 = 1, .opc1 = 3, .crn = 7, .crm = 7, .opc2 = 1,
.access = PL0_R, .type = ARM_CP_GCSPOPM },
+ { .name = "GCSPUSHX", .state = ARM_CP_STATE_AA64,
+ .opc0 = 1, .opc1 = 0, .crn = 7, .crm = 7, .opc2 = 4,
+ .access = PL1_W, .accessfn = access_gcspushx, .fgt = FGT_NGCSEPP,
+ .type = ARM_CP_GCSPUSHX },
};
void define_gcs_cpregs(ARMCPU *cpu)
diff --git a/target/arm/tcg/translate-a64.c b/target/arm/tcg/translate-a64.c
index c52b02b003..6fe6a77b34 100644
--- a/target/arm/tcg/translate-a64.c
+++ b/target/arm/tcg/translate-a64.c
@@ -2539,6 +2539,35 @@ static void gen_gcspopm(DisasContext *s, int rt)
tcg_gen_mov_i64(cpu_reg(s, rt), value);
}
+static void gen_gcspushx(DisasContext *s)
+{
+ TCGv_i64 gcspr = cpu_gcspr[s->current_el];
+ int spsr_idx = aarch64_banked_spsr_index(s->current_el);
+ int spsr_off = offsetof(CPUARMState, banked_spsr[spsr_idx]);
+ int elr_off = offsetof(CPUARMState, elr_el[s->current_el]);
+ int mmuidx = core_gcs_mem_index(s->mmu_idx);
+ MemOp mop = finalize_memop(s, MO_64 | MO_ALIGN);
+ TCGv_i64 addr = tcg_temp_new_i64();
+ TCGv_i64 tmp = tcg_temp_new_i64();
+
+ tcg_gen_addi_i64(addr, gcspr, -8);
+ tcg_gen_qemu_st_i64(cpu_reg(s, 30), addr, mmuidx, mop);
+
+ tcg_gen_ld_i64(tmp, tcg_env, spsr_off);
+ tcg_gen_addi_i64(addr, addr, -8);
+ tcg_gen_qemu_st_i64(tmp, addr, mmuidx, mop);
+
+ tcg_gen_ld_i64(tmp, tcg_env, elr_off);
+ tcg_gen_addi_i64(addr, addr, -8);
+ tcg_gen_qemu_st_i64(tmp, addr, mmuidx, mop);
+
+ tcg_gen_addi_i64(addr, addr, -8);
+ tcg_gen_qemu_st_i64(tcg_constant_i64(0b1001), addr, mmuidx, mop);
+
+ tcg_gen_mov_i64(gcspr, addr);
+ clear_pstate_bits(PSTATE_EXLOCK);
+}
+
/*
* Look up @key, returning the cpreg, which must exist.
* Additionally, the new cpreg must also be accessible.
@@ -2856,6 +2885,14 @@ static void handle_sys(DisasContext *s, bool isread,
gen_gcspopm(s, rt);
}
return;
+ case ARM_CP_GCSPUSHX:
+ /* Choose the CONSTRAINED UNPREDICTABLE for UNDEF. */
+ if (rt != 31) {
+ unallocated_encoding(s);
+ } else if (s->gcs_en) {
+ gen_gcspushx(s);
+ }
+ return;
default:
g_assert_not_reached();
}
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 54/76] target/arm: Implement GCSPOPX
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (52 preceding siblings ...)
2025-09-22 18:49 ` [PATCH v5 53/76] target/arm: Implement GCSPUSHX Richard Henderson
@ 2025-09-22 18:49 ` Richard Henderson
2025-09-22 18:49 ` [PATCH v5 55/76] target/arm: Implement GCSPOPCX Richard Henderson
` (22 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:49 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/cpregs.h | 1 +
target/arm/cpregs-gcs.c | 3 +++
target/arm/tcg/translate-a64.c | 35 ++++++++++++++++++++++++++++++++++
3 files changed, 39 insertions(+)
diff --git a/target/arm/cpregs.h b/target/arm/cpregs.h
index 909916b7fd..ccf45fd136 100644
--- a/target/arm/cpregs.h
+++ b/target/arm/cpregs.h
@@ -51,6 +51,7 @@ enum {
ARM_CP_GCSPUSHM = 0x0008,
ARM_CP_GCSPOPM = 0x0009,
ARM_CP_GCSPUSHX = 0x000a,
+ ARM_CP_GCSPOPX = 0x000b,
/* Flag: reads produce resetvalue; writes ignored. */
ARM_CP_CONST = 1 << 4,
diff --git a/target/arm/cpregs-gcs.c b/target/arm/cpregs-gcs.c
index e6c7025d02..5b5b895a09 100644
--- a/target/arm/cpregs-gcs.c
+++ b/target/arm/cpregs-gcs.c
@@ -120,6 +120,9 @@ static const ARMCPRegInfo gcs_reginfo[] = {
.opc0 = 1, .opc1 = 0, .crn = 7, .crm = 7, .opc2 = 4,
.access = PL1_W, .accessfn = access_gcspushx, .fgt = FGT_NGCSEPP,
.type = ARM_CP_GCSPUSHX },
+ { .name = "GCSPOPX", .state = ARM_CP_STATE_AA64,
+ .opc0 = 1, .opc1 = 0, .crn = 7, .crm = 7, .opc2 = 6,
+ .access = PL1_W, .type = ARM_CP_GCSPOPX },
};
void define_gcs_cpregs(ARMCPU *cpu)
diff --git a/target/arm/tcg/translate-a64.c b/target/arm/tcg/translate-a64.c
index 6fe6a77b34..c1852e2c5d 100644
--- a/target/arm/tcg/translate-a64.c
+++ b/target/arm/tcg/translate-a64.c
@@ -2568,6 +2568,33 @@ static void gen_gcspushx(DisasContext *s)
clear_pstate_bits(PSTATE_EXLOCK);
}
+static void gen_gcspopx(DisasContext *s)
+{
+ TCGv_i64 gcspr = cpu_gcspr[s->current_el];
+ int mmuidx = core_gcs_mem_index(s->mmu_idx);
+ MemOp mop = finalize_memop(s, MO_64 | MO_ALIGN);
+ TCGv_i64 addr = tcg_temp_new_i64();
+ TCGv_i64 tmp = tcg_temp_new_i64();
+ TCGLabel *fail_label =
+ delay_exception(s, EXCP_UDEF, syn_gcs_data_check(GCS_IT_GCSPOPX, 31));
+
+ /* The value at top-of-stack must be an exception token. */
+ tcg_gen_qemu_ld_i64(tmp, gcspr, mmuidx, mop);
+ tcg_gen_brcondi_i64(TCG_COND_NE, tmp, 0b1001, fail_label);
+
+ /*
+ * The other three values in the exception return record
+ * are ignored, but are loaded anyway to raise faults.
+ */
+ tcg_gen_addi_i64(addr, gcspr, 8);
+ tcg_gen_qemu_ld_i64(tmp, addr, mmuidx, mop);
+ tcg_gen_addi_i64(addr, addr, 8);
+ tcg_gen_qemu_ld_i64(tmp, addr, mmuidx, mop);
+ tcg_gen_addi_i64(addr, addr, 8);
+ tcg_gen_qemu_ld_i64(tmp, addr, mmuidx, mop);
+ tcg_gen_addi_i64(gcspr, addr, 8);
+}
+
/*
* Look up @key, returning the cpreg, which must exist.
* Additionally, the new cpreg must also be accessible.
@@ -2893,6 +2920,14 @@ static void handle_sys(DisasContext *s, bool isread,
gen_gcspushx(s);
}
return;
+ case ARM_CP_GCSPOPX:
+ /* Choose the CONSTRAINED UNPREDICTABLE for UNDEF. */
+ if (rt != 31) {
+ unallocated_encoding(s);
+ } else if (s->gcs_en) {
+ gen_gcspopx(s);
+ }
+ return;
default:
g_assert_not_reached();
}
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 55/76] target/arm: Implement GCSPOPCX
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (53 preceding siblings ...)
2025-09-22 18:49 ` [PATCH v5 54/76] target/arm: Implement GCSPOPX Richard Henderson
@ 2025-09-22 18:49 ` Richard Henderson
2025-09-22 18:49 ` [PATCH v5 56/76] target/arm: Implement GCSSS1 Richard Henderson
` (21 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:49 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/cpregs.h | 1 +
target/arm/cpregs-gcs.c | 17 +++++++++++
target/arm/tcg/translate-a64.c | 56 ++++++++++++++++++++++++++++++++++
3 files changed, 74 insertions(+)
diff --git a/target/arm/cpregs.h b/target/arm/cpregs.h
index ccf45fd136..6d9145109f 100644
--- a/target/arm/cpregs.h
+++ b/target/arm/cpregs.h
@@ -52,6 +52,7 @@ enum {
ARM_CP_GCSPOPM = 0x0009,
ARM_CP_GCSPUSHX = 0x000a,
ARM_CP_GCSPOPX = 0x000b,
+ ARM_CP_GCSPOPCX = 0x000c,
/* Flag: reads produce resetvalue; writes ignored. */
ARM_CP_CONST = 1 << 4,
diff --git a/target/arm/cpregs-gcs.c b/target/arm/cpregs-gcs.c
index 5b5b895a09..3795bf7f36 100644
--- a/target/arm/cpregs-gcs.c
+++ b/target/arm/cpregs-gcs.c
@@ -66,6 +66,19 @@ static CPAccessResult access_gcspushx(CPUARMState *env, const ARMCPRegInfo *ri,
return CP_ACCESS_OK;
}
+static CPAccessResult access_gcspopcx(CPUARMState *env, const ARMCPRegInfo *ri,
+ bool isread)
+{
+ /* Trap if lock not taken, and enabled. */
+ if (env->pstate & PSTATE_EXLOCK) {
+ int el = arm_current_el(env);
+ if (env->cp15.gcscr_el[el] & GCSCR_EXLOCKEN) {
+ return CP_ACCESS_EXLOCK;
+ }
+ }
+ return CP_ACCESS_OK;
+}
+
static const ARMCPRegInfo gcs_reginfo[] = {
{ .name = "GCSCRE0_EL1", .state = ARM_CP_STATE_AA64,
.opc0 = 3, .opc1 = 0, .crn = 2, .crm = 5, .opc2 = 2,
@@ -120,6 +133,10 @@ static const ARMCPRegInfo gcs_reginfo[] = {
.opc0 = 1, .opc1 = 0, .crn = 7, .crm = 7, .opc2 = 4,
.access = PL1_W, .accessfn = access_gcspushx, .fgt = FGT_NGCSEPP,
.type = ARM_CP_GCSPUSHX },
+ { .name = "GCSPOPCX", .state = ARM_CP_STATE_AA64,
+ .opc0 = 1, .opc1 = 0, .crn = 7, .crm = 7, .opc2 = 5,
+ .access = PL1_W, .accessfn = access_gcspopcx, .fgt = FGT_NGCSEPP,
+ .type = ARM_CP_GCSPOPCX },
{ .name = "GCSPOPX", .state = ARM_CP_STATE_AA64,
.opc0 = 1, .opc1 = 0, .crn = 7, .crm = 7, .opc2 = 6,
.access = PL1_W, .type = ARM_CP_GCSPOPX },
diff --git a/target/arm/tcg/translate-a64.c b/target/arm/tcg/translate-a64.c
index c1852e2c5d..c601ff790b 100644
--- a/target/arm/tcg/translate-a64.c
+++ b/target/arm/tcg/translate-a64.c
@@ -2568,6 +2568,54 @@ static void gen_gcspushx(DisasContext *s)
clear_pstate_bits(PSTATE_EXLOCK);
}
+static void gen_gcspopcx(DisasContext *s)
+{
+ TCGv_i64 gcspr = cpu_gcspr[s->current_el];
+ int spsr_idx = aarch64_banked_spsr_index(s->current_el);
+ int spsr_off = offsetof(CPUARMState, banked_spsr[spsr_idx]);
+ int elr_off = offsetof(CPUARMState, elr_el[s->current_el]);
+ int gcscr_off = offsetof(CPUARMState, cp15.gcscr_el[s->current_el]);
+ int pstate_off = offsetof(CPUARMState, pstate);
+ int mmuidx = core_gcs_mem_index(s->mmu_idx);
+ MemOp mop = finalize_memop(s, MO_64 | MO_ALIGN);
+ TCGv_i64 addr = tcg_temp_new_i64();
+ TCGv_i64 tmp1 = tcg_temp_new_i64();
+ TCGv_i64 tmp2 = tcg_temp_new_i64();
+ TCGLabel *fail_label =
+ delay_exception(s, EXCP_UDEF, syn_gcs_data_check(GCS_IT_GCSPOPCX, 31));
+
+ /* The value at top-of-stack must be an exception token. */
+ tcg_gen_qemu_ld_i64(tmp1, gcspr, mmuidx, mop);
+ tcg_gen_brcondi_i64(TCG_COND_NE, tmp1, 0b1001, fail_label);
+
+ /* Validate in turn, ELR ... */
+ tcg_gen_addi_i64(addr, gcspr, 8);
+ tcg_gen_qemu_ld_i64(tmp1, addr, mmuidx, mop);
+ tcg_gen_ld_i64(tmp2, tcg_env, elr_off);
+ tcg_gen_brcond_i64(TCG_COND_NE, tmp1, tmp2, fail_label);
+
+ /* ... SPSR ... */
+ tcg_gen_addi_i64(addr, addr, 8);
+ tcg_gen_qemu_ld_i64(tmp1, addr, mmuidx, mop);
+ tcg_gen_ld_i64(tmp2, tcg_env, spsr_off);
+ tcg_gen_brcond_i64(TCG_COND_NE, tmp1, tmp2, fail_label);
+
+ /* ... and LR. */
+ tcg_gen_addi_i64(addr, addr, 8);
+ tcg_gen_qemu_ld_i64(tmp1, addr, mmuidx, mop);
+ tcg_gen_brcond_i64(TCG_COND_NE, tmp1, cpu_reg(s, 30), fail_label);
+
+ /* Writeback stack pointer after pop. */
+ tcg_gen_addi_i64(gcspr, addr, 8);
+
+ /* PSTATE.EXLOCK = GetCurrentEXLOCKEN(). */
+ tcg_gen_ld_i64(tmp1, tcg_env, gcscr_off);
+ tcg_gen_ld_i64(tmp2, tcg_env, pstate_off);
+ tcg_gen_shri_i64(tmp1, tmp1, ctz64(GCSCR_EXLOCKEN));
+ tcg_gen_deposit_i64(tmp2, tmp2, tmp1, ctz64(PSTATE_EXLOCK), 1);
+ tcg_gen_st_i64(tmp2, tcg_env, pstate_off);
+}
+
static void gen_gcspopx(DisasContext *s)
{
TCGv_i64 gcspr = cpu_gcspr[s->current_el];
@@ -2920,6 +2968,14 @@ static void handle_sys(DisasContext *s, bool isread,
gen_gcspushx(s);
}
return;
+ case ARM_CP_GCSPOPCX:
+ /* Choose the CONSTRAINED UNPREDICTABLE for UNDEF. */
+ if (rt != 31) {
+ unallocated_encoding(s);
+ } else if (s->gcs_en) {
+ gen_gcspopcx(s);
+ }
+ return;
case ARM_CP_GCSPOPX:
/* Choose the CONSTRAINED UNPREDICTABLE for UNDEF. */
if (rt != 31) {
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 56/76] target/arm: Implement GCSSS1
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (54 preceding siblings ...)
2025-09-22 18:49 ` [PATCH v5 55/76] target/arm: Implement GCSPOPCX Richard Henderson
@ 2025-09-22 18:49 ` Richard Henderson
2025-09-22 18:49 ` [PATCH v5 57/76] target/arm: Implement GCSSS2 Richard Henderson
` (20 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:49 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/cpregs.h | 1 +
target/arm/cpregs-gcs.c | 3 +++
target/arm/tcg/translate-a64.c | 30 ++++++++++++++++++++++++++++++
3 files changed, 34 insertions(+)
diff --git a/target/arm/cpregs.h b/target/arm/cpregs.h
index 6d9145109f..e7e7050880 100644
--- a/target/arm/cpregs.h
+++ b/target/arm/cpregs.h
@@ -53,6 +53,7 @@ enum {
ARM_CP_GCSPUSHX = 0x000a,
ARM_CP_GCSPOPX = 0x000b,
ARM_CP_GCSPOPCX = 0x000c,
+ ARM_CP_GCSSS1 = 0x000d,
/* Flag: reads produce resetvalue; writes ignored. */
ARM_CP_CONST = 1 << 4,
diff --git a/target/arm/cpregs-gcs.c b/target/arm/cpregs-gcs.c
index 3795bf7f36..eda5c498d7 100644
--- a/target/arm/cpregs-gcs.c
+++ b/target/arm/cpregs-gcs.c
@@ -129,6 +129,9 @@ static const ARMCPRegInfo gcs_reginfo[] = {
{ .name = "GCSPOPM", .state = ARM_CP_STATE_AA64,
.opc0 = 1, .opc1 = 3, .crn = 7, .crm = 7, .opc2 = 1,
.access = PL0_R, .type = ARM_CP_GCSPOPM },
+ { .name = "GCSSS1", .state = ARM_CP_STATE_AA64,
+ .opc0 = 1, .opc1 = 3, .crn = 7, .crm = 7, .opc2 = 2,
+ .access = PL0_W, .type = ARM_CP_GCSSS1 },
{ .name = "GCSPUSHX", .state = ARM_CP_STATE_AA64,
.opc0 = 1, .opc1 = 0, .crn = 7, .crm = 7, .opc2 = 4,
.access = PL1_W, .accessfn = access_gcspushx, .fgt = FGT_NGCSEPP,
diff --git a/target/arm/tcg/translate-a64.c b/target/arm/tcg/translate-a64.c
index c601ff790b..54e7546c8f 100644
--- a/target/arm/tcg/translate-a64.c
+++ b/target/arm/tcg/translate-a64.c
@@ -2643,6 +2643,31 @@ static void gen_gcspopx(DisasContext *s)
tcg_gen_addi_i64(gcspr, addr, 8);
}
+static void gen_gcsss1(DisasContext *s, int rt)
+{
+ TCGv_i64 gcspr = cpu_gcspr[s->current_el];
+ int mmuidx = core_gcs_mem_index(s->mmu_idx);
+ MemOp mop = finalize_memop(s, MO_64 | MO_ALIGN);
+ TCGv_i64 inptr = cpu_reg(s, rt);
+ TCGv_i64 cmp = tcg_temp_new_i64();
+ TCGv_i64 new = tcg_temp_new_i64();
+ TCGv_i64 old = tcg_temp_new_i64();
+ TCGLabel *fail_label =
+ delay_exception(s, EXCP_UDEF, syn_gcs_data_check(GCS_IT_GCSSS1, rt));
+
+ /* Compute the valid cap entry that the new stack must have. */
+ tcg_gen_deposit_i64(cmp, inptr, tcg_constant_i64(1), 0, 12);
+ /* Compute the in-progress cap entry for the old stack. */
+ tcg_gen_deposit_i64(new, gcspr, tcg_constant_i64(5), 0, 3);
+
+ /* Swap the valid cap the with the in-progress cap. */
+ tcg_gen_atomic_cmpxchg_i64(old, inptr, cmp, new, mmuidx, mop);
+ tcg_gen_brcond_i64(TCG_COND_NE, old, cmp, fail_label);
+
+ /* The new stack had a valid cap: change gcspr. */
+ tcg_gen_andi_i64(gcspr, inptr, ~7);
+}
+
/*
* Look up @key, returning the cpreg, which must exist.
* Additionally, the new cpreg must also be accessible.
@@ -2984,6 +3009,11 @@ static void handle_sys(DisasContext *s, bool isread,
gen_gcspopx(s);
}
return;
+ case ARM_CP_GCSSS1:
+ if (s->gcs_en) {
+ gen_gcsss1(s, rt);
+ }
+ return;
default:
g_assert_not_reached();
}
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 57/76] target/arm: Implement GCSSS2
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (55 preceding siblings ...)
2025-09-22 18:49 ` [PATCH v5 56/76] target/arm: Implement GCSSS1 Richard Henderson
@ 2025-09-22 18:49 ` Richard Henderson
2025-09-22 18:49 ` [PATCH v5 58/76] target/arm: Add gcs record for BL Richard Henderson
` (19 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:49 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/cpregs.h | 1 +
target/arm/cpregs-gcs.c | 3 +++
target/arm/tcg/translate-a64.c | 34 ++++++++++++++++++++++++++++++++++
3 files changed, 38 insertions(+)
diff --git a/target/arm/cpregs.h b/target/arm/cpregs.h
index e7e7050880..732c07506d 100644
--- a/target/arm/cpregs.h
+++ b/target/arm/cpregs.h
@@ -54,6 +54,7 @@ enum {
ARM_CP_GCSPOPX = 0x000b,
ARM_CP_GCSPOPCX = 0x000c,
ARM_CP_GCSSS1 = 0x000d,
+ ARM_CP_GCSSS2 = 0x000e,
/* Flag: reads produce resetvalue; writes ignored. */
ARM_CP_CONST = 1 << 4,
diff --git a/target/arm/cpregs-gcs.c b/target/arm/cpregs-gcs.c
index eda5c498d7..1ed52a211a 100644
--- a/target/arm/cpregs-gcs.c
+++ b/target/arm/cpregs-gcs.c
@@ -132,6 +132,9 @@ static const ARMCPRegInfo gcs_reginfo[] = {
{ .name = "GCSSS1", .state = ARM_CP_STATE_AA64,
.opc0 = 1, .opc1 = 3, .crn = 7, .crm = 7, .opc2 = 2,
.access = PL0_W, .type = ARM_CP_GCSSS1 },
+ { .name = "GCSSS2", .state = ARM_CP_STATE_AA64,
+ .opc0 = 1, .opc1 = 3, .crn = 7, .crm = 7, .opc2 = 3,
+ .access = PL0_R, .type = ARM_CP_GCSSS2 },
{ .name = "GCSPUSHX", .state = ARM_CP_STATE_AA64,
.opc0 = 1, .opc1 = 0, .crn = 7, .crm = 7, .opc2 = 4,
.access = PL1_W, .accessfn = access_gcspushx, .fgt = FGT_NGCSEPP,
diff --git a/target/arm/tcg/translate-a64.c b/target/arm/tcg/translate-a64.c
index 54e7546c8f..dfa2a6d64c 100644
--- a/target/arm/tcg/translate-a64.c
+++ b/target/arm/tcg/translate-a64.c
@@ -2668,6 +2668,35 @@ static void gen_gcsss1(DisasContext *s, int rt)
tcg_gen_andi_i64(gcspr, inptr, ~7);
}
+static void gen_gcsss2(DisasContext *s, int rt)
+{
+ TCGv_i64 gcspr = cpu_gcspr[s->current_el];
+ int mmuidx = core_gcs_mem_index(s->mmu_idx);
+ MemOp mop = finalize_memop(s, MO_64 | MO_ALIGN);
+ TCGv_i64 outptr = tcg_temp_new_i64();
+ TCGv_i64 tmp = tcg_temp_new_i64();
+ TCGLabel *fail_label =
+ delay_exception(s, EXCP_UDEF, syn_gcs_data_check(GCS_IT_GCSSS2, rt));
+
+ /* Validate that the new stack has an in-progress cap. */
+ tcg_gen_qemu_ld_i64(outptr, gcspr, mmuidx, mop);
+ tcg_gen_andi_i64(tmp, outptr, 7);
+ tcg_gen_brcondi_i64(TCG_COND_NE, tmp, 5, fail_label);
+
+ /* Push a valid cap to the old stack. */
+ tcg_gen_andi_i64(outptr, outptr, ~7);
+ tcg_gen_addi_i64(outptr, outptr, -8);
+ tcg_gen_deposit_i64(tmp, outptr, tcg_constant_i64(1), 0, 12);
+ tcg_gen_qemu_st_i64(tmp, outptr, mmuidx, mop);
+ tcg_gen_mb(TCG_BAR_SC | TCG_MO_ALL);
+
+ /* Pop the in-progress cap from the new stack. */
+ tcg_gen_addi_i64(gcspr, gcspr, 8);
+
+ /* Return a pointer to the old stack cap. */
+ tcg_gen_mov_i64(cpu_reg(s, rt), outptr);
+}
+
/*
* Look up @key, returning the cpreg, which must exist.
* Additionally, the new cpreg must also be accessible.
@@ -3014,6 +3043,11 @@ static void handle_sys(DisasContext *s, bool isread,
gen_gcsss1(s, rt);
}
return;
+ case ARM_CP_GCSSS2:
+ if (s->gcs_en) {
+ gen_gcsss2(s, rt);
+ }
+ return;
default:
g_assert_not_reached();
}
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 58/76] target/arm: Add gcs record for BL
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (56 preceding siblings ...)
2025-09-22 18:49 ` [PATCH v5 57/76] target/arm: Implement GCSSS2 Richard Henderson
@ 2025-09-22 18:49 ` Richard Henderson
2025-09-22 18:49 ` [PATCH v5 59/76] target/arm: Add gcs record for BLR Richard Henderson
` (18 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:49 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/tcg/translate-a64.c | 9 ++++++++-
1 file changed, 8 insertions(+), 1 deletion(-)
diff --git a/target/arm/tcg/translate-a64.c b/target/arm/tcg/translate-a64.c
index dfa2a6d64c..6eca241f12 100644
--- a/target/arm/tcg/translate-a64.c
+++ b/target/arm/tcg/translate-a64.c
@@ -1678,7 +1678,14 @@ static bool trans_B(DisasContext *s, arg_i *a)
static bool trans_BL(DisasContext *s, arg_i *a)
{
- gen_pc_plus_diff(s, cpu_reg(s, 30), curr_insn_len(s));
+ TCGv_i64 link = tcg_temp_new_i64();
+
+ gen_pc_plus_diff(s, link, 4);
+ if (s->gcs_en) {
+ gen_add_gcs_record(s, link);
+ }
+ tcg_gen_mov_i64(cpu_reg(s, 30), link);
+
reset_btype(s);
gen_goto_tb(s, 0, a->imm);
return true;
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 59/76] target/arm: Add gcs record for BLR
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (57 preceding siblings ...)
2025-09-22 18:49 ` [PATCH v5 58/76] target/arm: Add gcs record for BL Richard Henderson
@ 2025-09-22 18:49 ` Richard Henderson
2025-09-22 18:49 ` [PATCH v5 60/76] target/arm: Add gcs record for BLR with PAuth Richard Henderson
` (17 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:49 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/tcg/translate-a64.c | 16 ++++++++--------
1 file changed, 8 insertions(+), 8 deletions(-)
diff --git a/target/arm/tcg/translate-a64.c b/target/arm/tcg/translate-a64.c
index 6eca241f12..79fc7b8c3d 100644
--- a/target/arm/tcg/translate-a64.c
+++ b/target/arm/tcg/translate-a64.c
@@ -1782,15 +1782,15 @@ static bool trans_BR(DisasContext *s, arg_r *a)
static bool trans_BLR(DisasContext *s, arg_r *a)
{
- TCGv_i64 dst = cpu_reg(s, a->rn);
- TCGv_i64 lr = cpu_reg(s, 30);
- if (dst == lr) {
- TCGv_i64 tmp = tcg_temp_new_i64();
- tcg_gen_mov_i64(tmp, dst);
- dst = tmp;
+ TCGv_i64 link = tcg_temp_new_i64();
+
+ gen_pc_plus_diff(s, link, 4);
+ if (s->gcs_en) {
+ gen_add_gcs_record(s, link);
}
- gen_pc_plus_diff(s, lr, curr_insn_len(s));
- gen_a64_set_pc(s, dst);
+ gen_a64_set_pc(s, cpu_reg(s, a->rn));
+ tcg_gen_mov_i64(cpu_reg(s, 30), link);
+
set_btype_for_blr(s);
s->base.is_jmp = DISAS_JUMP;
return true;
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 60/76] target/arm: Add gcs record for BLR with PAuth
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (58 preceding siblings ...)
2025-09-22 18:49 ` [PATCH v5 59/76] target/arm: Add gcs record for BLR Richard Henderson
@ 2025-09-22 18:49 ` Richard Henderson
2025-09-22 18:49 ` [PATCH v5 61/76] target/arm: Load gcs record for RET Richard Henderson
` (16 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:49 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/tcg/translate-a64.c | 31 ++++++++++++++++---------------
1 file changed, 16 insertions(+), 15 deletions(-)
diff --git a/target/arm/tcg/translate-a64.c b/target/arm/tcg/translate-a64.c
index 79fc7b8c3d..529c5bef0c 100644
--- a/target/arm/tcg/translate-a64.c
+++ b/target/arm/tcg/translate-a64.c
@@ -1842,21 +1842,21 @@ static bool trans_BRAZ(DisasContext *s, arg_braz *a)
static bool trans_BLRAZ(DisasContext *s, arg_braz *a)
{
- TCGv_i64 dst, lr;
+ TCGv_i64 dst, link;
if (!dc_isar_feature(aa64_pauth, s)) {
return false;
}
-
dst = auth_branch_target(s, cpu_reg(s, a->rn), tcg_constant_i64(0), !a->m);
- lr = cpu_reg(s, 30);
- if (dst == lr) {
- TCGv_i64 tmp = tcg_temp_new_i64();
- tcg_gen_mov_i64(tmp, dst);
- dst = tmp;
+
+ link = tcg_temp_new_i64();
+ gen_pc_plus_diff(s, link, 4);
+ if (s->gcs_en) {
+ gen_add_gcs_record(s, link);
}
- gen_pc_plus_diff(s, lr, curr_insn_len(s));
gen_a64_set_pc(s, dst);
+ tcg_gen_mov_i64(cpu_reg(s, 30), link);
+
set_btype_for_blr(s);
s->base.is_jmp = DISAS_JUMP;
return true;
@@ -1892,20 +1892,21 @@ static bool trans_BRA(DisasContext *s, arg_bra *a)
static bool trans_BLRA(DisasContext *s, arg_bra *a)
{
- TCGv_i64 dst, lr;
+ TCGv_i64 dst, link;
if (!dc_isar_feature(aa64_pauth, s)) {
return false;
}
dst = auth_branch_target(s, cpu_reg(s, a->rn), cpu_reg_sp(s, a->rm), !a->m);
- lr = cpu_reg(s, 30);
- if (dst == lr) {
- TCGv_i64 tmp = tcg_temp_new_i64();
- tcg_gen_mov_i64(tmp, dst);
- dst = tmp;
+
+ link = tcg_temp_new_i64();
+ gen_pc_plus_diff(s, link, 4);
+ if (s->gcs_en) {
+ gen_add_gcs_record(s, link);
}
- gen_pc_plus_diff(s, lr, curr_insn_len(s));
gen_a64_set_pc(s, dst);
+ tcg_gen_mov_i64(cpu_reg(s, 30), link);
+
set_btype_for_blr(s);
s->base.is_jmp = DISAS_JUMP;
return true;
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 61/76] target/arm: Load gcs record for RET
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (59 preceding siblings ...)
2025-09-22 18:49 ` [PATCH v5 60/76] target/arm: Add gcs record for BLR with PAuth Richard Henderson
@ 2025-09-22 18:49 ` Richard Henderson
2025-09-22 18:49 ` [PATCH v5 62/76] target/arm: Load gcs record for RET with PAuth Richard Henderson
` (15 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:49 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/tcg/translate-a64.c | 29 ++++++++++++++++++++++++++++-
1 file changed, 28 insertions(+), 1 deletion(-)
diff --git a/target/arm/tcg/translate-a64.c b/target/arm/tcg/translate-a64.c
index 529c5bef0c..462c61032b 100644
--- a/target/arm/tcg/translate-a64.c
+++ b/target/arm/tcg/translate-a64.c
@@ -444,6 +444,27 @@ static void gen_add_gcs_record(DisasContext *s, TCGv_i64 value)
tcg_gen_mov_i64(gcspr, addr);
}
+static void gen_load_check_gcs_record(DisasContext *s, TCGv_i64 target,
+ GCSInstructionType it, int rt)
+{
+ TCGv_i64 gcspr = cpu_gcspr[s->current_el];
+ int mmuidx = core_gcs_mem_index(s->mmu_idx);
+ MemOp mop = finalize_memop(s, MO_64 | MO_ALIGN);
+ TCGv_i64 rec_va = tcg_temp_new_i64();
+
+ tcg_gen_qemu_ld_i64(rec_va, clean_data_tbi(s, gcspr), mmuidx, mop);
+
+ if (s->gcs_rvcen) {
+ TCGLabel *fail_label =
+ delay_exception(s, EXCP_UDEF, syn_gcs_data_check(it, rt));
+
+ tcg_gen_brcond_i64(TCG_COND_NE, rec_va, target, fail_label);
+ }
+
+ gen_a64_set_pc(s, rec_va);
+ tcg_gen_addi_i64(gcspr, gcspr, 8);
+}
+
typedef struct DisasCompare64 {
TCGCond cond;
TCGv_i64 value;
@@ -1798,7 +1819,13 @@ static bool trans_BLR(DisasContext *s, arg_r *a)
static bool trans_RET(DisasContext *s, arg_r *a)
{
- gen_a64_set_pc(s, cpu_reg(s, a->rn));
+ TCGv_i64 target = cpu_reg(s, a->rn);
+
+ if (s->gcs_en) {
+ gen_load_check_gcs_record(s, target, GCS_IT_RET_nPauth, a->rn);
+ } else {
+ gen_a64_set_pc(s, target);
+ }
s->base.is_jmp = DISAS_JUMP;
return true;
}
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 62/76] target/arm: Load gcs record for RET with PAuth
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (60 preceding siblings ...)
2025-09-22 18:49 ` [PATCH v5 61/76] target/arm: Load gcs record for RET Richard Henderson
@ 2025-09-22 18:49 ` Richard Henderson
2025-09-22 18:49 ` [PATCH v5 63/76] target/arm: Copy EXLOCKEn to EXLOCK on exception to the same EL Richard Henderson
` (14 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:49 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/tcg/translate-a64.c | 7 ++++++-
1 file changed, 6 insertions(+), 1 deletion(-)
diff --git a/target/arm/tcg/translate-a64.c b/target/arm/tcg/translate-a64.c
index 462c61032b..aae6dcd548 100644
--- a/target/arm/tcg/translate-a64.c
+++ b/target/arm/tcg/translate-a64.c
@@ -1898,7 +1898,12 @@ static bool trans_RETA(DisasContext *s, arg_reta *a)
}
dst = auth_branch_target(s, cpu_reg(s, 30), cpu_X[31], !a->m);
- gen_a64_set_pc(s, dst);
+ if (s->gcs_en) {
+ GCSInstructionType it = a->m ? GCS_IT_RET_PauthB : GCS_IT_RET_PauthA;
+ gen_load_check_gcs_record(s, dst, it, 30);
+ } else {
+ gen_a64_set_pc(s, dst);
+ }
s->base.is_jmp = DISAS_JUMP;
return true;
}
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 63/76] target/arm: Copy EXLOCKEn to EXLOCK on exception to the same EL
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (61 preceding siblings ...)
2025-09-22 18:49 ` [PATCH v5 62/76] target/arm: Load gcs record for RET with PAuth Richard Henderson
@ 2025-09-22 18:49 ` Richard Henderson
2025-09-22 18:49 ` [PATCH v5 64/76] target/arm: Implement EXLOCK check during exception return Richard Henderson
` (13 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:49 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Per R_WTXBY, PSTATE.EXLOCK is 0 on an exception to a higher EL,
and copied from EXLOCKEn otherwise.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/helper.c | 9 +++++++--
1 file changed, 7 insertions(+), 2 deletions(-)
diff --git a/target/arm/helper.c b/target/arm/helper.c
index e90398acc9..d98f9b0d70 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -9057,8 +9057,13 @@ static void arm_cpu_do_interrupt_aarch64(CPUState *cs)
} else {
addr += 0x600;
}
- } else if (pstate_read(env) & PSTATE_SP) {
- addr += 0x200;
+ } else {
+ if (pstate_read(env) & PSTATE_SP) {
+ addr += 0x200;
+ }
+ if (is_a64(env) && (env->cp15.gcscr_el[new_el] & GCSCR_EXLOCKEN)) {
+ new_mode |= PSTATE_EXLOCK;
+ }
}
switch (cs->exception_index) {
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 64/76] target/arm: Implement EXLOCK check during exception return
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (62 preceding siblings ...)
2025-09-22 18:49 ` [PATCH v5 63/76] target/arm: Copy EXLOCKEn to EXLOCK on exception to the same EL Richard Henderson
@ 2025-09-22 18:49 ` Richard Henderson
2025-09-22 18:49 ` [PATCH v5 65/76] target/arm: Enable FEAT_GCS with -cpu max Richard Henderson
` (12 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:49 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/tcg/helper-a64.c | 11 +++++++++++
1 file changed, 11 insertions(+)
diff --git a/target/arm/tcg/helper-a64.c b/target/arm/tcg/helper-a64.c
index 08b7db7c46..ba1d775d81 100644
--- a/target/arm/tcg/helper-a64.c
+++ b/target/arm/tcg/helper-a64.c
@@ -675,6 +675,17 @@ void HELPER(exception_return)(CPUARMState *env, uint64_t new_pc)
goto illegal_return;
}
+ /*
+ * If GetCurrentEXLOCKEN, the exception return path must use GCSPOPCX,
+ * which will set PSTATE.EXLOCK. We need not explicitly check FEAT_GCS,
+ * because GCSCR_ELx cannot be set without it.
+ */
+ if (new_el == cur_el &&
+ (env->cp15.gcscr_el[cur_el] & GCSCR_EXLOCKEN) &&
+ !(env->pstate & PSTATE_EXLOCK)) {
+ goto illegal_return;
+ }
+
bql_lock();
arm_call_pre_el_change_hook(cpu);
bql_unlock();
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 65/76] target/arm: Enable FEAT_GCS with -cpu max
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (63 preceding siblings ...)
2025-09-22 18:49 ` [PATCH v5 64/76] target/arm: Implement EXLOCK check during exception return Richard Henderson
@ 2025-09-22 18:49 ` Richard Henderson
2025-09-22 18:49 ` [PATCH v5 66/76] linux-user/aarch64: Implement prctls for GCS Richard Henderson
` (11 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:49 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Thiago Jung Bauermann, Pierrick Bouvier
Tested-by: Thiago Jung Bauermann <thiago.bauermann@linaro.org>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/tcg/cpu64.c | 1 +
docs/system/arm/emulation.rst | 1 +
2 files changed, 2 insertions(+)
diff --git a/target/arm/tcg/cpu64.c b/target/arm/tcg/cpu64.c
index 76193826ce..8ebb0a91db 100644
--- a/target/arm/tcg/cpu64.c
+++ b/target/arm/tcg/cpu64.c
@@ -1206,6 +1206,7 @@ void aarch64_max_tcg_initfn(Object *obj)
t = FIELD_DP64(t, ID_AA64PFR1, SME, 2); /* FEAT_SME2 */
t = FIELD_DP64(t, ID_AA64PFR1, CSV2_FRAC, 0); /* FEAT_CSV2_3 */
t = FIELD_DP64(t, ID_AA64PFR1, NMI, 1); /* FEAT_NMI */
+ t = FIELD_DP64(t, ID_AA64PFR1, GCS, 1); /* FEAT_GCS */
SET_IDREG(isar, ID_AA64PFR1, t);
t = GET_IDREG(isar, ID_AA64MMFR0);
diff --git a/docs/system/arm/emulation.rst b/docs/system/arm/emulation.rst
index 4b07af9eac..21c187a464 100644
--- a/docs/system/arm/emulation.rst
+++ b/docs/system/arm/emulation.rst
@@ -73,6 +73,7 @@ the following architecture extensions:
- FEAT_FRINTTS (Floating-point to integer instructions)
- FEAT_FlagM (Flag manipulation instructions v2)
- FEAT_FlagM2 (Enhancements to flag manipulation instructions)
+- FEAT_GCS (Guarded Control Stack Extension)
- FEAT_GTG (Guest translation granule size)
- FEAT_HAFDBS (Hardware management of the access flag and dirty bit state)
- FEAT_HBC (Hinted conditional branches)
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 66/76] linux-user/aarch64: Implement prctls for GCS
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (64 preceding siblings ...)
2025-09-22 18:49 ` [PATCH v5 65/76] target/arm: Enable FEAT_GCS with -cpu max Richard Henderson
@ 2025-09-22 18:49 ` Richard Henderson
2025-09-22 18:49 ` [PATCH v5 67/76] linux-user/aarch64: Allocate new gcs stack on clone Richard Henderson
` (10 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:49 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
This is PR_GET_SHADOW_STACK_STATUS, PR_SET_SHADOW_STACK_STATUS,
and PR_LOCK_SHADOW_STACK_STATUS.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
linux-user/aarch64/gcs-internal.h | 38 ++++++++++++
linux-user/aarch64/target_prctl.h | 96 +++++++++++++++++++++++++++++++
linux-user/qemu.h | 5 ++
linux-user/syscall.c | 29 ++++++++++
4 files changed, 168 insertions(+)
create mode 100644 linux-user/aarch64/gcs-internal.h
diff --git a/linux-user/aarch64/gcs-internal.h b/linux-user/aarch64/gcs-internal.h
new file mode 100644
index 0000000000..e586c7e80e
--- /dev/null
+++ b/linux-user/aarch64/gcs-internal.h
@@ -0,0 +1,38 @@
+/*
+ * AArch64 gcs functions for linux-user
+ *
+ * SPDX-License-Identifier: GPL-2.0-or-later
+ */
+#ifndef AARCH64_GCS_INTERNAL_H
+#define AARCH64_GCS_INTERNAL_H
+
+#ifndef PR_SHADOW_STACK_ENABLE
+# define PR_SHADOW_STACK_ENABLE (1U << 0)
+# define PR_SHADOW_STACK_WRITE (1U << 1)
+# define PR_SHADOW_STACK_PUSH (1U << 2)
+#endif
+
+static inline uint64_t gcs_get_el0_mode(CPUArchState *env)
+{
+ uint64_t cr = env->cp15.gcscr_el[0];
+ abi_ulong flags = 0;
+
+ flags |= cr & GCSCR_PCRSEL ? PR_SHADOW_STACK_ENABLE : 0;
+ flags |= cr & GCSCR_STREN ? PR_SHADOW_STACK_WRITE : 0;
+ flags |= cr & GCSCR_PUSHMEN ? PR_SHADOW_STACK_PUSH : 0;
+
+ return flags;
+}
+
+static inline void gcs_set_el0_mode(CPUArchState *env, uint64_t flags)
+{
+ uint64_t cr = GCSCRE0_NTR;
+
+ cr |= flags & PR_SHADOW_STACK_ENABLE ? GCSCR_RVCHKEN | GCSCR_PCRSEL : 0;
+ cr |= flags & PR_SHADOW_STACK_WRITE ? GCSCR_STREN : 0;
+ cr |= flags & PR_SHADOW_STACK_PUSH ? GCSCR_PUSHMEN : 0;
+
+ env->cp15.gcscr_el[0] = cr;
+}
+
+#endif
diff --git a/linux-user/aarch64/target_prctl.h b/linux-user/aarch64/target_prctl.h
index ed75b9e4b5..621be5727f 100644
--- a/linux-user/aarch64/target_prctl.h
+++ b/linux-user/aarch64/target_prctl.h
@@ -6,8 +6,10 @@
#ifndef AARCH64_TARGET_PRCTL_H
#define AARCH64_TARGET_PRCTL_H
+#include "qemu/units.h"
#include "target/arm/cpu-features.h"
#include "mte_user_helper.h"
+#include "gcs-internal.h"
static abi_long do_prctl_sve_get_vl(CPUArchState *env)
{
@@ -206,4 +208,98 @@ static abi_long do_prctl_get_tagged_addr_ctrl(CPUArchState *env)
}
#define do_prctl_get_tagged_addr_ctrl do_prctl_get_tagged_addr_ctrl
+static abi_long do_prctl_get_shadow_stack_status(CPUArchState *env,
+ abi_long arg2)
+{
+ ARMCPU *cpu = env_archcpu(env);
+
+ if (!cpu_isar_feature(aa64_gcs, cpu)) {
+ return -TARGET_EINVAL;
+ }
+ return put_user_ual(gcs_get_el0_mode(env), arg2);
+}
+#define do_prctl_get_shadow_stack_status do_prctl_get_shadow_stack_status
+
+static abi_long gcs_alloc(abi_ulong hint, abi_ulong size)
+{
+ /*
+ * Without softmmu, we cannot protect GCS memory properly.
+ * Make do with normal read/write permissions. This at least allows
+ * emulation of correct programs which don't access the gcs stack
+ * with normal instructions.
+ */
+ return target_mmap(hint, size, PROT_READ | PROT_WRITE,
+ MAP_PRIVATE | MAP_ANONYMOUS |
+ (hint ? MAP_FIXED_NOREPLACE : 0), -1, 0);
+}
+
+static abi_ulong gcs_new_stack(TaskState *ts)
+{
+ /* Use guest_stack_size as a proxy for RLIMIT_STACK. */
+ abi_ulong size = MIN(MAX(guest_stack_size / 2, TARGET_PAGE_SIZE), 2 * GiB);
+ abi_ulong base = gcs_alloc(0, size);
+
+ if (base == -1) {
+ return -1;
+ }
+
+ ts->gcs_base = base;
+ ts->gcs_size = size;
+ return base + size - 8;
+}
+
+static abi_long do_prctl_set_shadow_stack_status(CPUArchState *env,
+ abi_long new_mode)
+{
+ ARMCPU *cpu = env_archcpu(env);
+ TaskState *ts = get_task_state(env_cpu(env));
+ abi_long cur_mode;
+
+ if (!cpu_isar_feature(aa64_gcs, cpu)) {
+ return -TARGET_EINVAL;
+ }
+ if (new_mode & ~(PR_SHADOW_STACK_ENABLE |
+ PR_SHADOW_STACK_WRITE |
+ PR_SHADOW_STACK_PUSH)) {
+ return -TARGET_EINVAL;
+ }
+
+ cur_mode = gcs_get_el0_mode(env);
+ if ((new_mode ^ cur_mode) & ts->gcs_el0_locked) {
+ return -TARGET_EBUSY;
+ }
+
+ if (new_mode & ~cur_mode & PR_SHADOW_STACK_ENABLE) {
+ abi_long gcspr;
+
+ if (ts->gcs_base || env->cp15.gcspr_el[0]) {
+ return -EINVAL;
+ }
+ gcspr = gcs_new_stack(ts);
+ if (gcspr == -1) {
+ return -TARGET_ENOMEM;
+ }
+ env->cp15.gcspr_el[0] = gcspr;
+ }
+
+ gcs_set_el0_mode(env, new_mode);
+ arm_rebuild_hflags(env);
+ return 0;
+}
+#define do_prctl_set_shadow_stack_status do_prctl_set_shadow_stack_status
+
+static abi_long do_prctl_lock_shadow_stack_status(CPUArchState *env,
+ abi_long arg2)
+{
+ ARMCPU *cpu = env_archcpu(env);
+ TaskState *ts = get_task_state(env_cpu(env));
+
+ if (!cpu_isar_feature(aa64_gcs, cpu)) {
+ return -EINVAL;
+ }
+ ts->gcs_el0_locked |= arg2;
+ return 0;
+}
+#define do_prctl_lock_shadow_stack_status do_prctl_lock_shadow_stack_status
+
#endif /* AARCH64_TARGET_PRCTL_H */
diff --git a/linux-user/qemu.h b/linux-user/qemu.h
index e4dca0c20f..8a0d240405 100644
--- a/linux-user/qemu.h
+++ b/linux-user/qemu.h
@@ -121,6 +121,11 @@ struct TaskState {
abi_ulong child_tidptr;
#ifdef TARGET_M68K
abi_ulong tp_value;
+#endif
+#if defined(TARGET_AARCH64)
+ vaddr gcs_base;
+ abi_ulong gcs_size;
+ abi_ulong gcs_el0_locked;
#endif
int used; /* non zero if used */
struct image_info *info;
diff --git a/linux-user/syscall.c b/linux-user/syscall.c
index 91360a072c..2f1e881046 100644
--- a/linux-user/syscall.c
+++ b/linux-user/syscall.c
@@ -6351,6 +6351,11 @@ abi_long do_arch_prctl(CPUX86State *env, int code, abi_ulong addr)
# define PR_SME_VL_LEN_MASK 0xffff
# define PR_SME_VL_INHERIT (1 << 17)
#endif
+#ifndef PR_GET_SHADOW_STACK_STATUS
+# define PR_GET_SHADOW_STACK_STATUS 74
+# define PR_SET_SHADOW_STACK_STATUS 75
+# define PR_LOCK_SHADOW_STACK_STATUS 76
+#endif
#include "target_prctl.h"
@@ -6397,6 +6402,15 @@ static abi_long do_prctl_inval1(CPUArchState *env, abi_long arg2)
#ifndef do_prctl_sme_set_vl
#define do_prctl_sme_set_vl do_prctl_inval1
#endif
+#ifndef do_prctl_get_shadow_stack_status
+#define do_prctl_get_shadow_stack_status do_prctl_inval1
+#endif
+#ifndef do_prctl_set_shadow_stack_status
+#define do_prctl_set_shadow_stack_status do_prctl_inval1
+#endif
+#ifndef do_prctl_lock_shadow_stack_status
+#define do_prctl_lock_shadow_stack_status do_prctl_inval1
+#endif
static abi_long do_prctl(CPUArchState *env, abi_long option, abi_long arg2,
abi_long arg3, abi_long arg4, abi_long arg5)
@@ -6467,6 +6481,21 @@ static abi_long do_prctl(CPUArchState *env, abi_long option, abi_long arg2,
return -TARGET_EINVAL;
}
return do_prctl_get_tagged_addr_ctrl(env);
+ case PR_GET_SHADOW_STACK_STATUS:
+ if (arg3 || arg4 || arg5) {
+ return -TARGET_EINVAL;
+ }
+ return do_prctl_get_shadow_stack_status(env, arg2);
+ case PR_SET_SHADOW_STACK_STATUS:
+ if (arg3 || arg4 || arg5) {
+ return -TARGET_EINVAL;
+ }
+ return do_prctl_set_shadow_stack_status(env, arg2);
+ case PR_LOCK_SHADOW_STACK_STATUS:
+ if (arg3 || arg4 || arg5) {
+ return -TARGET_EINVAL;
+ }
+ return do_prctl_lock_shadow_stack_status(env, arg2);
case PR_GET_UNALIGN:
return do_prctl_get_unalign(env, arg2);
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 67/76] linux-user/aarch64: Allocate new gcs stack on clone
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (65 preceding siblings ...)
2025-09-22 18:49 ` [PATCH v5 66/76] linux-user/aarch64: Implement prctls for GCS Richard Henderson
@ 2025-09-22 18:49 ` Richard Henderson
2025-09-22 18:49 ` [PATCH v5 68/76] linux-user/aarch64: Release gcs stack on thread exit Richard Henderson
` (9 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:49 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Allocate the new stack early, so that error reporting need
not clean up other objects.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
linux-user/syscall.c | 20 ++++++++++++++++++++
1 file changed, 20 insertions(+)
diff --git a/linux-user/syscall.c b/linux-user/syscall.c
index 2f1e881046..91210775ed 100644
--- a/linux-user/syscall.c
+++ b/linux-user/syscall.c
@@ -6652,6 +6652,21 @@ static int do_fork(CPUArchState *env, unsigned int flags, abi_ulong newsp,
ts = g_new0(TaskState, 1);
init_task_state(ts);
+#ifdef TARGET_AARCH64
+ /*
+ * If GCS is enabled in the parent thread, it is also enabled
+ * in the child thread, but with a newly allocated stack.
+ */
+ abi_long new_gcspr = 0;
+ if (env->cp15.gcscr_el[0] & GCSCR_PCRSEL) {
+ new_gcspr = gcs_new_stack(ts);
+ if (new_gcspr == -1) {
+ g_free(ts);
+ return -TARGET_ENOMEM;
+ }
+ }
+#endif
+
/* Grab a mutex so that thread setup appears atomic. */
pthread_mutex_lock(&clone_lock);
@@ -6676,6 +6691,11 @@ static int do_fork(CPUArchState *env, unsigned int flags, abi_ulong newsp,
ts->info = parent_ts->info;
ts->signal_mask = parent_ts->signal_mask;
+#ifdef TARGET_AARCH64
+ ts->gcs_el0_locked = parent_ts->gcs_el0_locked;
+ new_env->cp15.gcspr_el[0] = new_gcspr;
+#endif
+
if (flags & CLONE_CHILD_CLEARTID) {
ts->child_tidptr = child_tidptr;
}
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 68/76] linux-user/aarch64: Release gcs stack on thread exit
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (66 preceding siblings ...)
2025-09-22 18:49 ` [PATCH v5 67/76] linux-user/aarch64: Allocate new gcs stack on clone Richard Henderson
@ 2025-09-22 18:49 ` Richard Henderson
2025-09-22 18:49 ` [PATCH v5 69/76] linux-user/aarch64: Implement map_shadow_stack syscall Richard Henderson
` (8 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:49 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
linux-user/syscall.c | 6 ++++++
1 file changed, 6 insertions(+)
diff --git a/linux-user/syscall.c b/linux-user/syscall.c
index 91210775ed..dfe114ceb3 100644
--- a/linux-user/syscall.c
+++ b/linux-user/syscall.c
@@ -9309,6 +9309,12 @@ static abi_long do_syscall1(CPUArchState *cpu_env, int num, abi_long arg1,
FUTEX_WAKE, INT_MAX, NULL, NULL, 0);
}
+#ifdef TARGET_AARCH64
+ if (ts->gcs_base) {
+ target_munmap(ts->gcs_base, ts->gcs_size);
+ }
+#endif
+
object_unparent(OBJECT(cpu));
object_unref(OBJECT(cpu));
/*
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 69/76] linux-user/aarch64: Implement map_shadow_stack syscall
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (67 preceding siblings ...)
2025-09-22 18:49 ` [PATCH v5 68/76] linux-user/aarch64: Release gcs stack on thread exit Richard Henderson
@ 2025-09-22 18:49 ` Richard Henderson
2025-09-22 18:49 ` [PATCH v5 70/76] target/arm: Enable GCSPR_EL0 for read in user-mode Richard Henderson
` (7 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:49 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
linux-user/syscall.c | 59 ++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 59 insertions(+)
diff --git a/linux-user/syscall.c b/linux-user/syscall.c
index dfe114ceb3..0a7ce7a262 100644
--- a/linux-user/syscall.c
+++ b/linux-user/syscall.c
@@ -6356,6 +6356,12 @@ abi_long do_arch_prctl(CPUX86State *env, int code, abi_ulong addr)
# define PR_SET_SHADOW_STACK_STATUS 75
# define PR_LOCK_SHADOW_STACK_STATUS 76
#endif
+#ifndef SHADOW_STACK_SET_TOKEN
+# define SHADOW_STACK_SET_TOKEN (1u << 0)
+#endif
+#ifndef SHADOW_STACK_SET_MARKER
+# define SHADOW_STACK_SET_MARKER (1u << 1)
+#endif
#include "target_prctl.h"
@@ -6571,6 +6577,54 @@ static abi_long do_prctl(CPUArchState *env, abi_long option, abi_long arg2,
}
}
+#ifdef TARGET_AARCH64
+static abi_long do_map_shadow_stack(CPUArchState *env, abi_ulong addr,
+ abi_ulong size, abi_int flags)
+{
+ ARMCPU *cpu = env_archcpu(env);
+ abi_ulong alloc_size;
+
+ if (!cpu_isar_feature(aa64_gcs, cpu)) {
+ return -TARGET_EOPNOTSUPP;
+ }
+ if (flags & ~(SHADOW_STACK_SET_TOKEN | SHADOW_STACK_SET_MARKER)) {
+ return -TARGET_EINVAL;
+ }
+ if (addr & ~TARGET_PAGE_MASK) {
+ return -TARGET_EINVAL;
+ }
+ if (size == 8 || !QEMU_IS_ALIGNED(size, 8)) {
+ return -TARGET_EINVAL;
+ }
+
+ alloc_size = TARGET_PAGE_ALIGN(size);
+ if (alloc_size < size) {
+ return -TARGET_EOVERFLOW;
+ }
+
+ mmap_lock();
+ addr = gcs_alloc(addr, alloc_size);
+ if (addr != -1) {
+ if (flags & SHADOW_STACK_SET_TOKEN) {
+ abi_ptr cap_ptr = addr + size - 8;
+ uint64_t cap_val;
+
+ if (flags & SHADOW_STACK_SET_MARKER) {
+ /* Leave an extra empty frame at top-of-stack. */
+ cap_ptr -= 8;
+ }
+ cap_val = (cap_ptr & TARGET_PAGE_MASK) | 1;
+ if (put_user_u64(cap_val, cap_ptr)) {
+ /* Allocation succeeded above. */
+ g_assert_not_reached();
+ }
+ }
+ }
+ mmap_unlock();
+ return get_errno(addr);
+}
+#endif
+
#define NEW_STACK_SIZE 0x40000
@@ -13945,6 +13999,11 @@ static abi_long do_syscall1(CPUArchState *cpu_env, int num, abi_long arg1,
return do_riscv_hwprobe(cpu_env, arg1, arg2, arg3, arg4, arg5);
#endif
+#ifdef TARGET_AARCH64
+ case TARGET_NR_map_shadow_stack:
+ return do_map_shadow_stack(cpu_env, arg1, arg2, arg3);
+#endif
+
default:
qemu_log_mask(LOG_UNIMP, "Unsupported syscall: %d\n", num);
return -TARGET_ENOSYS;
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 70/76] target/arm: Enable GCSPR_EL0 for read in user-mode
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (68 preceding siblings ...)
2025-09-22 18:49 ` [PATCH v5 69/76] linux-user/aarch64: Implement map_shadow_stack syscall Richard Henderson
@ 2025-09-22 18:49 ` Richard Henderson
2025-09-22 18:49 ` [PATCH v5 71/76] linux-user/aarch64: Inject SIGSEGV for GCS faults Richard Henderson
` (6 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:49 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/cpu.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/target/arm/cpu.c b/target/arm/cpu.c
index 731b4d5725..09c27451fd 100644
--- a/target/arm/cpu.c
+++ b/target/arm/cpu.c
@@ -307,6 +307,10 @@ static void arm_cpu_reset_hold(Object *obj, ResetType type)
env->cp15.mdscr_el1 |= 1 << 12;
/* Enable FEAT_MOPS */
env->cp15.sctlr_el[1] |= SCTLR_MSCEN;
+ /* For Linux, GCSPR_EL0 is always readable. */
+ if (cpu_isar_feature(aa64_gcs, cpu)) {
+ env->cp15.gcscr_el[0] = GCSCRE0_NTR;
+ }
#else
/* Reset into the highest available EL */
if (arm_feature(env, ARM_FEATURE_EL3)) {
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 71/76] linux-user/aarch64: Inject SIGSEGV for GCS faults
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (69 preceding siblings ...)
2025-09-22 18:49 ` [PATCH v5 70/76] target/arm: Enable GCSPR_EL0 for read in user-mode Richard Henderson
@ 2025-09-22 18:49 ` Richard Henderson
2025-09-22 18:49 ` [PATCH v5 72/76] linux-user/aarch64: Generate GCS signal records Richard Henderson
` (5 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:49 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
linux-user/aarch64/target_signal.h | 1 +
linux-user/aarch64/cpu_loop.c | 5 +++++
2 files changed, 6 insertions(+)
diff --git a/linux-user/aarch64/target_signal.h b/linux-user/aarch64/target_signal.h
index 6f66a50bfd..e509ac1032 100644
--- a/linux-user/aarch64/target_signal.h
+++ b/linux-user/aarch64/target_signal.h
@@ -7,6 +7,7 @@
#define TARGET_SEGV_MTEAERR 8 /* Asynchronous ARM MTE error */
#define TARGET_SEGV_MTESERR 9 /* Synchronous ARM MTE exception */
+#define TARGET_SEGV_CPERR 10 /* Control protection fault */
#define TARGET_ARCH_HAS_SETUP_FRAME
#define TARGET_ARCH_HAS_SIGTRAMP_PAGE 1
diff --git a/linux-user/aarch64/cpu_loop.c b/linux-user/aarch64/cpu_loop.c
index 6060572eed..41d6b35498 100644
--- a/linux-user/aarch64/cpu_loop.c
+++ b/linux-user/aarch64/cpu_loop.c
@@ -89,6 +89,11 @@ static void signal_for_exception(CPUARMState *env, vaddr addr)
si_code = TARGET_ILL_ILLOPN;
break;
+ case EC_GCS:
+ si_signo = TARGET_SIGSEGV;
+ si_code = TARGET_SEGV_CPERR;
+ break;
+
case EC_MOP:
/*
* FIXME: The kernel fixes up wrong-option exceptions.
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 72/76] linux-user/aarch64: Generate GCS signal records
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (70 preceding siblings ...)
2025-09-22 18:49 ` [PATCH v5 71/76] linux-user/aarch64: Inject SIGSEGV for GCS faults Richard Henderson
@ 2025-09-22 18:49 ` Richard Henderson
2025-09-22 18:49 ` [PATCH v5 73/76] linux-user/aarch64: Enable GCS in HWCAP Richard Henderson
` (4 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:49 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Here we must push and pop a cap on the GCS stack as
well as the gcs record on the normal stack.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
linux-user/aarch64/signal.c | 138 ++++++++++++++++++++++++++++++++++--
1 file changed, 132 insertions(+), 6 deletions(-)
diff --git a/linux-user/aarch64/signal.c b/linux-user/aarch64/signal.c
index ef97be3ac7..f7edfa249e 100644
--- a/linux-user/aarch64/signal.c
+++ b/linux-user/aarch64/signal.c
@@ -22,6 +22,7 @@
#include "signal-common.h"
#include "linux-user/trace.h"
#include "target/arm/cpu-features.h"
+#include "gcs-internal.h"
struct target_sigcontext {
uint64_t fault_address;
@@ -152,6 +153,16 @@ struct target_zt_context {
QEMU_BUILD_BUG_ON(TARGET_ZT_SIG_REG_BYTES != \
sizeof_field(CPUARMState, za_state.zt0));
+#define TARGET_GCS_MAGIC 0x47435300
+#define GCS_SIGNAL_CAP(X) ((X) & TARGET_PAGE_MASK)
+
+struct target_gcs_context {
+ struct target_aarch64_ctx head;
+ uint64_t gcspr;
+ uint64_t features_enabled;
+ uint64_t reserved;
+};
+
struct target_rt_sigframe {
struct target_siginfo info;
struct target_ucontext uc;
@@ -322,6 +333,35 @@ static void target_setup_zt_record(struct target_zt_context *zt,
}
}
+static bool target_setup_gcs_record(struct target_gcs_context *ctx,
+ CPUARMState *env, uint64_t return_addr)
+{
+ uint64_t mode = gcs_get_el0_mode(env);
+ uint64_t gcspr = env->cp15.gcspr_el[0];
+
+ if (mode & PR_SHADOW_STACK_ENABLE) {
+ /* Push a cap for the signal frame. */
+ gcspr -= 8;
+ if (put_user_u64(GCS_SIGNAL_CAP(gcspr), gcspr)) {
+ return false;
+ }
+
+ /* Push a gcs entry for the trampoline. */
+ if (put_user_u64(return_addr, gcspr - 8)) {
+ return false;
+ }
+ env->cp15.gcspr_el[0] = gcspr - 8;
+ }
+
+ __put_user(TARGET_GCS_MAGIC, &ctx->head.magic);
+ __put_user(sizeof(*ctx), &ctx->head.size);
+ __put_user(gcspr, &ctx->gcspr);
+ __put_user(mode, &ctx->features_enabled);
+ __put_user(0, &ctx->reserved);
+
+ return true;
+}
+
static void target_restore_general_frame(CPUARMState *env,
struct target_rt_sigframe *sf)
{
@@ -502,6 +542,64 @@ static bool target_restore_zt_record(CPUARMState *env,
return true;
}
+static bool target_restore_gcs_record(CPUARMState *env,
+ struct target_gcs_context *ctx,
+ bool *rebuild_hflags)
+{
+ TaskState *ts = get_task_state(env_cpu(env));
+ uint64_t cur_mode = gcs_get_el0_mode(env);
+ uint64_t new_mode, gcspr;
+
+ __get_user(new_mode, &ctx->features_enabled);
+ __get_user(gcspr, &ctx->gcspr);
+
+ /*
+ * The kernel pushes the value through the hw register:
+ * write_sysreg_s(gcspr, SYS_GCSPR_EL0) in restore_gcs_context,
+ * then read_sysreg_s(SYS_GCSPR_EL0) in gcs_restore_signal.
+ * Since the bottom 3 bits are RES0, this can (CONSTRAINED UNPREDICTABLE)
+ * force align the value. Mirror the choice from gcspr_write().
+ */
+ gcspr &= ~7;
+
+ if (new_mode & ~(PR_SHADOW_STACK_ENABLE |
+ PR_SHADOW_STACK_WRITE |
+ PR_SHADOW_STACK_PUSH)) {
+ return false;
+ }
+ if ((new_mode ^ cur_mode) & ts->gcs_el0_locked) {
+ return false;
+ }
+ if (new_mode & ~cur_mode & PR_SHADOW_STACK_ENABLE) {
+ return false;
+ }
+
+ if (new_mode & PR_SHADOW_STACK_ENABLE) {
+ uint64_t cap;
+
+ /* Pop and clear the signal cap. */
+ if (get_user_u64(cap, gcspr)) {
+ return false;
+ }
+ if (cap != GCS_SIGNAL_CAP(gcspr)) {
+ return false;
+ }
+ if (put_user_u64(0, gcspr)) {
+ return false;
+ }
+ gcspr += 8;
+ } else {
+ new_mode = 0;
+ }
+
+ env->cp15.gcspr_el[0] = gcspr;
+ if (new_mode != cur_mode) {
+ *rebuild_hflags = true;
+ gcs_set_el0_mode(env, new_mode);
+ }
+ return true;
+}
+
static int target_restore_sigframe(CPUARMState *env,
struct target_rt_sigframe *sf)
{
@@ -511,8 +609,10 @@ static int target_restore_sigframe(CPUARMState *env,
struct target_za_context *za = NULL;
struct target_tpidr2_context *tpidr2 = NULL;
struct target_zt_context *zt = NULL;
+ struct target_gcs_context *gcs = NULL;
uint64_t extra_datap = 0;
bool used_extra = false;
+ bool rebuild_hflags = false;
int sve_size = 0;
int za_size = 0;
int zt_size = 0;
@@ -582,6 +682,15 @@ static int target_restore_sigframe(CPUARMState *env,
zt_size = size;
break;
+ case TARGET_GCS_MAGIC:
+ if (gcs
+ || size != sizeof(struct target_gcs_context)
+ || !cpu_isar_feature(aa64_gcs, env_archcpu(env))) {
+ goto err;
+ }
+ gcs = (struct target_gcs_context *)ctx;
+ break;
+
case TARGET_EXTRA_MAGIC:
if (extra || size != sizeof(struct target_extra_context)) {
goto err;
@@ -612,6 +721,10 @@ static int target_restore_sigframe(CPUARMState *env,
goto err;
}
+ if (gcs && !target_restore_gcs_record(env, gcs, &rebuild_hflags)) {
+ goto err;
+ }
+
/* SVE data, if present, overwrites FPSIMD data. */
if (sve && !target_restore_sve_record(env, sve, sve_size, &svcr)) {
goto err;
@@ -631,6 +744,9 @@ static int target_restore_sigframe(CPUARMState *env,
}
if (env->svcr != svcr) {
env->svcr = svcr;
+ rebuild_hflags = true;
+ }
+ if (rebuild_hflags) {
arm_rebuild_hflags(env);
}
unlock_user(extra, extra_datap, 0);
@@ -701,7 +817,7 @@ static void target_setup_frame(int usig, struct target_sigaction *ka,
uc.tuc_mcontext.__reserved),
};
int fpsimd_ofs, fr_ofs, sve_ofs = 0, za_ofs = 0, tpidr2_ofs = 0;
- int zt_ofs = 0, esr_ofs = 0;
+ int zt_ofs = 0, esr_ofs = 0, gcs_ofs = 0;
int sve_size = 0, za_size = 0, tpidr2_size = 0, zt_size = 0;
struct target_rt_sigframe *frame;
struct target_rt_frame_record *fr;
@@ -720,6 +836,11 @@ static void target_setup_frame(int usig, struct target_sigaction *ka,
&layout);
}
+ if (env->cp15.gcspr_el[0]) {
+ gcs_ofs = alloc_sigframe_space(sizeof(struct target_gcs_context),
+ &layout);
+ }
+
/* SVE state needs saving only if it exists. */
if (cpu_isar_feature(aa64_sve, env_archcpu(env)) ||
cpu_isar_feature(aa64_sme, env_archcpu(env))) {
@@ -779,6 +900,12 @@ static void target_setup_frame(int usig, struct target_sigaction *ka,
goto give_sigsegv;
}
+ if (ka->sa_flags & TARGET_SA_RESTORER) {
+ return_addr = ka->sa_restorer;
+ } else {
+ return_addr = default_rt_sigreturn;
+ }
+
target_setup_general_frame(frame, env, set);
target_setup_fpsimd_record((void *)frame + fpsimd_ofs, env);
if (esr_ofs) {
@@ -786,6 +913,10 @@ static void target_setup_frame(int usig, struct target_sigaction *ka,
/* Leave ESR_EL1 clear while it's not relevant. */
env->cp15.esr_el[1] = 0;
}
+ if (gcs_ofs &&
+ !target_setup_gcs_record((void *)frame + gcs_ofs, env, return_addr)) {
+ goto give_sigsegv;
+ }
target_setup_end_record((void *)frame + layout.std_end_ofs);
if (layout.extra_ofs) {
target_setup_extra_record((void *)frame + layout.extra_ofs,
@@ -811,11 +942,6 @@ static void target_setup_frame(int usig, struct target_sigaction *ka,
__put_user(env->xregs[29], &fr->fp);
__put_user(env->xregs[30], &fr->lr);
- if (ka->sa_flags & TARGET_SA_RESTORER) {
- return_addr = ka->sa_restorer;
- } else {
- return_addr = default_rt_sigreturn;
- }
env->xregs[0] = usig;
env->xregs[29] = frame_addr + fr_ofs;
env->xregs[30] = return_addr;
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 73/76] linux-user/aarch64: Enable GCS in HWCAP
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (71 preceding siblings ...)
2025-09-22 18:49 ` [PATCH v5 72/76] linux-user/aarch64: Generate GCS signal records Richard Henderson
@ 2025-09-22 18:49 ` Richard Henderson
2025-09-22 18:49 ` [PATCH v5 74/76] tests/tcg/aarch64: Add gcsstr Richard Henderson
` (3 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:49 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
linux-user/aarch64/elfload.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/linux-user/aarch64/elfload.c b/linux-user/aarch64/elfload.c
index 77d03b50e1..3af5a37776 100644
--- a/linux-user/aarch64/elfload.c
+++ b/linux-user/aarch64/elfload.c
@@ -169,6 +169,7 @@ abi_ulong get_elf_hwcap(CPUState *cs)
GET_FEATURE_ID(aa64_dcpop, ARM_HWCAP_A64_DCPOP);
GET_FEATURE_ID(aa64_rcpc_8_3, ARM_HWCAP_A64_LRCPC);
GET_FEATURE_ID(aa64_rcpc_8_4, ARM_HWCAP_A64_ILRCPC);
+ GET_FEATURE_ID(aa64_gcs, ARM_HWCAP_A64_GCS);
return hwcaps;
}
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 74/76] tests/tcg/aarch64: Add gcsstr
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (72 preceding siblings ...)
2025-09-22 18:49 ` [PATCH v5 73/76] linux-user/aarch64: Enable GCS in HWCAP Richard Henderson
@ 2025-09-22 18:49 ` Richard Henderson
2025-09-22 18:49 ` [PATCH v5 75/76] tests/tcg/aarch64: Add gcspushm Richard Henderson
` (2 subsequent siblings)
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:49 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Add some infrastructure for testing gcs in userspace.
Validate successful and trapped executions of GCSSTR.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tests/tcg/aarch64/gcs.h | 71 +++++++++++++++++++++++++++++++
tests/tcg/aarch64/gcsstr.c | 48 +++++++++++++++++++++
tests/tcg/aarch64/Makefile.target | 5 +++
3 files changed, 124 insertions(+)
create mode 100644 tests/tcg/aarch64/gcs.h
create mode 100644 tests/tcg/aarch64/gcsstr.c
diff --git a/tests/tcg/aarch64/gcs.h b/tests/tcg/aarch64/gcs.h
new file mode 100644
index 0000000000..99cb4d4e38
--- /dev/null
+++ b/tests/tcg/aarch64/gcs.h
@@ -0,0 +1,71 @@
+/*
+ * Linux kernel fallback API definitions for GCS and test helpers.
+ *
+ * Copyright (c) 2025 Linaro Ltd
+ * SPDX-License-Identifier: GPL-2.0-or-later
+ */
+
+#include <assert.h>
+#include <string.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <stdint.h>
+#include <unistd.h>
+#include <errno.h>
+#include <signal.h>
+#include <sys/mman.h>
+#include <sys/prctl.h>
+#include <sys/syscall.h>
+
+#ifndef PR_GET_SHADOW_STACK_STATUS
+#define PR_GET_SHADOW_STACK_STATUS 74
+#endif
+#ifndef PR_SET_SHADOW_STACK_STATUS
+#define PR_SET_SHADOW_STACK_STATUS 75
+#endif
+#ifndef PR_LOCK_SHADOW_STACK_STATUS
+#define PR_LOCK_SHADOW_STACK_STATUS 76
+#endif
+#ifndef PR_SHADOW_STACK_ENABLE
+# define PR_SHADOW_STACK_ENABLE (1 << 0)
+# define PR_SHADOW_STACK_WRITE (1 << 1)
+# define PR_SHADOW_STACK_PUSH (1 << 2)
+#endif
+#ifndef SHADOW_STACK_SET_TOKEN
+#define SHADOW_STACK_SET_TOKEN (1 << 0)
+#endif
+#ifndef SHADOW_STACK_SET_MARKER
+#define SHADOW_STACK_SET_MARKER (1 << 1)
+#endif
+#ifndef SEGV_CPERR
+#define SEGV_CPERR 10
+#endif
+#ifndef __NR_map_shadow_stack
+#define __NR_map_shadow_stack 453
+#endif
+
+/*
+ * Macros, and implement the syscall inline, lest we fail
+ * the checked return from any function call.
+ */
+#define enable_gcs(flags) \
+ do { \
+ register long num __asm__ ("x8") = __NR_prctl; \
+ register long arg1 __asm__ ("x0") = PR_SET_SHADOW_STACK_STATUS; \
+ register long arg2 __asm__ ("x1") = PR_SHADOW_STACK_ENABLE | flags; \
+ register long arg3 __asm__ ("x2") = 0; \
+ register long arg4 __asm__ ("x3") = 0; \
+ register long arg5 __asm__ ("x4") = 0; \
+ asm volatile("svc #0" \
+ : "+r"(arg1) \
+ : "r"(arg2), "r"(arg3), "r"(arg4), "r"(arg5), "r"(num) \
+ : "memory", "cc"); \
+ if (arg1) { \
+ errno = -arg1; \
+ perror("PR_SET_SHADOW_STACK_STATUS"); \
+ exit(2); \
+ } \
+ } while (0)
+
+#define gcspr() \
+ ({ uint64_t *r; asm volatile("mrs %0, s3_3_c2_c5_1" : "=r"(r)); r; })
diff --git a/tests/tcg/aarch64/gcsstr.c b/tests/tcg/aarch64/gcsstr.c
new file mode 100644
index 0000000000..b045aee925
--- /dev/null
+++ b/tests/tcg/aarch64/gcsstr.c
@@ -0,0 +1,48 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+
+#include "gcs.h"
+
+/*
+ * A single garbage store to the gcs stack.
+ * The asm inside must be unique, so disallow inlining.
+ */
+void __attribute__((noinline))
+test_gcsstr(void)
+{
+ register uint64_t *ptr __asm__("x0") = gcspr();
+ /* GCSSTR x1, x0 */
+ __asm__("inst_gcsstr: .inst 0xd91f1c01" : : "r"(--ptr));
+}
+
+static void test_sigsegv(int sig, siginfo_t *info, void *vuc)
+{
+ ucontext_t *uc = vuc;
+ uint64_t inst_gcsstr;
+
+ __asm__("adr %0, inst_gcsstr" : "=r"(inst_gcsstr));
+ assert(uc->uc_mcontext.pc == inst_gcsstr);
+ assert(info->si_code == SEGV_CPERR);
+ /* TODO: Dig for ESR and verify syndrome. */
+ exit(0);
+}
+
+int main()
+{
+ struct sigaction sa = {
+ .sa_sigaction = test_sigsegv,
+ .sa_flags = SA_SIGINFO,
+ };
+
+ /* Enable GCSSTR and test the store succeeds. */
+ enable_gcs(PR_SHADOW_STACK_WRITE);
+ test_gcsstr();
+
+ /* Disable GCSSTR and test the resulting sigsegv. */
+ enable_gcs(0);
+ if (sigaction(SIGSEGV, &sa, NULL) < 0) {
+ perror("sigaction");
+ exit(1);
+ }
+ test_gcsstr();
+ abort();
+}
diff --git a/tests/tcg/aarch64/Makefile.target b/tests/tcg/aarch64/Makefile.target
index 1755874bee..5e1b3a3385 100644
--- a/tests/tcg/aarch64/Makefile.target
+++ b/tests/tcg/aarch64/Makefile.target
@@ -75,6 +75,11 @@ AARCH64_TESTS += $(SME_TESTS)
$(SME_TESTS): CFLAGS += $(CROSS_AS_HAS_ARMV9_SME)
endif
+# GCS Tests
+GCS_TESTS += gcsstr
+AARCH64_TESTS += $(GCS_TESTS)
+$(GCS_TESTS): gcs.h
+
# System Registers Tests
AARCH64_TESTS += sysregs
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 75/76] tests/tcg/aarch64: Add gcspushm
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (73 preceding siblings ...)
2025-09-22 18:49 ` [PATCH v5 74/76] tests/tcg/aarch64: Add gcsstr Richard Henderson
@ 2025-09-22 18:49 ` Richard Henderson
2025-09-22 18:49 ` [PATCH v5 76/76] tests/tcg/aarch64: Add gcsss Richard Henderson
2025-09-23 23:58 ` [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:49 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Validate successful and trapped executions of GCSPUSHM, GCSPOPM.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tests/tcg/aarch64/gcspushm.c | 71 +++++++++++++++++++++++++++++++
tests/tcg/aarch64/Makefile.target | 2 +-
2 files changed, 72 insertions(+), 1 deletion(-)
create mode 100644 tests/tcg/aarch64/gcspushm.c
diff --git a/tests/tcg/aarch64/gcspushm.c b/tests/tcg/aarch64/gcspushm.c
new file mode 100644
index 0000000000..c330417a2f
--- /dev/null
+++ b/tests/tcg/aarch64/gcspushm.c
@@ -0,0 +1,71 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+
+#include "gcs.h"
+
+
+#define GCSPUSHM "sys #3, c7, c7, #0, %[push]"
+#define GCSPOPM "sysl %[pop], #3, c7, c7, #1"
+
+static void test_sigsegv(int sig, siginfo_t *info, void *vuc)
+{
+ ucontext_t *uc = vuc;
+ uint64_t inst_sigsegv;
+
+ __asm__("adr %0, inst_sigsegv" : "=r"(inst_sigsegv));
+ assert(uc->uc_mcontext.pc == inst_sigsegv);
+ assert(info->si_code == SEGV_CPERR);
+ /* TODO: Dig for ESR and verify syndrome. */
+ uc->uc_mcontext.pc += 4;
+}
+
+static void test_sigill(int sig, siginfo_t *info, void *vuc)
+{
+ ucontext_t *uc = vuc;
+ uint64_t inst_sigill;
+
+ __asm__("adr %0, inst_sigill" : "=r"(inst_sigill));
+ assert(uc->uc_mcontext.pc == inst_sigill);
+ assert(info->si_code == ILL_ILLOPC);
+ uc->uc_mcontext.pc += 4;
+}
+
+int main()
+{
+ struct sigaction sa = { .sa_flags = SA_SIGINFO };
+ uint64_t old, new;
+
+ sa.sa_sigaction = test_sigsegv;
+ if (sigaction(SIGSEGV, &sa, NULL) < 0) {
+ perror("sigaction");
+ exit(1);
+ }
+
+ sa.sa_sigaction = test_sigill;
+ if (sigaction(SIGILL, &sa, NULL) < 0) {
+ perror("sigaction");
+ exit(1);
+ }
+
+ /* Pushm is disabled -- SIGILL via EC_SYSTEMREGISTERTRAP */
+ asm volatile("inst_sigill:\t" GCSPUSHM
+ : : [push] "r" (1));
+
+ enable_gcs(PR_SHADOW_STACK_PUSH);
+
+ /* Valid value -- low 2 bits clear */
+ old = 0xdeadbeeffeedcaec;
+ asm volatile(GCSPUSHM "\n\t" GCSPOPM
+ : [pop] "=r" (new)
+ : [push] "r" (old)
+ : "memory");
+ assert(old == new);
+
+ /* Invalid value -- SIGSEGV via EC_GCS */
+ asm volatile(GCSPUSHM "\n"
+ "inst_sigsegv:\t" GCSPOPM
+ : [pop] "=r" (new)
+ : [push] "r" (1)
+ : "memory");
+
+ exit(0);
+}
diff --git a/tests/tcg/aarch64/Makefile.target b/tests/tcg/aarch64/Makefile.target
index 5e1b3a3385..fddb7bc9cd 100644
--- a/tests/tcg/aarch64/Makefile.target
+++ b/tests/tcg/aarch64/Makefile.target
@@ -76,7 +76,7 @@ $(SME_TESTS): CFLAGS += $(CROSS_AS_HAS_ARMV9_SME)
endif
# GCS Tests
-GCS_TESTS += gcsstr
+GCS_TESTS += gcsstr gcspushm
AARCH64_TESTS += $(GCS_TESTS)
$(GCS_TESTS): gcs.h
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH v5 76/76] tests/tcg/aarch64: Add gcsss
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (74 preceding siblings ...)
2025-09-22 18:49 ` [PATCH v5 75/76] tests/tcg/aarch64: Add gcspushm Richard Henderson
@ 2025-09-22 18:49 ` Richard Henderson
2025-09-23 23:58 ` [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-22 18:49 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm, Pierrick Bouvier
Validate stack switching and recursion depth.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tests/tcg/aarch64/gcs.h | 9 ++++
tests/tcg/aarch64/gcsss.c | 74 +++++++++++++++++++++++++++++++
tests/tcg/aarch64/Makefile.target | 2 +-
3 files changed, 84 insertions(+), 1 deletion(-)
create mode 100644 tests/tcg/aarch64/gcsss.c
diff --git a/tests/tcg/aarch64/gcs.h b/tests/tcg/aarch64/gcs.h
index 99cb4d4e38..e821963cb8 100644
--- a/tests/tcg/aarch64/gcs.h
+++ b/tests/tcg/aarch64/gcs.h
@@ -69,3 +69,12 @@
#define gcspr() \
({ uint64_t *r; asm volatile("mrs %0, s3_3_c2_c5_1" : "=r"(r)); r; })
+
+#define gcsss1(val) \
+ do { \
+ asm volatile("sys #3, c7, c7, #2, %0" : : "r"(val) : "memory"); \
+ } while (0)
+
+#define gcsss2() \
+ ({ uint64_t *r; \
+ asm volatile("sysl %0, #3, c7, c7, #3" : "=r"(r) : : "memory"); r; })
diff --git a/tests/tcg/aarch64/gcsss.c b/tests/tcg/aarch64/gcsss.c
new file mode 100644
index 0000000000..9550c68e7e
--- /dev/null
+++ b/tests/tcg/aarch64/gcsss.c
@@ -0,0 +1,74 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+
+#include "gcs.h"
+
+#define IN_PROGRESS(X) ((uint64_t)(X) | 5)
+#define CAP(X) (((uint64_t)(X) & ~0xfff) + 1)
+
+static uint64_t * __attribute__((noinline)) recurse(size_t index)
+{
+ if (index == 0) {
+ return gcspr();
+ }
+ return recurse(index - 1);
+}
+
+int main()
+{
+ void *tmp;
+ uint64_t *alt_stack, *alt_cap;
+ uint64_t *orig_pr, *orig_cap;
+ uint64_t *bottom;
+ size_t pagesize = getpagesize();
+ size_t words;
+
+ enable_gcs(0);
+ orig_pr = gcspr();
+
+ /* Allocate a guard page before and after. */
+ tmp = mmap(0, 3 * pagesize, PROT_NONE, MAP_ANON | MAP_PRIVATE, -1, 0);
+ assert(tmp != MAP_FAILED);
+
+ /* map_shadow_stack won't replace existing mappings */
+ munmap(tmp + pagesize, pagesize);
+
+ /* Allocate a new stack between the guards. */
+ alt_stack = (uint64_t *)
+ syscall(__NR_map_shadow_stack, tmp + pagesize, pagesize,
+ SHADOW_STACK_SET_TOKEN);
+ assert(alt_stack == tmp + pagesize);
+
+ words = pagesize / 8;
+ alt_cap = alt_stack + words - 1;
+
+ /* SHADOW_STACK_SET_TOKEN set the cap. */
+ assert(*alt_cap == CAP(alt_cap));
+
+ /* Swap to the alt stack, one step at a time. */
+ gcsss1(alt_cap);
+
+ assert(gcspr() == alt_cap);
+ assert(*alt_cap == IN_PROGRESS(orig_pr));
+
+ orig_cap = gcsss2();
+
+ assert(orig_cap == orig_pr - 1);
+ assert(*orig_cap == CAP(orig_cap));
+ assert(gcspr() == alt_stack + words);
+
+ /* We should be able to use the whole stack. */
+ bottom = recurse(words - 1);
+ assert(bottom == alt_stack);
+
+ /* We should be back where we started. */
+ assert(gcspr() == alt_stack + words);
+
+ /* Swap back to the original stack. */
+ gcsss1(orig_cap);
+ tmp = gcsss2();
+
+ assert(gcspr() == orig_pr);
+ assert(tmp == alt_cap);
+
+ exit(0);
+}
diff --git a/tests/tcg/aarch64/Makefile.target b/tests/tcg/aarch64/Makefile.target
index fddb7bc9cd..55ce34e45e 100644
--- a/tests/tcg/aarch64/Makefile.target
+++ b/tests/tcg/aarch64/Makefile.target
@@ -76,7 +76,7 @@ $(SME_TESTS): CFLAGS += $(CROSS_AS_HAS_ARMV9_SME)
endif
# GCS Tests
-GCS_TESTS += gcsstr gcspushm
+GCS_TESTS += gcsstr gcspushm gcsss
AARCH64_TESTS += $(GCS_TESTS)
$(GCS_TESTS): gcs.h
--
2.43.0
^ permalink raw reply related [flat|nested] 78+ messages in thread
* Re: [PATCH v5 00/76] target/arm: Implement FEAT_GCS
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
` (75 preceding siblings ...)
2025-09-22 18:49 ` [PATCH v5 76/76] tests/tcg/aarch64: Add gcsss Richard Henderson
@ 2025-09-23 23:58 ` Richard Henderson
76 siblings, 0 replies; 78+ messages in thread
From: Richard Henderson @ 2025-09-23 23:58 UTC (permalink / raw)
To: qemu-devel; +Cc: qemu-arm
On 9/22/25 11:48, Richard Henderson wrote:
> Based-on: 20250916142238.664316-1-richard.henderson@linaro.org
> ("[PATCH v2 00/36] target/arm: Reorg VHE redirection")
>
> Changes from v4:
> - Master now includes FEAT_ATS1A.
> - Rebase on VHE reorg
> - Trivial adjustments to S1PIE and GCS regs for REG12 redirect.
> - Simplifies EXLOCK access test.
> - Migration fixes based on review.
>
> Note that tests/functional/aarch64/test_rme_* will fail.
> The TF-A images need updating again for at least FEAT_S1PIE
> (EL1 accessing PIRE0_EL1 trapping to EL3 is the fatal fault).
>
> I'm toying with updating the QEMU build to match the FVP build,
> so that all features are runtime enabled.
>
>
> r~
>
>
> Richard Henderson (76):
...> include/hw/core/cpu: Introduce MMUIdxMap
> include/hw/core/cpu: Introduce cpu_tlb_fast
> include/hw/core/cpu: Invert the indexing into CPUTLBDescFast
> target/hppa: Adjust mmu indexes to begin with 0
I'm cherry-picking these 4 into tcg-next.
r~
^ permalink raw reply [flat|nested] 78+ messages in thread
end of thread, other threads:[~2025-09-23 23:58 UTC | newest]
Thread overview: 78+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2025-09-22 18:48 [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
2025-09-22 18:48 ` [PATCH v5 01/76] target/arm: Add isar feature test for FEAT_S1PIE, FEAT_S2PIE Richard Henderson
2025-09-22 18:48 ` [PATCH v5 02/76] target/arm: Enable TCR2_ELx.PIE Richard Henderson
2025-09-22 18:48 ` [PATCH v5 03/76] target/arm: Implement PIR_ELx, PIRE0_ELx, S2PIR_EL2 registers Richard Henderson
2025-09-22 18:48 ` [PATCH v5 04/76] target/arm: Force HPD for stage2 translations Richard Henderson
2025-09-22 18:48 ` [PATCH v5 05/76] target/arm: Cache NV1 early in get_phys_addr_lpae Richard Henderson
2025-09-22 18:48 ` [PATCH v5 06/76] target/arm: Populate PIE in aa64_va_parameters Richard Henderson
2025-09-22 18:48 ` [PATCH v5 07/76] target/arm: Implement get_S1prot_indirect Richard Henderson
2025-09-22 18:48 ` [PATCH v5 08/76] target/arm: Implement get_S2prot_indirect Richard Henderson
2025-09-22 18:48 ` [PATCH v5 09/76] target/arm: Expand CPUARMState.exception.syndrome to 64 bits Richard Henderson
2025-09-22 18:48 ` [PATCH v5 10/76] target/arm: Expand syndrome parameter to raise_exception* Richard Henderson
2025-09-22 18:48 ` [PATCH v5 11/76] target/arm: Implement dirtybit check for PIE Richard Henderson
2025-09-22 18:48 ` [PATCH v5 12/76] target/arm: Enable FEAT_S1PIE and FEAT_S2PIE on -cpu max Richard Henderson
2025-09-22 18:48 ` [PATCH v5 13/76] include/hw/core/cpu: Introduce MMUIdxMap Richard Henderson
2025-09-22 18:48 ` [PATCH v5 14/76] include/hw/core/cpu: Introduce cpu_tlb_fast Richard Henderson
2025-09-22 18:48 ` [PATCH v5 15/76] include/hw/core/cpu: Invert the indexing into CPUTLBDescFast Richard Henderson
2025-09-22 18:48 ` [PATCH v5 16/76] target/hppa: Adjust mmu indexes to begin with 0 Richard Henderson
2025-09-22 18:48 ` [PATCH v5 17/76] include/exec/memopidx: Adjust for 32 mmu indexes Richard Henderson
2025-09-22 18:48 ` [PATCH v5 18/76] include/hw/core/cpu: Widen MMUIdxMap Richard Henderson
2025-09-22 18:48 ` [PATCH v5 19/76] target/arm: Split out mmuidx.h from cpu.h Richard Henderson
2025-09-22 18:48 ` [PATCH v5 20/76] target/arm: Convert arm_mmu_idx_to_el from switch to table Richard Henderson
2025-09-22 18:48 ` [PATCH v5 21/76] target/arm: Remove unused env argument from regime_el Richard Henderson
2025-09-22 18:48 ` [PATCH v5 22/76] target/arm: Convert regime_el from switch to table Richard Henderson
2025-09-22 18:48 ` [PATCH v5 23/76] target/arm: Convert regime_has_2_ranges " Richard Henderson
2025-09-22 18:48 ` [PATCH v5 24/76] target/arm: Remove unused env argument from regime_is_pan Richard Henderson
2025-09-22 18:48 ` [PATCH v5 25/76] target/arm: Convert regime_is_pan from switch to table Richard Henderson
2025-09-22 18:48 ` [PATCH v5 26/76] target/arm: Remove unused env argument from regime_is_user Richard Henderson
2025-09-22 18:48 ` [PATCH v5 27/76] target/arm: Convert regime_is_user from switch to table Richard Henderson
2025-09-22 18:48 ` [PATCH v5 28/76] target/arm: Convert arm_mmu_idx_is_stage1_of_2 " Richard Henderson
2025-09-22 18:48 ` [PATCH v5 29/76] target/arm: Convert regime_is_stage2 " Richard Henderson
2025-09-22 18:48 ` [PATCH v5 30/76] target/arm: Introduce mmu indexes for GCS Richard Henderson
2025-09-22 18:48 ` [PATCH v5 31/76] target/arm: Introduce regime_to_gcs Richard Henderson
2025-09-22 18:48 ` [PATCH v5 32/76] target/arm: Support page protections for GCS mmu indexes Richard Henderson
2025-09-22 18:48 ` [PATCH v5 33/76] target/arm: Implement gcs bit for data abort Richard Henderson
2025-09-22 18:48 ` [PATCH v5 34/76] target/arm: Add GCS cpregs Richard Henderson
2025-09-22 18:48 ` [PATCH v5 35/76] target/arm: Add GCS enable and trap levels to DisasContext Richard Henderson
2025-09-22 18:48 ` [PATCH v5 36/76] target/arm: Implement FEAT_CHK Richard Henderson
2025-09-22 18:48 ` [PATCH v5 37/76] target/arm: Make helper_exception_return system-only Richard Henderson
2025-09-22 18:48 ` [PATCH v5 38/76] target/arm: Export cpsr_{read_for, write_from}_spsr_elx Richard Henderson
2025-09-22 18:48 ` [PATCH v5 39/76] target/arm: Expand pstate to 64 bits Richard Henderson
2025-09-22 18:48 ` [PATCH v5 40/76] target/arm: Add syndrome data for EC_GCS Richard Henderson
2025-09-22 18:48 ` [PATCH v5 41/76] target/arm: Add arm_hcr_el2_nvx_eff Richard Henderson
2025-09-22 18:48 ` [PATCH v5 42/76] target/arm: Use arm_hcr_el2_nvx_eff in access_nv1 Richard Henderson
2025-09-22 18:48 ` [PATCH v5 43/76] target/arm: Split out access_nv1_with_nvx Richard Henderson
2025-09-22 18:48 ` [PATCH v5 44/76] target/arm: Implement EXLOCKException for ELR_ELx and SPSR_ELx Richard Henderson
2025-09-22 18:48 ` [PATCH v5 45/76] target/arm: Split {full,core}_a64_user_mem_index Richard Henderson
2025-09-22 18:48 ` [PATCH v5 46/76] target/arm: Introduce delay_exception{_el} Richard Henderson
2025-09-22 18:48 ` [PATCH v5 47/76] target/arm: Emit HSTR trap exception out of line Richard Henderson
2025-09-22 18:48 ` [PATCH v5 48/76] target/arm: Emit v7m LTPSIZE " Richard Henderson
2025-09-22 18:48 ` [PATCH v5 49/76] target/arm: Implement GCSSTR, GCSSTTR Richard Henderson
2025-09-22 18:48 ` [PATCH v5 50/76] target/arm: Implement GCSB Richard Henderson
2025-09-22 18:48 ` [PATCH v5 51/76] target/arm: Implement GCSPUSHM Richard Henderson
2025-09-22 18:49 ` [PATCH v5 52/76] target/arm: Implement GCSPOPM Richard Henderson
2025-09-22 18:49 ` [PATCH v5 53/76] target/arm: Implement GCSPUSHX Richard Henderson
2025-09-22 18:49 ` [PATCH v5 54/76] target/arm: Implement GCSPOPX Richard Henderson
2025-09-22 18:49 ` [PATCH v5 55/76] target/arm: Implement GCSPOPCX Richard Henderson
2025-09-22 18:49 ` [PATCH v5 56/76] target/arm: Implement GCSSS1 Richard Henderson
2025-09-22 18:49 ` [PATCH v5 57/76] target/arm: Implement GCSSS2 Richard Henderson
2025-09-22 18:49 ` [PATCH v5 58/76] target/arm: Add gcs record for BL Richard Henderson
2025-09-22 18:49 ` [PATCH v5 59/76] target/arm: Add gcs record for BLR Richard Henderson
2025-09-22 18:49 ` [PATCH v5 60/76] target/arm: Add gcs record for BLR with PAuth Richard Henderson
2025-09-22 18:49 ` [PATCH v5 61/76] target/arm: Load gcs record for RET Richard Henderson
2025-09-22 18:49 ` [PATCH v5 62/76] target/arm: Load gcs record for RET with PAuth Richard Henderson
2025-09-22 18:49 ` [PATCH v5 63/76] target/arm: Copy EXLOCKEn to EXLOCK on exception to the same EL Richard Henderson
2025-09-22 18:49 ` [PATCH v5 64/76] target/arm: Implement EXLOCK check during exception return Richard Henderson
2025-09-22 18:49 ` [PATCH v5 65/76] target/arm: Enable FEAT_GCS with -cpu max Richard Henderson
2025-09-22 18:49 ` [PATCH v5 66/76] linux-user/aarch64: Implement prctls for GCS Richard Henderson
2025-09-22 18:49 ` [PATCH v5 67/76] linux-user/aarch64: Allocate new gcs stack on clone Richard Henderson
2025-09-22 18:49 ` [PATCH v5 68/76] linux-user/aarch64: Release gcs stack on thread exit Richard Henderson
2025-09-22 18:49 ` [PATCH v5 69/76] linux-user/aarch64: Implement map_shadow_stack syscall Richard Henderson
2025-09-22 18:49 ` [PATCH v5 70/76] target/arm: Enable GCSPR_EL0 for read in user-mode Richard Henderson
2025-09-22 18:49 ` [PATCH v5 71/76] linux-user/aarch64: Inject SIGSEGV for GCS faults Richard Henderson
2025-09-22 18:49 ` [PATCH v5 72/76] linux-user/aarch64: Generate GCS signal records Richard Henderson
2025-09-22 18:49 ` [PATCH v5 73/76] linux-user/aarch64: Enable GCS in HWCAP Richard Henderson
2025-09-22 18:49 ` [PATCH v5 74/76] tests/tcg/aarch64: Add gcsstr Richard Henderson
2025-09-22 18:49 ` [PATCH v5 75/76] tests/tcg/aarch64: Add gcspushm Richard Henderson
2025-09-22 18:49 ` [PATCH v5 76/76] tests/tcg/aarch64: Add gcsss Richard Henderson
2025-09-23 23:58 ` [PATCH v5 00/76] target/arm: Implement FEAT_GCS Richard Henderson
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).