Most of this is the Neon decodetree patches, followed by Edgar's versal cleanups. thanks -- PMM The following changes since commit 2ef486e76d64436be90f7359a3071fb2a56ce835: Merge remote-tracking branch 'remotes/marcel/tags/rdma-pull-request' into staging (2020-05-03 14:12:56 +0100) are available in the Git repository at: https://git.linaro.org/people/pmaydell/qemu-arm.git tags/pull-target-arm-20200504 for you to fetch changes up to 9aefc6cf9b73f66062d2f914a0136756e7a28211: target/arm: Move gen_ function typedefs to translate.h (2020-05-04 12:59:26 +0100) ---------------------------------------------------------------- target-arm queue: * Start of conversion of Neon insns to decodetree * versal board: support SD and RTC * Implement ARMv8.2-TTS2UXN * Make VQDMULL undefined when U=1 * Some minor code cleanups ---------------------------------------------------------------- Edgar E. Iglesias (11): hw/arm: versal: Remove inclusion of arm_gicv3_common.h hw/arm: versal: Move misplaced comment hw/arm: versal-virt: Fix typo xlnx-ve -> xlnx-versal hw/arm: versal: Embed the UARTs into the SoC type hw/arm: versal: Embed the GEMs into the SoC type hw/arm: versal: Embed the ADMAs into the SoC type hw/arm: versal: Embed the APUs into the SoC type hw/arm: versal: Add support for SD hw/arm: versal: Add support for the RTC hw/arm: versal-virt: Add support for SD hw/arm: versal-virt: Add support for the RTC Fredrik Strupe (1): target/arm: Make VQDMULL undefined when U=1 Peter Maydell (25): target/arm: Don't use a TLB for ARMMMUIdx_Stage2 target/arm: Use enum constant in get_phys_addr_lpae() call target/arm: Add new 's1_is_el0' argument to get_phys_addr_lpae() target/arm: Implement ARMv8.2-TTS2UXN target/arm: Use correct variable for setting 'max' cpu's ID_AA64DFR0 target/arm/translate-vfp.inc.c: Remove duplicate simd_r32 check target/arm: Don't allow Thumb Neon insns without FEATURE_NEON target/arm: Add stubs for AArch32 Neon decodetree target/arm: Convert VCMLA (vector) to decodetree target/arm: Convert VCADD (vector) to decodetree target/arm: Convert V[US]DOT (vector) to decodetree target/arm: Convert VFM[AS]L (vector) to decodetree target/arm: Convert VCMLA (scalar) to decodetree target/arm: Convert V[US]DOT (scalar) to decodetree target/arm: Convert VFM[AS]L (scalar) to decodetree target/arm: Convert Neon load/store multiple structures to decodetree target/arm: Convert Neon 'load single structure to all lanes' to decodetree target/arm: Convert Neon 'load/store single structure' to decodetree target/arm: Convert Neon 3-reg-same VADD/VSUB to decodetree target/arm: Convert Neon 3-reg-same logic ops to decodetree target/arm: Convert Neon 3-reg-same VMAX/VMIN to decodetree target/arm: Convert Neon 3-reg-same comparisons to decodetree target/arm: Convert Neon 3-reg-same VQADD/VQSUB to decodetree target/arm: Convert Neon 3-reg-same VMUL, VMLA, VMLS, VSHL to decodetree target/arm: Move gen_ function typedefs to translate.h Philippe Mathieu-Daudé (2): hw/arm/mps2-tz: Use TYPE_IOTKIT instead of hardcoded string target/arm: Use uint64_t for midr field in CPU state struct include/hw/arm/xlnx-versal.h | 31 +- target/arm/cpu-param.h | 2 +- target/arm/cpu.h | 38 ++- target/arm/translate-a64.h | 9 - target/arm/translate.h | 26 ++ target/arm/neon-dp.decode | 86 +++++ target/arm/neon-ls.decode | 52 +++ target/arm/neon-shared.decode | 66 ++++ hw/arm/mps2-tz.c | 2 +- hw/arm/xlnx-versal-virt.c | 74 ++++- hw/arm/xlnx-versal.c | 115 +++++-- target/arm/cpu.c | 3 +- target/arm/cpu64.c | 8 +- target/arm/helper.c | 183 ++++------ target/arm/translate-a64.c | 17 - target/arm/translate-neon.inc.c | 714 +++++++++++++++++++++++++++++++++++++++ target/arm/translate-vfp.inc.c | 6 - target/arm/translate.c | 716 +++------------------------------------- target/arm/Makefile.objs | 18 + 19 files changed, 1302 insertions(+), 864 deletions(-) create mode 100644 target/arm/neon-dp.decode create mode 100644 target/arm/neon-ls.decode create mode 100644 target/arm/neon-shared.decode create mode 100644 target/arm/translate-neon.inc.c From: Fredrik Strupe <fredrik@strupe.net> According to Arm ARM, VQDMULL is only valid when U=0, while having U=1 is unallocated. Signed-off-by: Fredrik Strupe <fredrik@strupe.net> Fixes: 695272dcb976 ("target-arm: Handle UNDEF cases for Neon 3-regs-different-widths") Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) {0, 0, 0, 0}, /* VMLSL */ {0, 0, 0, 9}, /* VQDMLSL */ {0, 0, 0, 0}, /* Integer VMULL */ - {0, 0, 0, 1}, /* VQDMULL */ + {0, 0, 0, 9}, /* VQDMULL */ {0, 0, 0, 0xa}, /* Polynomial VMULL */ {0, 0, 0, 7}, /* Reserved: always UNDEF */ }; -- 2.20.1 From: Philippe Mathieu-Daudé <f4bug@amsat.org> By using the TYPE_* definitions for devices, we can: - quickly find where devices are used with 'git-grep' - easily rename a device (one-line change). Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Message-id: 20200428154650.21991-1-f4bug@amsat.org Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- hw/arm/mps2-tz.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/hw/arm/mps2-tz.c b/hw/arm/mps2-tz.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/mps2-tz.c +++ b/hw/arm/mps2-tz.c @@ -XXX,XX +XXX,XX @@ static void mps2tz_common_init(MachineState *machine) exit(EXIT_FAILURE); } - sysbus_init_child_obj(OBJECT(machine), "iotkit", &mms->iotkit, + sysbus_init_child_obj(OBJECT(machine), TYPE_IOTKIT, &mms->iotkit, sizeof(mms->iotkit), mmc->armsse_type); iotkitdev = DEVICE(&mms->iotkit); object_property_set_link(OBJECT(&mms->iotkit), OBJECT(system_memory), -- 2.20.1 We define ARMMMUIdx_Stage2 as being an MMU index which uses a QEMU TLB. However we never actually use the TLB -- all stage 2 lookups are done by direct calls to get_phys_addr_lpae() followed by a physical address load via address_space_ld*(). Remove Stage2 from the list of ARM MMU indexes which correspond to real core MMU indexes, and instead put it in the set of "NOTLB" ARM MMU indexes. This allows us to drop NB_MMU_MODES to 11. It also means we can safely add support for the ARMv8.3-TTS2UXN extension, which adds permission bits to the stage 2 descriptors which define execute permission separatel for EL0 and EL1; supporting that while keeping Stage2 in a QEMU TLB would require us to use separate TLBs for "Stage2 for an EL0 access" and "Stage2 for an EL1 access", which is a lot of extra complication given we aren't even using the QEMU TLB. In the process of updating the comment on our MMU index use, fix a couple of other minor errors: * NS EL2 EL2&0 was missing from the list in the comment * some text hadn't been updated from when we bumped NB_MMU_MODES above 8 Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200330210400.11724-2-peter.maydell@linaro.org --- target/arm/cpu-param.h | 2 +- target/arm/cpu.h | 21 +++++--- target/arm/helper.c | 112 ++++------------------------------------- 3 files changed, 27 insertions(+), 108 deletions(-) diff --git a/target/arm/cpu-param.h b/target/arm/cpu-param.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu-param.h +++ b/target/arm/cpu-param.h @@ -XXX,XX +XXX,XX @@ # define TARGET_PAGE_BITS_MIN 10 #endif -#define NB_MMU_MODES 12 +#define NB_MMU_MODES 11 #endif diff --git a/target/arm/cpu.h b/target/arm/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.h +++ b/target/arm/cpu.h @@ -XXX,XX +XXX,XX @@ bool write_cpustate_to_list(ARMCPU *cpu, bool kvm_sync); * 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. @@ -XXX,XX +XXX,XX @@ bool write_cpustate_to_list(ARMCPU *cpu, bool kvm_sync); * NS EL1 EL1&0 stage 1+2 (aka NS PL1) * NS EL1 EL1&0 stage 1+2 +PAN * NS EL0 EL2&0 + * NS EL2 EL2&0 * NS EL2 EL2&0 +PAN * NS EL2 (aka NS PL2) * S EL0 EL1&0 (aka S PL0) * S EL1 EL1&0 (not used if EL3 is 32 bit) * S EL1 EL1&0 +PAN * S EL3 (aka S PL1) - * NS EL1&0 stage 2 * - * for a total of 12 different mmu_idx. + * for a total of 11 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 NS EL0 and NS EL1 (and @@ -XXX,XX +XXX,XX @@ bool write_cpustate_to_list(ARMCPU *cpu, bool kvm_sync); * 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 setting NB_MMU_MODES to more than 8. The lower bits of + * 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 @@ -XXX,XX +XXX,XX @@ typedef enum ARMMMUIdx { ARMMMUIdx_SE10_1_PAN = 9 | ARM_MMU_IDX_A, ARMMMUIdx_SE3 = 10 | ARM_MMU_IDX_A, - ARMMMUIdx_Stage2 = 11 | 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. @@ -XXX,XX +XXX,XX @@ 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, + /* + * Not allocated a TLB: used only for second stage of an S12 page + * table walk, or for descriptor loads during first stage of an S1 + * page table walk. Note that if we ever want to have a TLB for this + * then various TLB flush insns which currently are no-ops or flush + * only stage 1 MMU indexes will need to change to flush stage 2. + */ + ARMMMUIdx_Stage2 = 3 | ARM_MMU_IDX_NOTLB, /* * M-profile. @@ -XXX,XX +XXX,XX @@ typedef enum ARMMMUIdxBit { TO_CORE_BIT(SE10_1), TO_CORE_BIT(SE10_1_PAN), TO_CORE_BIT(SE3), - TO_CORE_BIT(Stage2), TO_CORE_BIT(MUser), TO_CORE_BIT(MPriv), diff --git a/target/arm/helper.c b/target/arm/helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper.c +++ b/target/arm/helper.c @@ -XXX,XX +XXX,XX @@ static void tlbiall_nsnh_write(CPUARMState *env, const ARMCPRegInfo *ri, tlb_flush_by_mmuidx(cs, ARMMMUIdxBit_E10_1 | ARMMMUIdxBit_E10_1_PAN | - ARMMMUIdxBit_E10_0 | - ARMMMUIdxBit_Stage2); + ARMMMUIdxBit_E10_0); } static void tlbiall_nsnh_is_write(CPUARMState *env, const ARMCPRegInfo *ri, @@ -XXX,XX +XXX,XX @@ static void tlbiall_nsnh_is_write(CPUARMState *env, const ARMCPRegInfo *ri, tlb_flush_by_mmuidx_all_cpus_synced(cs, ARMMMUIdxBit_E10_1 | ARMMMUIdxBit_E10_1_PAN | - ARMMMUIdxBit_E10_0 | - ARMMMUIdxBit_Stage2); + ARMMMUIdxBit_E10_0); } -static void tlbiipas2_write(CPUARMState *env, const ARMCPRegInfo *ri, - uint64_t value) -{ - /* Invalidate by IPA. This has to invalidate any structures that - * contain only stage 2 translation information, but does not need - * to apply to structures that contain combined stage 1 and stage 2 - * translation information. - * This must NOP if EL2 isn't implemented or SCR_EL3.NS is zero. - */ - CPUState *cs = env_cpu(env); - uint64_t pageaddr; - - if (!arm_feature(env, ARM_FEATURE_EL2) || !(env->cp15.scr_el3 & SCR_NS)) { - return; - } - - pageaddr = sextract64(value << 12, 0, 40); - - tlb_flush_page_by_mmuidx(cs, pageaddr, ARMMMUIdxBit_Stage2); -} - -static void tlbiipas2_is_write(CPUARMState *env, const ARMCPRegInfo *ri, - uint64_t value) -{ - CPUState *cs = env_cpu(env); - uint64_t pageaddr; - - if (!arm_feature(env, ARM_FEATURE_EL2) || !(env->cp15.scr_el3 & SCR_NS)) { - return; - } - - pageaddr = sextract64(value << 12, 0, 40); - - tlb_flush_page_by_mmuidx_all_cpus_synced(cs, pageaddr, - ARMMMUIdxBit_Stage2); -} static void tlbiall_hyp_write(CPUARMState *env, const ARMCPRegInfo *ri, uint64_t value) @@ -XXX,XX +XXX,XX @@ static void vttbr_write(CPUARMState *env, const ARMCPRegInfo *ri, tlb_flush_by_mmuidx(cs, ARMMMUIdxBit_E10_1 | ARMMMUIdxBit_E10_1_PAN | - ARMMMUIdxBit_E10_0 | - ARMMMUIdxBit_Stage2); + ARMMMUIdxBit_E10_0); raw_write(env, ri, value); } } @@ -XXX,XX +XXX,XX @@ static int alle1_tlbmask(CPUARMState *env) return ARMMMUIdxBit_SE10_1 | ARMMMUIdxBit_SE10_1_PAN | ARMMMUIdxBit_SE10_0; - } else if (arm_feature(env, ARM_FEATURE_EL2)) { - return ARMMMUIdxBit_E10_1 | - ARMMMUIdxBit_E10_1_PAN | - ARMMMUIdxBit_E10_0 | - ARMMMUIdxBit_Stage2; } else { return ARMMMUIdxBit_E10_1 | ARMMMUIdxBit_E10_1_PAN | @@ -XXX,XX +XXX,XX @@ static void tlbi_aa64_vae3is_write(CPUARMState *env, const ARMCPRegInfo *ri, ARMMMUIdxBit_SE3); } -static void tlbi_aa64_ipas2e1_write(CPUARMState *env, const ARMCPRegInfo *ri, - uint64_t value) -{ - /* Invalidate by IPA. This has to invalidate any structures that - * contain only stage 2 translation information, but does not need - * to apply to structures that contain combined stage 1 and stage 2 - * translation information. - * This must NOP if EL2 isn't implemented or SCR_EL3.NS is zero. - */ - ARMCPU *cpu = env_archcpu(env); - CPUState *cs = CPU(cpu); - uint64_t pageaddr; - - if (!arm_feature(env, ARM_FEATURE_EL2) || !(env->cp15.scr_el3 & SCR_NS)) { - return; - } - - pageaddr = sextract64(value << 12, 0, 48); - - tlb_flush_page_by_mmuidx(cs, pageaddr, ARMMMUIdxBit_Stage2); -} - -static void tlbi_aa64_ipas2e1is_write(CPUARMState *env, const ARMCPRegInfo *ri, - uint64_t value) -{ - CPUState *cs = env_cpu(env); - uint64_t pageaddr; - - if (!arm_feature(env, ARM_FEATURE_EL2) || !(env->cp15.scr_el3 & SCR_NS)) { - return; - } - - pageaddr = sextract64(value << 12, 0, 48); - - tlb_flush_page_by_mmuidx_all_cpus_synced(cs, pageaddr, - ARMMMUIdxBit_Stage2); -} - static CPAccessResult aa64_zva_access(CPUARMState *env, const ARMCPRegInfo *ri, bool isread) { @@ -XXX,XX +XXX,XX @@ static const ARMCPRegInfo v8_cp_reginfo[] = { .writefn = tlbi_aa64_vae1_write }, { .name = "TLBI_IPAS2E1IS", .state = ARM_CP_STATE_AA64, .opc0 = 1, .opc1 = 4, .crn = 8, .crm = 0, .opc2 = 1, - .access = PL2_W, .type = ARM_CP_NO_RAW, - .writefn = tlbi_aa64_ipas2e1is_write }, + .access = PL2_W, .type = ARM_CP_NOP }, { .name = "TLBI_IPAS2LE1IS", .state = ARM_CP_STATE_AA64, .opc0 = 1, .opc1 = 4, .crn = 8, .crm = 0, .opc2 = 5, - .access = PL2_W, .type = ARM_CP_NO_RAW, - .writefn = tlbi_aa64_ipas2e1is_write }, + .access = PL2_W, .type = ARM_CP_NOP }, { .name = "TLBI_ALLE1IS", .state = ARM_CP_STATE_AA64, .opc0 = 1, .opc1 = 4, .crn = 8, .crm = 3, .opc2 = 4, .access = PL2_W, .type = ARM_CP_NO_RAW, @@ -XXX,XX +XXX,XX @@ static const ARMCPRegInfo v8_cp_reginfo[] = { .writefn = tlbi_aa64_alle1is_write }, { .name = "TLBI_IPAS2E1", .state = ARM_CP_STATE_AA64, .opc0 = 1, .opc1 = 4, .crn = 8, .crm = 4, .opc2 = 1, - .access = PL2_W, .type = ARM_CP_NO_RAW, - .writefn = tlbi_aa64_ipas2e1_write }, + .access = PL2_W, .type = ARM_CP_NOP }, { .name = "TLBI_IPAS2LE1", .state = ARM_CP_STATE_AA64, .opc0 = 1, .opc1 = 4, .crn = 8, .crm = 4, .opc2 = 5, - .access = PL2_W, .type = ARM_CP_NO_RAW, - .writefn = tlbi_aa64_ipas2e1_write }, + .access = PL2_W, .type = ARM_CP_NOP }, { .name = "TLBI_ALLE1", .state = ARM_CP_STATE_AA64, .opc0 = 1, .opc1 = 4, .crn = 8, .crm = 7, .opc2 = 4, .access = PL2_W, .type = ARM_CP_NO_RAW, @@ -XXX,XX +XXX,XX @@ static const ARMCPRegInfo v8_cp_reginfo[] = { .writefn = tlbimva_hyp_is_write }, { .name = "TLBIIPAS2", .cp = 15, .opc1 = 4, .crn = 8, .crm = 4, .opc2 = 1, - .type = ARM_CP_NO_RAW, .access = PL2_W, - .writefn = tlbiipas2_write }, + .type = ARM_CP_NOP, .access = PL2_W }, { .name = "TLBIIPAS2IS", .cp = 15, .opc1 = 4, .crn = 8, .crm = 0, .opc2 = 1, - .type = ARM_CP_NO_RAW, .access = PL2_W, - .writefn = tlbiipas2_is_write }, + .type = ARM_CP_NOP, .access = PL2_W }, { .name = "TLBIIPAS2L", .cp = 15, .opc1 = 4, .crn = 8, .crm = 4, .opc2 = 5, - .type = ARM_CP_NO_RAW, .access = PL2_W, - .writefn = tlbiipas2_write }, + .type = ARM_CP_NOP, .access = PL2_W }, { .name = "TLBIIPAS2LIS", .cp = 15, .opc1 = 4, .crn = 8, .crm = 0, .opc2 = 5, - .type = ARM_CP_NO_RAW, .access = PL2_W, - .writefn = tlbiipas2_is_write }, + .type = ARM_CP_NOP, .access = PL2_W }, /* 32 bit cache operations */ { .name = "ICIALLUIS", .cp = 15, .opc1 = 0, .crn = 7, .crm = 1, .opc2 = 0, .type = ARM_CP_NOP, .access = PL1_W, .accessfn = aa64_cacheop_pou_access }, -- 2.20.1 The access_type argument to get_phys_addr_lpae() is an MMUAccessType; use the enum constant MMU_DATA_LOAD rather than a literal 0 when we call it in S1_ptw_translate(). Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200330210400.11724-3-peter.maydell@linaro.org --- target/arm/helper.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/target/arm/helper.c b/target/arm/helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper.c +++ b/target/arm/helper.c @@ -XXX,XX +XXX,XX @@ static hwaddr S1_ptw_translate(CPUARMState *env, ARMMMUIdx mmu_idx, pcacheattrs = &cacheattrs; } - ret = get_phys_addr_lpae(env, addr, 0, ARMMMUIdx_Stage2, &s2pa, - &txattrs, &s2prot, &s2size, fi, pcacheattrs); + ret = get_phys_addr_lpae(env, addr, MMU_DATA_LOAD, ARMMMUIdx_Stage2, + &s2pa, &txattrs, &s2prot, &s2size, fi, + pcacheattrs); if (ret) { assert(fi->type != ARMFault_None); fi->s2addr = addr; -- 2.20.1 For ARMv8.2-TTS2UXN, the stage 2 page table walk wants to know whether the stage 1 access is for EL0 or not, because whether exec permission is given can depend on whether this is an EL0 or EL1 access. Add a new argument to get_phys_addr_lpae() so the call sites can pass this information in. Since get_phys_addr_lpae() doesn't already have a doc comment, add one so we have a place to put the documentation of the semantics of the new s1_is_el0 argument. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200330210400.11724-4-peter.maydell@linaro.org --- target/arm/helper.c | 29 ++++++++++++++++++++++++++++- 1 file changed, 28 insertions(+), 1 deletion(-) diff --git a/target/arm/helper.c b/target/arm/helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper.c +++ b/target/arm/helper.c @@ -XXX,XX +XXX,XX @@ static bool get_phys_addr_lpae(CPUARMState *env, target_ulong address, MMUAccessType access_type, ARMMMUIdx mmu_idx, + bool s1_is_el0, hwaddr *phys_ptr, MemTxAttrs *txattrs, int *prot, target_ulong *page_size_ptr, ARMMMUFaultInfo *fi, ARMCacheAttrs *cacheattrs); @@ -XXX,XX +XXX,XX @@ static hwaddr S1_ptw_translate(CPUARMState *env, ARMMMUIdx mmu_idx, } ret = get_phys_addr_lpae(env, addr, MMU_DATA_LOAD, ARMMMUIdx_Stage2, + false, &s2pa, &txattrs, &s2prot, &s2size, fi, pcacheattrs); if (ret) { @@ -XXX,XX +XXX,XX @@ static ARMVAParameters aa32_va_parameters(CPUARMState *env, uint32_t va, }; } +/** + * get_phys_addr_lpae: perform one stage of page table walk, LPAE format + * + * Returns false if the translation was successful. Otherwise, phys_ptr, attrs, + * prot and page_size may not be filled in, and the populated fsr value provides + * information on why the translation aborted, in the format of a long-format + * DFSR/IFSR fault register, with the following caveats: + * * the WnR bit is never set (the caller must do this). + * + * @env: CPUARMState + * @address: virtual address to get physical address for + * @access_type: MMU_DATA_LOAD, MMU_DATA_STORE or MMU_INST_FETCH + * @mmu_idx: MMU index indicating required translation regime + * @s1_is_el0: if @mmu_idx is ARMMMUIdx_Stage2 (so this is a stage 2 page table + * walk), must be true if this is stage 2 of a stage 1+2 walk for an + * EL0 access). If @mmu_idx is anything else, @s1_is_el0 is ignored. + * @phys_ptr: set to the physical address corresponding to the virtual address + * @attrs: set to the memory transaction attributes to use + * @prot: set to the permissions for the page containing phys_ptr + * @page_size_ptr: set to the size of the page containing phys_ptr + * @fi: set to fault info if the translation fails + * @cacheattrs: (if non-NULL) set to the cacheability/shareability attributes + */ static bool get_phys_addr_lpae(CPUARMState *env, target_ulong address, MMUAccessType access_type, ARMMMUIdx mmu_idx, + bool s1_is_el0, hwaddr *phys_ptr, MemTxAttrs *txattrs, int *prot, target_ulong *page_size_ptr, ARMMMUFaultInfo *fi, ARMCacheAttrs *cacheattrs) @@ -XXX,XX +XXX,XX @@ bool get_phys_addr(CPUARMState *env, target_ulong address, /* S1 is done. Now do S2 translation. */ ret = get_phys_addr_lpae(env, ipa, access_type, ARMMMUIdx_Stage2, + mmu_idx == ARMMMUIdx_E10_0, phys_ptr, attrs, &s2_prot, page_size, fi, cacheattrs != NULL ? &cacheattrs2 : NULL); @@ -XXX,XX +XXX,XX @@ bool get_phys_addr(CPUARMState *env, target_ulong address, } if (regime_using_lpae_format(env, mmu_idx)) { - return get_phys_addr_lpae(env, address, access_type, mmu_idx, + return get_phys_addr_lpae(env, address, access_type, mmu_idx, false, phys_ptr, attrs, prot, page_size, fi, cacheattrs); } else if (regime_sctlr(env, mmu_idx) & SCTLR_XP) { -- 2.20.1 The ARMv8.2-TTS2UXN feature extends the XN field in stage 2 translation table descriptors from just bit [54] to bits [54:53], allowing stage 2 to control execution permissions separately for EL0 and EL1. Implement the new semantics of the XN field and enable the feature for our 'max' CPU. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200330210400.11724-5-peter.maydell@linaro.org --- target/arm/cpu.h | 15 +++++++++++++++ target/arm/cpu.c | 1 + target/arm/cpu64.c | 2 ++ target/arm/helper.c | 37 +++++++++++++++++++++++++++++++------ 4 files changed, 49 insertions(+), 6 deletions(-) diff --git a/target/arm/cpu.h b/target/arm/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.h +++ b/target/arm/cpu.h @@ -XXX,XX +XXX,XX @@ static inline bool isar_feature_aa32_ccidx(const ARMISARegisters *id) return FIELD_EX32(id->id_mmfr4, ID_MMFR4, CCIDX) != 0; } +static inline bool isar_feature_aa32_tts2uxn(const ARMISARegisters *id) +{ + return FIELD_EX32(id->id_mmfr4, ID_MMFR4, XNX) != 0; +} + /* * 64-bit feature tests via id registers. */ @@ -XXX,XX +XXX,XX @@ static inline bool isar_feature_aa64_ccidx(const ARMISARegisters *id) return FIELD_EX64(id->id_aa64mmfr2, ID_AA64MMFR2, CCIDX) != 0; } +static inline bool isar_feature_aa64_tts2uxn(const ARMISARegisters *id) +{ + return FIELD_EX64(id->id_aa64mmfr1, ID_AA64MMFR1, XNX) != 0; +} + /* * Feature tests for "does this exist in either 32-bit or 64-bit?" */ @@ -XXX,XX +XXX,XX @@ static inline bool isar_feature_any_ccidx(const ARMISARegisters *id) return isar_feature_aa64_ccidx(id) || isar_feature_aa32_ccidx(id); } +static inline bool isar_feature_any_tts2uxn(const ARMISARegisters *id) +{ + return isar_feature_aa64_tts2uxn(id) || isar_feature_aa32_tts2uxn(id); +} + /* * Forward to the above feature tests given an ARMCPU pointer. */ diff --git a/target/arm/cpu.c b/target/arm/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.c +++ b/target/arm/cpu.c @@ -XXX,XX +XXX,XX @@ static void arm_max_initfn(Object *obj) t = FIELD_DP32(t, ID_MMFR4, HPDS, 1); /* AA32HPD */ t = FIELD_DP32(t, ID_MMFR4, AC2, 1); /* ACTLR2, HACTLR2 */ t = FIELD_DP32(t, ID_MMFR4, CNP, 1); /* TTCNP */ + t = FIELD_DP32(t, ID_MMFR4, XNX, 1); /* TTS2UXN */ cpu->isar.id_mmfr4 = t; } #endif diff --git a/target/arm/cpu64.c b/target/arm/cpu64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu64.c +++ b/target/arm/cpu64.c @@ -XXX,XX +XXX,XX @@ static void aarch64_max_initfn(Object *obj) t = FIELD_DP64(t, ID_AA64MMFR1, VH, 1); t = FIELD_DP64(t, ID_AA64MMFR1, PAN, 2); /* ATS1E1 */ t = FIELD_DP64(t, ID_AA64MMFR1, VMIDBITS, 2); /* VMID16 */ + t = FIELD_DP64(t, ID_AA64MMFR1, XNX, 1); /* TTS2UXN */ cpu->isar.id_aa64mmfr1 = t; t = cpu->isar.id_aa64mmfr2; @@ -XXX,XX +XXX,XX @@ static void aarch64_max_initfn(Object *obj) u = FIELD_DP32(u, ID_MMFR4, HPDS, 1); /* AA32HPD */ u = FIELD_DP32(u, ID_MMFR4, AC2, 1); /* ACTLR2, HACTLR2 */ u = FIELD_DP32(u, ID_MMFR4, CNP, 1); /* TTCNP */ + u = FIELD_DP32(u, ID_MMFR4, XNX, 1); /* TTS2UXN */ cpu->isar.id_mmfr4 = u; u = cpu->isar.id_aa64dfr0; diff --git a/target/arm/helper.c b/target/arm/helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper.c +++ b/target/arm/helper.c @@ -XXX,XX +XXX,XX @@ simple_ap_to_rw_prot(CPUARMState *env, ARMMMUIdx mmu_idx, int ap) * * @env: CPUARMState * @s2ap: The 2-bit stage2 access permissions (S2AP) - * @xn: XN (execute-never) bit + * @xn: XN (execute-never) bits + * @s1_is_el0: true if this is S2 of an S1+2 walk for EL0 */ -static int get_S2prot(CPUARMState *env, int s2ap, int xn) +static int get_S2prot(CPUARMState *env, int s2ap, int xn, bool s1_is_el0) { int prot = 0; @@ -XXX,XX +XXX,XX @@ static int get_S2prot(CPUARMState *env, int s2ap, int xn) if (s2ap & 2) { prot |= PAGE_WRITE; } - if (!xn) { - if (arm_el_is_aa64(env, 2) || prot & PAGE_READ) { + + if (cpu_isar_feature(any_tts2uxn, env_archcpu(env))) { + switch (xn) { + case 0: prot |= PAGE_EXEC; + break; + case 1: + if (s1_is_el0) { + prot |= PAGE_EXEC; + } + break; + case 2: + break; + case 3: + if (!s1_is_el0) { + prot |= PAGE_EXEC; + } + break; + default: + g_assert_not_reached(); + } + } else { + if (!extract32(xn, 1, 1)) { + if (arm_el_is_aa64(env, 2) || prot & PAGE_READ) { + prot |= PAGE_EXEC; + } } } return prot; @@ -XXX,XX +XXX,XX @@ static bool get_phys_addr_lpae(CPUARMState *env, target_ulong address, } ap = extract32(attrs, 4, 2); - xn = extract32(attrs, 12, 1); if (mmu_idx == ARMMMUIdx_Stage2) { ns = true; - *prot = get_S2prot(env, ap, xn); + xn = extract32(attrs, 11, 2); + *prot = get_S2prot(env, ap, xn, s1_is_el0); } else { ns = extract32(attrs, 3, 1); + xn = extract32(attrs, 12, 1); pxn = extract32(attrs, 11, 1); *prot = get_S1prot(env, mmu_idx, aarch64, ap, ns, xn, pxn); } -- 2.20.1 In aarch64_max_initfn() we update both 32-bit and 64-bit ID registers. The intended pattern is that for 64-bit ID registers we use FIELD_DP64 and the uint64_t 't' register, while 32-bit ID registers use FIELD_DP32 and the uint32_t 'u' register. For ID_AA64DFR0 we accidentally used 'u', meaning that the top 32 bits of this 64-bit ID register would end up always zero. Luckily at the moment that's what they should be anyway, so this bug has no visible effects. Use the right-sized variable. Fixes: 3bec78447a958d481991 Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Laurent Desnogues <laurent.desnogues@gmail.com> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Message-id: 20200423110915.10527-1-peter.maydell@linaro.org --- target/arm/cpu64.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/target/arm/cpu64.c b/target/arm/cpu64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu64.c +++ b/target/arm/cpu64.c @@ -XXX,XX +XXX,XX @@ static void aarch64_max_initfn(Object *obj) u = FIELD_DP32(u, ID_MMFR4, XNX, 1); /* TTS2UXN */ cpu->isar.id_mmfr4 = u; - u = cpu->isar.id_aa64dfr0; - u = FIELD_DP64(u, ID_AA64DFR0, PMUVER, 5); /* v8.4-PMU */ - cpu->isar.id_aa64dfr0 = u; + t = cpu->isar.id_aa64dfr0; + t = FIELD_DP64(t, ID_AA64DFR0, PMUVER, 5); /* v8.4-PMU */ + cpu->isar.id_aa64dfr0 = t; u = cpu->isar.id_dfr0; u = FIELD_DP32(u, ID_DFR0, PERFMON, 5); /* v8.4-PMU */ -- 2.20.1 From: Philippe Mathieu-Daudé <f4bug@amsat.org> MIDR_EL1 is a 64-bit system register with the top 32-bit being RES0. Represent it in QEMU's ARMCPU struct with a uint64_t, not a uint32_t. This fixes an error when compiling with -Werror=conversion because we were manipulating the register value using a local uint64_t variable: target/arm/cpu64.c: In function ‘aarch64_max_initfn’: target/arm/cpu64.c:628:21: error: conversion from ‘uint64_t’ {aka ‘long unsigned int’} to ‘uint32_t’ {aka ‘unsigned int’} may change value [-Werror=conversion] 628 | cpu->midr = t; | ^ and future-proofs us against a possible future architecture change using some of the top 32 bits. Suggested-by: Laurent Desnogues <laurent.desnogues@gmail.com> Suggested-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Laurent Desnogues <laurent.desnogues@gmail.com> Message-id: 20200428172634.29707-1-f4bug@amsat.org Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/cpu.h | 2 +- target/arm/cpu.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/target/arm/cpu.h b/target/arm/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.h +++ b/target/arm/cpu.h @@ -XXX,XX +XXX,XX @@ struct ARMCPU { uint64_t id_aa64dfr0; uint64_t id_aa64dfr1; } isar; - uint32_t midr; + uint64_t midr; uint32_t revidr; uint32_t reset_fpsid; uint32_t ctr; diff --git a/target/arm/cpu.c b/target/arm/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.c +++ b/target/arm/cpu.c @@ -XXX,XX +XXX,XX @@ static const ARMCPUInfo arm_cpus[] = { static Property arm_cpu_properties[] = { DEFINE_PROP_BOOL("start-powered-off", ARMCPU, start_powered_off, false), DEFINE_PROP_UINT32("psci-conduit", ARMCPU, psci_conduit, 0), - DEFINE_PROP_UINT32("midr", ARMCPU, midr, 0), + DEFINE_PROP_UINT64("midr", ARMCPU, midr, 0), DEFINE_PROP_UINT64("mp-affinity", ARMCPU, mp_affinity, ARM64_AFFINITY_INVALID), DEFINE_PROP_INT32("node-id", ARMCPU, node_id, CPU_UNSET_NUMA_NODE_ID), -- 2.20.1 From: "Edgar E. Iglesias" <edgar.iglesias@xilinx.com> Remove inclusion of arm_gicv3_common.h, this already gets included via xlnx-versal.h. Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Alistair Francis <alistair.francis@wdc.com> Reviewed-by: Luc Michel <luc.michel@greensocs.com> Message-id: 20200427181649.26851-2-edgar.iglesias@gmail.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- hw/arm/xlnx-versal.c | 1 - 1 file changed, 1 deletion(-) diff --git a/hw/arm/xlnx-versal.c b/hw/arm/xlnx-versal.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/xlnx-versal.c +++ b/hw/arm/xlnx-versal.c @@ -XXX,XX +XXX,XX @@ #include "hw/arm/boot.h" #include "kvm_arm.h" #include "hw/misc/unimp.h" -#include "hw/intc/arm_gicv3_common.h" #include "hw/arm/xlnx-versal.h" #include "hw/char/pl011.h" -- 2.20.1 From: "Edgar E. Iglesias" <edgar.iglesias@xilinx.com> Move misplaced comment. Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Alistair Francis <alistair.francis@wdc.com> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Luc Michel <luc.michel@greensocs.com> Message-id: 20200427181649.26851-3-edgar.iglesias@gmail.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- hw/arm/xlnx-versal.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/hw/arm/xlnx-versal.c b/hw/arm/xlnx-versal.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/xlnx-versal.c +++ b/hw/arm/xlnx-versal.c @@ -XXX,XX +XXX,XX @@ static void versal_create_apu_cpus(Versal *s) obj = object_new(XLNX_VERSAL_ACPU_TYPE); if (!obj) { - /* Secondary CPUs start in PSCI powered-down state */ error_report("Unable to create apu.cpu[%d] of type %s", i, XLNX_VERSAL_ACPU_TYPE); exit(EXIT_FAILURE); @@ -XXX,XX +XXX,XX @@ static void versal_create_apu_cpus(Versal *s) object_property_set_int(obj, s->cfg.psci_conduit, "psci-conduit", &error_abort); if (i) { + /* Secondary CPUs start in PSCI powered-down state */ object_property_set_bool(obj, true, "start-powered-off", &error_abort); } -- 2.20.1 From: "Edgar E. Iglesias" <edgar.iglesias@xilinx.com> Fix typo xlnx-ve -> xlnx-versal. Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Alistair Francis <alistair.francis@wdc.com> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Luc Michel <luc.michel@greensocs.com> Message-id: 20200427181649.26851-4-edgar.iglesias@gmail.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- hw/arm/xlnx-versal-virt.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/hw/arm/xlnx-versal-virt.c b/hw/arm/xlnx-versal-virt.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/xlnx-versal-virt.c +++ b/hw/arm/xlnx-versal-virt.c @@ -XXX,XX +XXX,XX @@ static void versal_virt_init(MachineState *machine) psci_conduit = QEMU_PSCI_CONDUIT_SMC; } - sysbus_init_child_obj(OBJECT(machine), "xlnx-ve", &s->soc, + sysbus_init_child_obj(OBJECT(machine), "xlnx-versal", &s->soc, sizeof(s->soc), TYPE_XLNX_VERSAL); object_property_set_link(OBJECT(&s->soc), OBJECT(machine->ram), "ddr", &error_abort); -- 2.20.1 From: "Edgar E. Iglesias" <edgar.iglesias@xilinx.com> Embed the UARTs into the SoC type. Suggested-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Alistair Francis <alistair.francis@wdc.com> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Luc Michel <luc.michel@greensocs.com> Message-id: 20200427181649.26851-5-edgar.iglesias@gmail.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- include/hw/arm/xlnx-versal.h | 3 ++- hw/arm/xlnx-versal.c | 12 ++++++------ 2 files changed, 8 insertions(+), 7 deletions(-) diff --git a/include/hw/arm/xlnx-versal.h b/include/hw/arm/xlnx-versal.h index XXXXXXX..XXXXXXX 100644 --- a/include/hw/arm/xlnx-versal.h +++ b/include/hw/arm/xlnx-versal.h @@ -XXX,XX +XXX,XX @@ #include "hw/sysbus.h" #include "hw/arm/boot.h" #include "hw/intc/arm_gicv3.h" +#include "hw/char/pl011.h" #define TYPE_XLNX_VERSAL "xlnx-versal" #define XLNX_VERSAL(obj) OBJECT_CHECK(Versal, (obj), TYPE_XLNX_VERSAL) @@ -XXX,XX +XXX,XX @@ typedef struct Versal { MemoryRegion mr_ocm; struct { - SysBusDevice *uart[XLNX_VERSAL_NR_UARTS]; + PL011State uart[XLNX_VERSAL_NR_UARTS]; SysBusDevice *gem[XLNX_VERSAL_NR_GEMS]; SysBusDevice *adma[XLNX_VERSAL_NR_ADMAS]; } iou; diff --git a/hw/arm/xlnx-versal.c b/hw/arm/xlnx-versal.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/xlnx-versal.c +++ b/hw/arm/xlnx-versal.c @@ -XXX,XX +XXX,XX @@ #include "kvm_arm.h" #include "hw/misc/unimp.h" #include "hw/arm/xlnx-versal.h" -#include "hw/char/pl011.h" #define XLNX_VERSAL_ACPU_TYPE ARM_CPU_TYPE_NAME("cortex-a72") #define GEM_REVISION 0x40070106 @@ -XXX,XX +XXX,XX @@ static void versal_create_uarts(Versal *s, qemu_irq *pic) DeviceState *dev; MemoryRegion *mr; - dev = qdev_create(NULL, TYPE_PL011); - s->lpd.iou.uart[i] = SYS_BUS_DEVICE(dev); + sysbus_init_child_obj(OBJECT(s), name, + &s->lpd.iou.uart[i], sizeof(s->lpd.iou.uart[i]), + TYPE_PL011); + dev = DEVICE(&s->lpd.iou.uart[i]); qdev_prop_set_chr(dev, "chardev", serial_hd(i)); - object_property_add_child(OBJECT(s), name, OBJECT(dev), &error_fatal); qdev_init_nofail(dev); - mr = sysbus_mmio_get_region(s->lpd.iou.uart[i], 0); + mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0); memory_region_add_subregion(&s->mr_ps, addrs[i], mr); - sysbus_connect_irq(s->lpd.iou.uart[i], 0, pic[irqs[i]]); + sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, pic[irqs[i]]); g_free(name); } } -- 2.20.1 From: "Edgar E. Iglesias" <edgar.iglesias@xilinx.com> Embed the GEMs into the SoC type. Suggested-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Alistair Francis <alistair.francis@wdc.com> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Luc Michel <luc.michel@greensocs.com> Message-id: 20200427181649.26851-6-edgar.iglesias@gmail.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- include/hw/arm/xlnx-versal.h | 3 ++- hw/arm/xlnx-versal.c | 15 ++++++++------- 2 files changed, 10 insertions(+), 8 deletions(-) diff --git a/include/hw/arm/xlnx-versal.h b/include/hw/arm/xlnx-versal.h index XXXXXXX..XXXXXXX 100644 --- a/include/hw/arm/xlnx-versal.h +++ b/include/hw/arm/xlnx-versal.h @@ -XXX,XX +XXX,XX @@ #include "hw/arm/boot.h" #include "hw/intc/arm_gicv3.h" #include "hw/char/pl011.h" +#include "hw/net/cadence_gem.h" #define TYPE_XLNX_VERSAL "xlnx-versal" #define XLNX_VERSAL(obj) OBJECT_CHECK(Versal, (obj), TYPE_XLNX_VERSAL) @@ -XXX,XX +XXX,XX @@ typedef struct Versal { struct { PL011State uart[XLNX_VERSAL_NR_UARTS]; - SysBusDevice *gem[XLNX_VERSAL_NR_GEMS]; + CadenceGEMState gem[XLNX_VERSAL_NR_GEMS]; SysBusDevice *adma[XLNX_VERSAL_NR_ADMAS]; } iou; } lpd; diff --git a/hw/arm/xlnx-versal.c b/hw/arm/xlnx-versal.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/xlnx-versal.c +++ b/hw/arm/xlnx-versal.c @@ -XXX,XX +XXX,XX @@ static void versal_create_gems(Versal *s, qemu_irq *pic) DeviceState *dev; MemoryRegion *mr; - dev = qdev_create(NULL, "cadence_gem"); - s->lpd.iou.gem[i] = SYS_BUS_DEVICE(dev); - object_property_add_child(OBJECT(s), name, OBJECT(dev), &error_fatal); + sysbus_init_child_obj(OBJECT(s), name, + &s->lpd.iou.gem[i], sizeof(s->lpd.iou.gem[i]), + TYPE_CADENCE_GEM); + dev = DEVICE(&s->lpd.iou.gem[i]); if (nd->used) { qemu_check_nic_model(nd, "cadence_gem"); qdev_set_nic_properties(dev, nd); } - object_property_set_int(OBJECT(s->lpd.iou.gem[i]), + object_property_set_int(OBJECT(dev), 2, "num-priority-queues", &error_abort); - object_property_set_link(OBJECT(s->lpd.iou.gem[i]), + object_property_set_link(OBJECT(dev), OBJECT(&s->mr_ps), "dma", &error_abort); qdev_init_nofail(dev); - mr = sysbus_mmio_get_region(s->lpd.iou.gem[i], 0); + mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0); memory_region_add_subregion(&s->mr_ps, addrs[i], mr); - sysbus_connect_irq(s->lpd.iou.gem[i], 0, pic[irqs[i]]); + sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, pic[irqs[i]]); g_free(name); } } -- 2.20.1 From: "Edgar E. Iglesias" <edgar.iglesias@xilinx.com> Embed the ADMAs into the SoC type. Suggested-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Alistair Francis <alistair.francis@wdc.com> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Luc Michel <luc.michel@greensocs.com> Message-id: 20200427181649.26851-7-edgar.iglesias@gmail.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- include/hw/arm/xlnx-versal.h | 3 ++- hw/arm/xlnx-versal.c | 14 +++++++------- 2 files changed, 9 insertions(+), 8 deletions(-) diff --git a/include/hw/arm/xlnx-versal.h b/include/hw/arm/xlnx-versal.h index XXXXXXX..XXXXXXX 100644 --- a/include/hw/arm/xlnx-versal.h +++ b/include/hw/arm/xlnx-versal.h @@ -XXX,XX +XXX,XX @@ #include "hw/arm/boot.h" #include "hw/intc/arm_gicv3.h" #include "hw/char/pl011.h" +#include "hw/dma/xlnx-zdma.h" #include "hw/net/cadence_gem.h" #define TYPE_XLNX_VERSAL "xlnx-versal" @@ -XXX,XX +XXX,XX @@ typedef struct Versal { struct { PL011State uart[XLNX_VERSAL_NR_UARTS]; CadenceGEMState gem[XLNX_VERSAL_NR_GEMS]; - SysBusDevice *adma[XLNX_VERSAL_NR_ADMAS]; + XlnxZDMA adma[XLNX_VERSAL_NR_ADMAS]; } iou; } lpd; diff --git a/hw/arm/xlnx-versal.c b/hw/arm/xlnx-versal.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/xlnx-versal.c +++ b/hw/arm/xlnx-versal.c @@ -XXX,XX +XXX,XX @@ static void versal_create_admas(Versal *s, qemu_irq *pic) DeviceState *dev; MemoryRegion *mr; - dev = qdev_create(NULL, "xlnx.zdma"); - s->lpd.iou.adma[i] = SYS_BUS_DEVICE(dev); - object_property_set_int(OBJECT(s->lpd.iou.adma[i]), 128, "bus-width", - &error_abort); - object_property_add_child(OBJECT(s), name, OBJECT(dev), &error_fatal); + sysbus_init_child_obj(OBJECT(s), name, + &s->lpd.iou.adma[i], sizeof(s->lpd.iou.adma[i]), + TYPE_XLNX_ZDMA); + dev = DEVICE(&s->lpd.iou.adma[i]); + object_property_set_int(OBJECT(dev), 128, "bus-width", &error_abort); qdev_init_nofail(dev); - mr = sysbus_mmio_get_region(s->lpd.iou.adma[i], 0); + mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0); memory_region_add_subregion(&s->mr_ps, MM_ADMA_CH0 + i * MM_ADMA_CH0_SIZE, mr); - sysbus_connect_irq(s->lpd.iou.adma[i], 0, pic[VERSAL_ADMA_IRQ_0 + i]); + sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, pic[VERSAL_ADMA_IRQ_0 + i]); g_free(name); } } -- 2.20.1 From: "Edgar E. Iglesias" <edgar.iglesias@xilinx.com> Embed the APUs into the SoC type. Suggested-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Alistair Francis <alistair.francis@wdc.com> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Luc Michel <luc.michel@greensocs.com> Message-id: 20200427181649.26851-8-edgar.iglesias@gmail.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- include/hw/arm/xlnx-versal.h | 2 +- hw/arm/xlnx-versal-virt.c | 4 ++-- hw/arm/xlnx-versal.c | 19 +++++-------------- 3 files changed, 8 insertions(+), 17 deletions(-) diff --git a/include/hw/arm/xlnx-versal.h b/include/hw/arm/xlnx-versal.h index XXXXXXX..XXXXXXX 100644 --- a/include/hw/arm/xlnx-versal.h +++ b/include/hw/arm/xlnx-versal.h @@ -XXX,XX +XXX,XX @@ typedef struct Versal { struct { struct { MemoryRegion mr; - ARMCPU *cpu[XLNX_VERSAL_NR_ACPUS]; + ARMCPU cpu[XLNX_VERSAL_NR_ACPUS]; GICv3State gic; } apu; } fpd; diff --git a/hw/arm/xlnx-versal-virt.c b/hw/arm/xlnx-versal-virt.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/xlnx-versal-virt.c +++ b/hw/arm/xlnx-versal-virt.c @@ -XXX,XX +XXX,XX @@ static void versal_virt_init(MachineState *machine) s->binfo.get_dtb = versal_virt_get_dtb; s->binfo.modify_dtb = versal_virt_modify_dtb; if (machine->kernel_filename) { - arm_load_kernel(s->soc.fpd.apu.cpu[0], machine, &s->binfo); + arm_load_kernel(&s->soc.fpd.apu.cpu[0], machine, &s->binfo); } else { - AddressSpace *as = arm_boot_address_space(s->soc.fpd.apu.cpu[0], + AddressSpace *as = arm_boot_address_space(&s->soc.fpd.apu.cpu[0], &s->binfo); /* Some boot-loaders (e.g u-boot) don't like blobs at address 0 (NULL). * Offset things by 4K. */ diff --git a/hw/arm/xlnx-versal.c b/hw/arm/xlnx-versal.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/xlnx-versal.c +++ b/hw/arm/xlnx-versal.c @@ -XXX,XX +XXX,XX @@ static void versal_create_apu_cpus(Versal *s) for (i = 0; i < ARRAY_SIZE(s->fpd.apu.cpu); i++) { Object *obj; - char *name; - - obj = object_new(XLNX_VERSAL_ACPU_TYPE); - if (!obj) { - error_report("Unable to create apu.cpu[%d] of type %s", - i, XLNX_VERSAL_ACPU_TYPE); - exit(EXIT_FAILURE); - } - - name = g_strdup_printf("apu-cpu[%d]", i); - object_property_add_child(OBJECT(s), name, obj, &error_fatal); - g_free(name); + object_initialize_child(OBJECT(s), "apu-cpu[*]", + &s->fpd.apu.cpu[i], sizeof(s->fpd.apu.cpu[i]), + XLNX_VERSAL_ACPU_TYPE, &error_abort, NULL); + obj = OBJECT(&s->fpd.apu.cpu[i]); object_property_set_int(obj, s->cfg.psci_conduit, "psci-conduit", &error_abort); if (i) { @@ -XXX,XX +XXX,XX @@ static void versal_create_apu_cpus(Versal *s) object_property_set_link(obj, OBJECT(&s->fpd.apu.mr), "memory", &error_abort); object_property_set_bool(obj, true, "realized", &error_fatal); - s->fpd.apu.cpu[i] = ARM_CPU(obj); } } @@ -XXX,XX +XXX,XX @@ static void versal_create_apu_gic(Versal *s, qemu_irq *pic) } for (i = 0; i < nr_apu_cpus; i++) { - DeviceState *cpudev = DEVICE(s->fpd.apu.cpu[i]); + DeviceState *cpudev = DEVICE(&s->fpd.apu.cpu[i]); int ppibase = XLNX_VERSAL_NR_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS; qemu_irq maint_irq; int ti; -- 2.20.1 From: "Edgar E. Iglesias" <edgar.iglesias@xilinx.com> Add support for SD. Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Alistair Francis <alistair.francis@wdc.com> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Luc Michel <luc.michel@greensocs.com> Message-id: 20200427181649.26851-9-edgar.iglesias@gmail.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- include/hw/arm/xlnx-versal.h | 12 ++++++++++++ hw/arm/xlnx-versal.c | 31 +++++++++++++++++++++++++++++++ 2 files changed, 43 insertions(+) diff --git a/include/hw/arm/xlnx-versal.h b/include/hw/arm/xlnx-versal.h index XXXXXXX..XXXXXXX 100644 --- a/include/hw/arm/xlnx-versal.h +++ b/include/hw/arm/xlnx-versal.h @@ -XXX,XX +XXX,XX @@ #include "hw/sysbus.h" #include "hw/arm/boot.h" +#include "hw/sd/sdhci.h" #include "hw/intc/arm_gicv3.h" #include "hw/char/pl011.h" #include "hw/dma/xlnx-zdma.h" @@ -XXX,XX +XXX,XX @@ #define XLNX_VERSAL_NR_UARTS 2 #define XLNX_VERSAL_NR_GEMS 2 #define XLNX_VERSAL_NR_ADMAS 8 +#define XLNX_VERSAL_NR_SDS 2 #define XLNX_VERSAL_NR_IRQS 192 typedef struct Versal { @@ -XXX,XX +XXX,XX @@ typedef struct Versal { } iou; } lpd; + /* The Platform Management Controller subsystem. */ + struct { + struct { + SDHCIState sd[XLNX_VERSAL_NR_SDS]; + } iou; + } pmc; + struct { MemoryRegion *mr_ddr; uint32_t psci_conduit; @@ -XXX,XX +XXX,XX @@ typedef struct Versal { #define VERSAL_GEM1_IRQ_0 58 #define VERSAL_GEM1_WAKE_IRQ_0 59 #define VERSAL_ADMA_IRQ_0 60 +#define VERSAL_SD0_IRQ_0 126 /* Architecturally reserved IRQs suitable for virtualization. */ #define VERSAL_RSVD_IRQ_FIRST 111 @@ -XXX,XX +XXX,XX @@ typedef struct Versal { #define MM_FPD_CRF 0xfd1a0000U #define MM_FPD_CRF_SIZE 0x140000 +#define MM_PMC_SD0 0xf1040000U +#define MM_PMC_SD0_SIZE 0x10000 #define MM_PMC_CRP 0xf1260000U #define MM_PMC_CRP_SIZE 0x10000 #endif diff --git a/hw/arm/xlnx-versal.c b/hw/arm/xlnx-versal.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/xlnx-versal.c +++ b/hw/arm/xlnx-versal.c @@ -XXX,XX +XXX,XX @@ static void versal_create_admas(Versal *s, qemu_irq *pic) } } +#define SDHCI_CAPABILITIES 0x280737ec6481 /* Same as on ZynqMP. */ +static void versal_create_sds(Versal *s, qemu_irq *pic) +{ + int i; + + for (i = 0; i < ARRAY_SIZE(s->pmc.iou.sd); i++) { + DeviceState *dev; + MemoryRegion *mr; + + sysbus_init_child_obj(OBJECT(s), "sd[*]", + &s->pmc.iou.sd[i], sizeof(s->pmc.iou.sd[i]), + TYPE_SYSBUS_SDHCI); + dev = DEVICE(&s->pmc.iou.sd[i]); + + object_property_set_uint(OBJECT(dev), + 3, "sd-spec-version", &error_fatal); + object_property_set_uint(OBJECT(dev), SDHCI_CAPABILITIES, "capareg", + &error_fatal); + object_property_set_uint(OBJECT(dev), UHS_I, "uhs", &error_fatal); + qdev_init_nofail(dev); + + mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0); + memory_region_add_subregion(&s->mr_ps, + MM_PMC_SD0 + i * MM_PMC_SD0_SIZE, mr); + + sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, + pic[VERSAL_SD0_IRQ_0 + i * 2]); + } +} + /* This takes the board allocated linear DDR memory and creates aliases * for each split DDR range/aperture on the Versal address map. */ @@ -XXX,XX +XXX,XX @@ static void versal_realize(DeviceState *dev, Error **errp) versal_create_uarts(s, pic); versal_create_gems(s, pic); versal_create_admas(s, pic); + versal_create_sds(s, pic); versal_map_ddr(s); versal_unimp(s); -- 2.20.1 From: "Edgar E. Iglesias" <edgar.iglesias@xilinx.com> hw/arm: versal: Add support for the RTC. Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Alistair Francis <alistair.francis@wdc.com> Reviewed-by: Luc Michel <luc.michel@greensocs.com> Message-id: 20200427181649.26851-10-edgar.iglesias@gmail.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- include/hw/arm/xlnx-versal.h | 8 ++++++++ hw/arm/xlnx-versal.c | 21 +++++++++++++++++++++ 2 files changed, 29 insertions(+) diff --git a/include/hw/arm/xlnx-versal.h b/include/hw/arm/xlnx-versal.h index XXXXXXX..XXXXXXX 100644 --- a/include/hw/arm/xlnx-versal.h +++ b/include/hw/arm/xlnx-versal.h @@ -XXX,XX +XXX,XX @@ #include "hw/char/pl011.h" #include "hw/dma/xlnx-zdma.h" #include "hw/net/cadence_gem.h" +#include "hw/rtc/xlnx-zynqmp-rtc.h" #define TYPE_XLNX_VERSAL "xlnx-versal" #define XLNX_VERSAL(obj) OBJECT_CHECK(Versal, (obj), TYPE_XLNX_VERSAL) @@ -XXX,XX +XXX,XX @@ typedef struct Versal { struct { SDHCIState sd[XLNX_VERSAL_NR_SDS]; } iou; + + XlnxZynqMPRTC rtc; } pmc; struct { @@ -XXX,XX +XXX,XX @@ typedef struct Versal { #define VERSAL_GEM1_IRQ_0 58 #define VERSAL_GEM1_WAKE_IRQ_0 59 #define VERSAL_ADMA_IRQ_0 60 +#define VERSAL_RTC_APB_ERR_IRQ 121 #define VERSAL_SD0_IRQ_0 126 +#define VERSAL_RTC_ALARM_IRQ 142 +#define VERSAL_RTC_SECONDS_IRQ 143 /* Architecturally reserved IRQs suitable for virtualization. */ #define VERSAL_RSVD_IRQ_FIRST 111 @@ -XXX,XX +XXX,XX @@ typedef struct Versal { #define MM_PMC_SD0_SIZE 0x10000 #define MM_PMC_CRP 0xf1260000U #define MM_PMC_CRP_SIZE 0x10000 +#define MM_PMC_RTC 0xf12a0000 +#define MM_PMC_RTC_SIZE 0x10000 #endif diff --git a/hw/arm/xlnx-versal.c b/hw/arm/xlnx-versal.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/xlnx-versal.c +++ b/hw/arm/xlnx-versal.c @@ -XXX,XX +XXX,XX @@ static void versal_create_sds(Versal *s, qemu_irq *pic) } } +static void versal_create_rtc(Versal *s, qemu_irq *pic) +{ + SysBusDevice *sbd; + MemoryRegion *mr; + + sysbus_init_child_obj(OBJECT(s), "rtc", &s->pmc.rtc, sizeof(s->pmc.rtc), + TYPE_XLNX_ZYNQMP_RTC); + sbd = SYS_BUS_DEVICE(&s->pmc.rtc); + qdev_init_nofail(DEVICE(sbd)); + + mr = sysbus_mmio_get_region(sbd, 0); + memory_region_add_subregion(&s->mr_ps, MM_PMC_RTC, mr); + + /* + * TODO: Connect the ALARM and SECONDS interrupts once our RTC model + * supports them. + */ + sysbus_connect_irq(sbd, 1, pic[VERSAL_RTC_APB_ERR_IRQ]); +} + /* This takes the board allocated linear DDR memory and creates aliases * for each split DDR range/aperture on the Versal address map. */ @@ -XXX,XX +XXX,XX @@ static void versal_realize(DeviceState *dev, Error **errp) versal_create_gems(s, pic); versal_create_admas(s, pic); versal_create_sds(s, pic); + versal_create_rtc(s, pic); versal_map_ddr(s); versal_unimp(s); -- 2.20.1 From: "Edgar E. Iglesias" <edgar.iglesias@xilinx.com> Add support for SD. Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Alistair Francis <alistair.francis@wdc.com> Reviewed-by: Luc Michel <luc.michel@greensocs.com> Message-id: 20200427181649.26851-11-edgar.iglesias@gmail.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- hw/arm/xlnx-versal-virt.c | 46 +++++++++++++++++++++++++++++++++++++++ 1 file changed, 46 insertions(+) diff --git a/hw/arm/xlnx-versal-virt.c b/hw/arm/xlnx-versal-virt.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/xlnx-versal-virt.c +++ b/hw/arm/xlnx-versal-virt.c @@ -XXX,XX +XXX,XX @@ #include "hw/arm/sysbus-fdt.h" #include "hw/arm/fdt.h" #include "cpu.h" +#include "hw/qdev-properties.h" #include "hw/arm/xlnx-versal.h" #define TYPE_XLNX_VERSAL_VIRT_MACHINE MACHINE_TYPE_NAME("xlnx-versal-virt") @@ -XXX,XX +XXX,XX @@ static void fdt_add_zdma_nodes(VersalVirt *s) } } +static void fdt_add_sd_nodes(VersalVirt *s) +{ + const char clocknames[] = "clk_xin\0clk_ahb"; + const char compat[] = "arasan,sdhci-8.9a"; + int i; + + for (i = ARRAY_SIZE(s->soc.pmc.iou.sd) - 1; i >= 0; i--) { + uint64_t addr = MM_PMC_SD0 + MM_PMC_SD0_SIZE * i; + char *name = g_strdup_printf("/sdhci@%" PRIx64, addr); + + qemu_fdt_add_subnode(s->fdt, name); + + qemu_fdt_setprop_cells(s->fdt, name, "clocks", + s->phandle.clk_25Mhz, s->phandle.clk_25Mhz); + qemu_fdt_setprop(s->fdt, name, "clock-names", + clocknames, sizeof(clocknames)); + qemu_fdt_setprop_cells(s->fdt, name, "interrupts", + GIC_FDT_IRQ_TYPE_SPI, VERSAL_SD0_IRQ_0 + i * 2, + GIC_FDT_IRQ_FLAGS_LEVEL_HI); + qemu_fdt_setprop_sized_cells(s->fdt, name, "reg", + 2, addr, 2, MM_PMC_SD0_SIZE); + qemu_fdt_setprop(s->fdt, name, "compatible", compat, sizeof(compat)); + g_free(name); + } +} + static void fdt_nop_memory_nodes(void *fdt, Error **errp) { Error *err = NULL; @@ -XXX,XX +XXX,XX @@ static void create_virtio_regions(VersalVirt *s) } } +static void sd_plugin_card(SDHCIState *sd, DriveInfo *di) +{ + BlockBackend *blk = di ? blk_by_legacy_dinfo(di) : NULL; + DeviceState *card; + + card = qdev_create(qdev_get_child_bus(DEVICE(sd), "sd-bus"), TYPE_SD_CARD); + object_property_add_child(OBJECT(sd), "card[*]", OBJECT(card), + &error_fatal); + qdev_prop_set_drive(card, "drive", blk, &error_fatal); + object_property_set_bool(OBJECT(card), true, "realized", &error_fatal); +} + static void versal_virt_init(MachineState *machine) { VersalVirt *s = XLNX_VERSAL_VIRT_MACHINE(machine); int psci_conduit = QEMU_PSCI_CONDUIT_DISABLED; + int i; /* * If the user provides an Operating System to be loaded, we expect them @@ -XXX,XX +XXX,XX @@ static void versal_virt_init(MachineState *machine) fdt_add_gic_nodes(s); fdt_add_timer_nodes(s); fdt_add_zdma_nodes(s); + fdt_add_sd_nodes(s); fdt_add_cpu_nodes(s, psci_conduit); fdt_add_clk_node(s, "/clk125", 125000000, s->phandle.clk_125Mhz); fdt_add_clk_node(s, "/clk25", 25000000, s->phandle.clk_25Mhz); @@ -XXX,XX +XXX,XX @@ static void versal_virt_init(MachineState *machine) memory_region_add_subregion_overlap(get_system_memory(), 0, &s->soc.fpd.apu.mr, 0); + /* Plugin SD cards. */ + for (i = 0; i < ARRAY_SIZE(s->soc.pmc.iou.sd); i++) { + sd_plugin_card(&s->soc.pmc.iou.sd[i], drive_get_next(IF_SD)); + } + s->binfo.ram_size = machine->ram_size; s->binfo.loader_start = 0x0; s->binfo.get_dtb = versal_virt_get_dtb; -- 2.20.1 From: "Edgar E. Iglesias" <edgar.iglesias@xilinx.com> Add support for the RTC. Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Alistair Francis <alistair.francis@wdc.com> Reviewed-by: Luc Michel <luc.michel@greensocs.com> Message-id: 20200427181649.26851-12-edgar.iglesias@gmail.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- hw/arm/xlnx-versal-virt.c | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+) diff --git a/hw/arm/xlnx-versal-virt.c b/hw/arm/xlnx-versal-virt.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/xlnx-versal-virt.c +++ b/hw/arm/xlnx-versal-virt.c @@ -XXX,XX +XXX,XX @@ static void fdt_add_sd_nodes(VersalVirt *s) } } +static void fdt_add_rtc_node(VersalVirt *s) +{ + const char compat[] = "xlnx,zynqmp-rtc"; + const char interrupt_names[] = "alarm\0sec"; + char *name = g_strdup_printf("/rtc@%x", MM_PMC_RTC); + + qemu_fdt_add_subnode(s->fdt, name); + + qemu_fdt_setprop_cells(s->fdt, name, "interrupts", + GIC_FDT_IRQ_TYPE_SPI, VERSAL_RTC_ALARM_IRQ, + GIC_FDT_IRQ_FLAGS_LEVEL_HI, + GIC_FDT_IRQ_TYPE_SPI, VERSAL_RTC_SECONDS_IRQ, + GIC_FDT_IRQ_FLAGS_LEVEL_HI); + qemu_fdt_setprop(s->fdt, name, "interrupt-names", + interrupt_names, sizeof(interrupt_names)); + qemu_fdt_setprop_sized_cells(s->fdt, name, "reg", + 2, MM_PMC_RTC, 2, MM_PMC_RTC_SIZE); + qemu_fdt_setprop(s->fdt, name, "compatible", compat, sizeof(compat)); + g_free(name); +} + static void fdt_nop_memory_nodes(void *fdt, Error **errp) { Error *err = NULL; @@ -XXX,XX +XXX,XX @@ static void versal_virt_init(MachineState *machine) fdt_add_timer_nodes(s); fdt_add_zdma_nodes(s); fdt_add_sd_nodes(s); + fdt_add_rtc_node(s); fdt_add_cpu_nodes(s, psci_conduit); fdt_add_clk_node(s, "/clk125", 125000000, s->phandle.clk_125Mhz); fdt_add_clk_node(s, "/clk25", 25000000, s->phandle.clk_25Mhz); -- 2.20.1 Somewhere along theline we accidentally added a duplicate "using D16-D31 when they don't exist" check to do_vfm_dp() (probably an artifact of a patchseries rebase). Remove it. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Message-id: 20200430181003.21682-2-peter.maydell@linaro.org --- target/arm/translate-vfp.inc.c | 6 ------ 1 file changed, 6 deletions(-) diff --git a/target/arm/translate-vfp.inc.c b/target/arm/translate-vfp.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-vfp.inc.c +++ b/target/arm/translate-vfp.inc.c @@ -XXX,XX +XXX,XX @@ static bool do_vfm_dp(DisasContext *s, arg_VFMA_dp *a, bool neg_n, bool neg_d) return false; } - /* UNDEF accesses to D16-D31 if they don't exist. */ - if (!dc_isar_feature(aa32_simd_r32, s) && - ((a->vd | a->vn | a->vm) & 0x10)) { - return false; - } - if (!vfp_access_check(s)) { return true; } -- 2.20.1 We were accidentally permitting decode of Thumb Neon insns even if the CPU didn't have the FEATURE_NEON bit set, because the feature check was being done before the call to disas_neon_data_insn() and disas_neon_ls_insn() in the Arm decoder but was omitted from the Thumb decoder. Push the feature bit check down into the called functions so it is done for both Arm and Thumb encodings. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Message-id: 20200430181003.21682-3-peter.maydell@linaro.org --- target/arm/translate.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static int disas_neon_ls_insn(DisasContext *s, uint32_t insn) TCGv_i32 tmp2; TCGv_i64 tmp64; + if (!arm_dc_feature(s, ARM_FEATURE_NEON)) { + return 1; + } + /* FIXME: this access check should not take precedence over UNDEF * for invalid encodings; we will generate incorrect syndrome information * for attempts to execute invalid vfp/neon encodings with FP disabled. @@ -XXX,XX +XXX,XX @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) TCGv_ptr ptr1, ptr2, ptr3; TCGv_i64 tmp64; + if (!arm_dc_feature(s, ARM_FEATURE_NEON)) { + return 1; + } + /* FIXME: this access check should not take precedence over UNDEF * for invalid encodings; we will generate incorrect syndrome information * for attempts to execute invalid vfp/neon encodings with FP disabled. @@ -XXX,XX +XXX,XX @@ static void disas_arm_insn(DisasContext *s, unsigned int insn) if (((insn >> 25) & 7) == 1) { /* NEON Data processing. */ - if (!arm_dc_feature(s, ARM_FEATURE_NEON)) { - goto illegal_op; - } - if (disas_neon_data_insn(s, insn)) { goto illegal_op; } @@ -XXX,XX +XXX,XX @@ static void disas_arm_insn(DisasContext *s, unsigned int insn) } if ((insn & 0x0f100000) == 0x04000000) { /* NEON load/store. */ - if (!arm_dc_feature(s, ARM_FEATURE_NEON)) { - goto illegal_op; - } - if (disas_neon_ls_insn(s, insn)) { goto illegal_op; } -- 2.20.1 Add the infrastructure for building and invoking a decodetree decoder for the AArch32 Neon encodings. At the moment the new decoder covers nothing, so we always fall back to the existing hand-written decode. We follow the same pattern we did for the VFP decodetree conversion (commit 78e138bc1f672c145ef6ace74617d and following): code that deals with Neon will be moving gradually out to translate-neon.vfp.inc, which we #include into translate.c. In order to share the decode files between A32 and T32, we split Neon into 3 parts: * data-processing * load-store * 'shared' encodings The first two groups of instructions have similar but not identical A32 and T32 encodings, so we need to manually transform the T32 encoding into the A32 one before calling the decoder; the third group covers the Neon instructions which are identical in A32 and T32. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-4-peter.maydell@linaro.org --- target/arm/neon-dp.decode | 29 ++++++++++++++++++++++++++ target/arm/neon-ls.decode | 29 ++++++++++++++++++++++++++ target/arm/neon-shared.decode | 27 +++++++++++++++++++++++++ target/arm/translate-neon.inc.c | 32 +++++++++++++++++++++++++++++ target/arm/translate.c | 36 +++++++++++++++++++++++++++++++-- target/arm/Makefile.objs | 18 +++++++++++++++++ 6 files changed, 169 insertions(+), 2 deletions(-) create mode 100644 target/arm/neon-dp.decode create mode 100644 target/arm/neon-ls.decode create mode 100644 target/arm/neon-shared.decode create mode 100644 target/arm/translate-neon.inc.c diff --git a/target/arm/neon-dp.decode b/target/arm/neon-dp.decode new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/target/arm/neon-dp.decode @@ -XXX,XX +XXX,XX @@ +# AArch32 Neon data-processing instruction descriptions +# +# Copyright (c) 2020 Linaro, Ltd +# +# This library is free software; you can redistribute it and/or +# modify it under the terms of the GNU Lesser General Public +# License as published by the Free Software Foundation; either +# version 2 of the License, or (at your option) any later version. +# +# This library is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +# Lesser General Public License for more details. +# +# You should have received a copy of the GNU Lesser General Public +# License along with this library; if not, see <http://www.gnu.org/licenses/>. + +# +# This file is processed by scripts/decodetree.py +# + +# Encodings for Neon data processing instructions where the T32 encoding +# is a simple transformation of the A32 encoding. +# More specifically, this file covers instructions where the A32 encoding is +# 0b1111_001p_qqqq_qqqq_qqqq_qqqq_qqqq_qqqq +# and the T32 encoding is +# 0b111p_1111_qqqq_qqqq_qqqq_qqqq_qqqq_qqqq +# This file works on the A32 encoding only; calling code for T32 has to +# transform the insn into the A32 version first. diff --git a/target/arm/neon-ls.decode b/target/arm/neon-ls.decode new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/target/arm/neon-ls.decode @@ -XXX,XX +XXX,XX @@ +# AArch32 Neon load/store instruction descriptions +# +# Copyright (c) 2020 Linaro, Ltd +# +# This library is free software; you can redistribute it and/or +# modify it under the terms of the GNU Lesser General Public +# License as published by the Free Software Foundation; either +# version 2 of the License, or (at your option) any later version. +# +# This library is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +# Lesser General Public License for more details. +# +# You should have received a copy of the GNU Lesser General Public +# License along with this library; if not, see <http://www.gnu.org/licenses/>. + +# +# This file is processed by scripts/decodetree.py +# + +# Encodings for Neon load/store instructions where the T32 encoding +# is a simple transformation of the A32 encoding. +# More specifically, this file covers instructions where the A32 encoding is +# 0b1111_0100_xxx0_xxxx_xxxx_xxxx_xxxx_xxxx +# and the T32 encoding is +# 0b1111_1001_xxx0_xxxx_xxxx_xxxx_xxxx_xxxx +# This file works on the A32 encoding only; calling code for T32 has to +# transform the insn into the A32 version first. diff --git a/target/arm/neon-shared.decode b/target/arm/neon-shared.decode new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/target/arm/neon-shared.decode @@ -XXX,XX +XXX,XX @@ +# AArch32 Neon instruction descriptions +# +# Copyright (c) 2020 Linaro, Ltd +# +# This library is free software; you can redistribute it and/or +# modify it under the terms of the GNU Lesser General Public +# License as published by the Free Software Foundation; either +# version 2 of the License, or (at your option) any later version. +# +# This library is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +# Lesser General Public License for more details. +# +# You should have received a copy of the GNU Lesser General Public +# License along with this library; if not, see <http://www.gnu.org/licenses/>. + +# +# This file is processed by scripts/decodetree.py +# + +# Encodings for Neon instructions whose encoding is the same for +# both A32 and T32. + +# More specifically, this covers: +# 2reg scalar ext: 0b1111_1110_xxxx_xxxx_xxxx_1x0x_xxxx_xxxx +# 3same ext: 0b1111_110x_xxxx_xxxx_xxxx_1x0x_xxxx_xxxx diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ +/* + * ARM translation: AArch32 Neon instructions + * + * Copyright (c) 2003 Fabrice Bellard + * Copyright (c) 2005-2007 CodeSourcery + * Copyright (c) 2007 OpenedHand, Ltd. + * Copyright (c) 2020 Linaro, Ltd. + * + * This library is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2 of the License, or (at your option) any later version. + * + * This library is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with this library; if not, see <http://www.gnu.org/licenses/>. + */ + +/* + * This file is intended to be included from translate.c; it uses + * some macros and definitions provided by that file. + * It might be possible to convert it to a standalone .c file eventually. + */ + +/* Include the generated Neon decoder */ +#include "decode-neon-dp.inc.c" +#include "decode-neon-ls.inc.c" +#include "decode-neon-shared.inc.c" diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static TCGv_ptr vfp_reg_ptr(bool dp, int reg) #define ARM_CP_RW_BIT (1 << 20) -/* Include the VFP decoder */ +/* Include the VFP and Neon decoders */ #include "translate-vfp.inc.c" +#include "translate-neon.inc.c" static inline void iwmmxt_load_reg(TCGv_i64 var, int reg) { @@ -XXX,XX +XXX,XX @@ static void disas_arm_insn(DisasContext *s, unsigned int insn) /* Unconditional instructions. */ /* TODO: Perhaps merge these into one decodetree output file. */ if (disas_a32_uncond(s, insn) || - disas_vfp_uncond(s, insn)) { + disas_vfp_uncond(s, insn) || + disas_neon_dp(s, insn) || + disas_neon_ls(s, insn) || + disas_neon_shared(s, insn)) { return; } /* fall back to legacy decoder */ @@ -XXX,XX +XXX,XX @@ static void disas_thumb2_insn(DisasContext *s, uint32_t insn) ARCH(6T2); } + if ((insn & 0xef000000) == 0xef000000) { + /* + * T32 encodings 0b111p_1111_qqqq_qqqq_qqqq_qqqq_qqqq_qqqq + * transform into + * A32 encodings 0b1111_001p_qqqq_qqqq_qqqq_qqqq_qqqq_qqqq + */ + uint32_t a32_insn = (insn & 0xe2ffffff) | + ((insn & (1 << 28)) >> 4) | (1 << 28); + + if (disas_neon_dp(s, a32_insn)) { + return; + } + } + + if ((insn & 0xff100000) == 0xf9000000) { + /* + * T32 encodings 0b1111_1001_ppp0_qqqq_qqqq_qqqq_qqqq_qqqq + * transform into + * A32 encodings 0b1111_0100_ppp0_qqqq_qqqq_qqqq_qqqq_qqqq + */ + uint32_t a32_insn = (insn & 0x00ffffff) | 0xf4000000; + + if (disas_neon_ls(s, a32_insn)) { + return; + } + } + /* * TODO: Perhaps merge these into one decodetree output file. * Note disas_vfp is written for a32 with cond field in the @@ -XXX,XX +XXX,XX @@ static void disas_thumb2_insn(DisasContext *s, uint32_t insn) */ if (disas_t32(s, insn) || disas_vfp_uncond(s, insn) || + disas_neon_shared(s, insn) || ((insn >> 28) == 0xe && disas_vfp(s, insn))) { return; } diff --git a/target/arm/Makefile.objs b/target/arm/Makefile.objs index XXXXXXX..XXXXXXX 100644 --- a/target/arm/Makefile.objs +++ b/target/arm/Makefile.objs @@ -XXX,XX +XXX,XX @@ target/arm/decode-sve.inc.c: $(SRC_PATH)/target/arm/sve.decode $(DECODETREE) $(PYTHON) $(DECODETREE) --decode disas_sve -o $@ $<,\ "GEN", $(TARGET_DIR)$@) +target/arm/decode-neon-shared.inc.c: $(SRC_PATH)/target/arm/neon-shared.decode $(DECODETREE) + $(call quiet-command,\ + $(PYTHON) $(DECODETREE) --static-decode disas_neon_shared -o $@ $<,\ + "GEN", $(TARGET_DIR)$@) + +target/arm/decode-neon-dp.inc.c: $(SRC_PATH)/target/arm/neon-dp.decode $(DECODETREE) + $(call quiet-command,\ + $(PYTHON) $(DECODETREE) --static-decode disas_neon_dp -o $@ $<,\ + "GEN", $(TARGET_DIR)$@) + +target/arm/decode-neon-ls.inc.c: $(SRC_PATH)/target/arm/neon-ls.decode $(DECODETREE) + $(call quiet-command,\ + $(PYTHON) $(DECODETREE) --static-decode disas_neon_ls -o $@ $<,\ + "GEN", $(TARGET_DIR)$@) + target/arm/decode-vfp.inc.c: $(SRC_PATH)/target/arm/vfp.decode $(DECODETREE) $(call quiet-command,\ $(PYTHON) $(DECODETREE) --static-decode disas_vfp -o $@ $<,\ @@ -XXX,XX +XXX,XX @@ target/arm/decode-t16.inc.c: $(SRC_PATH)/target/arm/t16.decode $(DECODETREE) "GEN", $(TARGET_DIR)$@) target/arm/translate-sve.o: target/arm/decode-sve.inc.c +target/arm/translate.o: target/arm/decode-neon-shared.inc.c +target/arm/translate.o: target/arm/decode-neon-dp.inc.c +target/arm/translate.o: target/arm/decode-neon-ls.inc.c target/arm/translate.o: target/arm/decode-vfp.inc.c target/arm/translate.o: target/arm/decode-vfp-uncond.inc.c target/arm/translate.o: target/arm/decode-a32.inc.c -- 2.20.1 Convert the VCMLA (vector) insns in the 3same extension group to decodetree. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-5-peter.maydell@linaro.org --- target/arm/neon-shared.decode | 11 ++++++++++ target/arm/translate-neon.inc.c | 37 +++++++++++++++++++++++++++++++++ target/arm/translate.c | 11 +--------- 3 files changed, 49 insertions(+), 10 deletions(-) diff --git a/target/arm/neon-shared.decode b/target/arm/neon-shared.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-shared.decode +++ b/target/arm/neon-shared.decode @@ -XXX,XX +XXX,XX @@ # More specifically, this covers: # 2reg scalar ext: 0b1111_1110_xxxx_xxxx_xxxx_1x0x_xxxx_xxxx # 3same ext: 0b1111_110x_xxxx_xxxx_xxxx_1x0x_xxxx_xxxx + +# VFP/Neon register fields; same as vfp.decode +%vm_dp 5:1 0:4 +%vm_sp 0:4 5:1 +%vn_dp 7:1 16:4 +%vn_sp 16:4 7:1 +%vd_dp 22:1 12:4 +%vd_sp 12:4 22:1 + +VCMLA 1111 110 rot:2 . 1 size:1 .... .... 1000 . q:1 . 0 .... \ + vm=%vm_dp vn=%vn_dp vd=%vd_dp diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ #include "decode-neon-dp.inc.c" #include "decode-neon-ls.inc.c" #include "decode-neon-shared.inc.c" + +static bool trans_VCMLA(DisasContext *s, arg_VCMLA *a) +{ + int opr_sz; + TCGv_ptr fpst; + gen_helper_gvec_3_ptr *fn_gvec_ptr; + + if (!dc_isar_feature(aa32_vcma, s) + || (!a->size && !dc_isar_feature(aa32_fp16_arith, s))) { + return false; + } + + /* UNDEF accesses to D16-D31 if they don't exist. */ + if (!dc_isar_feature(aa32_simd_r32, s) && + ((a->vd | a->vn | a->vm) & 0x10)) { + return false; + } + + if ((a->vn | a->vm | a->vd) & a->q) { + return false; + } + + if (!vfp_access_check(s)) { + return true; + } + + opr_sz = (1 + a->q) * 8; + fpst = get_fpstatus_ptr(1); + fn_gvec_ptr = a->size ? gen_helper_gvec_fcmlas : gen_helper_gvec_fcmlah; + tcg_gen_gvec_3_ptr(vfp_reg_offset(1, a->vd), + vfp_reg_offset(1, a->vn), + vfp_reg_offset(1, a->vm), + fpst, opr_sz, opr_sz, a->rot, + fn_gvec_ptr); + tcg_temp_free_ptr(fpst); + return true; +} diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static int disas_neon_insn_3same_ext(DisasContext *s, uint32_t insn) bool is_long = false, q = extract32(insn, 6, 1); bool ptr_is_env = false; - if ((insn & 0xfe200f10) == 0xfc200800) { - /* VCMLA -- 1111 110R R.1S .... .... 1000 ...0 .... */ - int size = extract32(insn, 20, 1); - data = extract32(insn, 23, 2); /* rot */ - if (!dc_isar_feature(aa32_vcma, s) - || (!size && !dc_isar_feature(aa32_fp16_arith, s))) { - return 1; - } - fn_gvec_ptr = size ? gen_helper_gvec_fcmlas : gen_helper_gvec_fcmlah; - } else if ((insn & 0xfea00f10) == 0xfc800800) { + if ((insn & 0xfea00f10) == 0xfc800800) { /* VCADD -- 1111 110R 1.0S .... .... 1000 ...0 .... */ int size = extract32(insn, 20, 1); data = extract32(insn, 24, 1); /* rot */ -- 2.20.1 Convert the VCADD (vector) insns to decodetree. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-6-peter.maydell@linaro.org --- target/arm/neon-shared.decode | 3 +++ target/arm/translate-neon.inc.c | 37 +++++++++++++++++++++++++++++++++ target/arm/translate.c | 11 +--------- 3 files changed, 41 insertions(+), 10 deletions(-) diff --git a/target/arm/neon-shared.decode b/target/arm/neon-shared.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-shared.decode +++ b/target/arm/neon-shared.decode @@ -XXX,XX +XXX,XX @@ VCMLA 1111 110 rot:2 . 1 size:1 .... .... 1000 . q:1 . 0 .... \ vm=%vm_dp vn=%vn_dp vd=%vd_dp + +VCADD 1111 110 rot:1 1 . 0 size:1 .... .... 1000 . q:1 . 0 .... \ + vm=%vm_dp vn=%vn_dp vd=%vd_dp diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ static bool trans_VCMLA(DisasContext *s, arg_VCMLA *a) tcg_temp_free_ptr(fpst); return true; } + +static bool trans_VCADD(DisasContext *s, arg_VCADD *a) +{ + int opr_sz; + TCGv_ptr fpst; + gen_helper_gvec_3_ptr *fn_gvec_ptr; + + if (!dc_isar_feature(aa32_vcma, s) + || (!a->size && !dc_isar_feature(aa32_fp16_arith, s))) { + return false; + } + + /* UNDEF accesses to D16-D31 if they don't exist. */ + if (!dc_isar_feature(aa32_simd_r32, s) && + ((a->vd | a->vn | a->vm) & 0x10)) { + return false; + } + + if ((a->vn | a->vm | a->vd) & a->q) { + return false; + } + + if (!vfp_access_check(s)) { + return true; + } + + opr_sz = (1 + a->q) * 8; + fpst = get_fpstatus_ptr(1); + fn_gvec_ptr = a->size ? gen_helper_gvec_fcadds : gen_helper_gvec_fcaddh; + tcg_gen_gvec_3_ptr(vfp_reg_offset(1, a->vd), + vfp_reg_offset(1, a->vn), + vfp_reg_offset(1, a->vm), + fpst, opr_sz, opr_sz, a->rot, + fn_gvec_ptr); + tcg_temp_free_ptr(fpst); + return true; +} diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static int disas_neon_insn_3same_ext(DisasContext *s, uint32_t insn) bool is_long = false, q = extract32(insn, 6, 1); bool ptr_is_env = false; - if ((insn & 0xfea00f10) == 0xfc800800) { - /* VCADD -- 1111 110R 1.0S .... .... 1000 ...0 .... */ - int size = extract32(insn, 20, 1); - data = extract32(insn, 24, 1); /* rot */ - if (!dc_isar_feature(aa32_vcma, s) - || (!size && !dc_isar_feature(aa32_fp16_arith, s))) { - return 1; - } - fn_gvec_ptr = size ? gen_helper_gvec_fcadds : gen_helper_gvec_fcaddh; - } else if ((insn & 0xfeb00f00) == 0xfc200d00) { + if ((insn & 0xfeb00f00) == 0xfc200d00) { /* V[US]DOT -- 1111 1100 0.10 .... .... 1101 .Q.U .... */ bool u = extract32(insn, 4, 1); if (!dc_isar_feature(aa32_dp, s)) { -- 2.20.1 Convert the V[US]DOT (vector) insns to decodetree. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-7-peter.maydell@linaro.org --- target/arm/neon-shared.decode | 4 ++++ target/arm/translate-neon.inc.c | 32 ++++++++++++++++++++++++++++++++ target/arm/translate.c | 9 +-------- 3 files changed, 37 insertions(+), 8 deletions(-) diff --git a/target/arm/neon-shared.decode b/target/arm/neon-shared.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-shared.decode +++ b/target/arm/neon-shared.decode @@ -XXX,XX +XXX,XX @@ VCMLA 1111 110 rot:2 . 1 size:1 .... .... 1000 . q:1 . 0 .... \ VCADD 1111 110 rot:1 1 . 0 size:1 .... .... 1000 . q:1 . 0 .... \ vm=%vm_dp vn=%vn_dp vd=%vd_dp + +# VUDOT and VSDOT +VDOT 1111 110 00 . 10 .... .... 1101 . q:1 . u:1 .... \ + vm=%vm_dp vn=%vn_dp vd=%vd_dp diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ static bool trans_VCADD(DisasContext *s, arg_VCADD *a) tcg_temp_free_ptr(fpst); return true; } + +static bool trans_VDOT(DisasContext *s, arg_VDOT *a) +{ + int opr_sz; + gen_helper_gvec_3 *fn_gvec; + + if (!dc_isar_feature(aa32_dp, s)) { + return false; + } + + /* UNDEF accesses to D16-D31 if they don't exist. */ + if (!dc_isar_feature(aa32_simd_r32, s) && + ((a->vd | a->vn | a->vm) & 0x10)) { + return false; + } + + if ((a->vn | a->vm | a->vd) & a->q) { + return false; + } + + if (!vfp_access_check(s)) { + return true; + } + + opr_sz = (1 + a->q) * 8; + fn_gvec = a->u ? gen_helper_gvec_udot_b : gen_helper_gvec_sdot_b; + tcg_gen_gvec_3_ool(vfp_reg_offset(1, a->vd), + vfp_reg_offset(1, a->vn), + vfp_reg_offset(1, a->vm), + opr_sz, opr_sz, 0, fn_gvec); + return true; +} diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static int disas_neon_insn_3same_ext(DisasContext *s, uint32_t insn) bool is_long = false, q = extract32(insn, 6, 1); bool ptr_is_env = false; - if ((insn & 0xfeb00f00) == 0xfc200d00) { - /* V[US]DOT -- 1111 1100 0.10 .... .... 1101 .Q.U .... */ - bool u = extract32(insn, 4, 1); - if (!dc_isar_feature(aa32_dp, s)) { - return 1; - } - fn_gvec = u ? gen_helper_gvec_udot_b : gen_helper_gvec_sdot_b; - } else if ((insn & 0xff300f10) == 0xfc200810) { + if ((insn & 0xff300f10) == 0xfc200810) { /* VFM[AS]L -- 1111 1100 S.10 .... .... 1000 .Q.1 .... */ int is_s = extract32(insn, 23, 1); if (!dc_isar_feature(aa32_fhm, s)) { -- 2.20.1 Convert the VFM[AS]L (vector) insns to decodetree. This is the last insn in the legacy decoder for the 3same_ext group, so we can delete the legacy decoder function for the group entirely. Note that in disas_thumb2_insn() the parts of this encoding space where the decodetree decoder returns false will correctly be directed to illegal_op by the "(insn & (1 << 28))" check so they won't fall into disas_coproc_insn() by mistake. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-8-peter.maydell@linaro.org --- target/arm/neon-shared.decode | 6 +++ target/arm/translate-neon.inc.c | 31 +++++++++++ target/arm/translate.c | 92 +-------------------------------- 3 files changed, 38 insertions(+), 91 deletions(-) diff --git a/target/arm/neon-shared.decode b/target/arm/neon-shared.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-shared.decode +++ b/target/arm/neon-shared.decode @@ -XXX,XX +XXX,XX @@ VCADD 1111 110 rot:1 1 . 0 size:1 .... .... 1000 . q:1 . 0 .... \ # VUDOT and VSDOT VDOT 1111 110 00 . 10 .... .... 1101 . q:1 . u:1 .... \ vm=%vm_dp vn=%vn_dp vd=%vd_dp + +# VFM[AS]L +VFML 1111 110 0 s:1 . 10 .... .... 1000 . 0 . 1 .... \ + vm=%vm_sp vn=%vn_sp vd=%vd_dp q=0 +VFML 1111 110 0 s:1 . 10 .... .... 1000 . 1 . 1 .... \ + vm=%vm_dp vn=%vn_dp vd=%vd_dp q=1 diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ static bool trans_VDOT(DisasContext *s, arg_VDOT *a) opr_sz, opr_sz, 0, fn_gvec); return true; } + +static bool trans_VFML(DisasContext *s, arg_VFML *a) +{ + int opr_sz; + + if (!dc_isar_feature(aa32_fhm, s)) { + return false; + } + + /* UNDEF accesses to D16-D31 if they don't exist. */ + if (!dc_isar_feature(aa32_simd_r32, s) && + (a->vd & 0x10)) { + return false; + } + + if (a->vd & a->q) { + return false; + } + + if (!vfp_access_check(s)) { + return true; + } + + opr_sz = (1 + a->q) * 8; + tcg_gen_gvec_3_ptr(vfp_reg_offset(1, a->vd), + vfp_reg_offset(a->q, a->vn), + vfp_reg_offset(a->q, a->vm), + cpu_env, opr_sz, opr_sz, a->s, /* is_2 == 0 */ + gen_helper_gvec_fmlal_a32); + return true; +} diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) return 0; } -/* Advanced SIMD three registers of the same length extension. - * 31 25 23 22 20 16 12 11 10 9 8 3 0 - * +---------------+-----+---+-----+----+----+---+----+---+----+---------+----+ - * | 1 1 1 1 1 1 0 | op1 | D | op2 | Vn | Vd | 1 | o3 | 0 | o4 | N Q M U | Vm | - * +---------------+-----+---+-----+----+----+---+----+---+----+---------+----+ - */ -static int disas_neon_insn_3same_ext(DisasContext *s, uint32_t insn) -{ - gen_helper_gvec_3 *fn_gvec = NULL; - gen_helper_gvec_3_ptr *fn_gvec_ptr = NULL; - int rd, rn, rm, opr_sz; - int data = 0; - int off_rn, off_rm; - bool is_long = false, q = extract32(insn, 6, 1); - bool ptr_is_env = false; - - if ((insn & 0xff300f10) == 0xfc200810) { - /* VFM[AS]L -- 1111 1100 S.10 .... .... 1000 .Q.1 .... */ - int is_s = extract32(insn, 23, 1); - if (!dc_isar_feature(aa32_fhm, s)) { - return 1; - } - is_long = true; - data = is_s; /* is_2 == 0 */ - fn_gvec_ptr = gen_helper_gvec_fmlal_a32; - ptr_is_env = true; - } else { - return 1; - } - - VFP_DREG_D(rd, insn); - if (rd & q) { - return 1; - } - if (q || !is_long) { - VFP_DREG_N(rn, insn); - VFP_DREG_M(rm, insn); - if ((rn | rm) & q & !is_long) { - return 1; - } - off_rn = vfp_reg_offset(1, rn); - off_rm = vfp_reg_offset(1, rm); - } else { - rn = VFP_SREG_N(insn); - rm = VFP_SREG_M(insn); - off_rn = vfp_reg_offset(0, rn); - off_rm = vfp_reg_offset(0, rm); - } - - if (s->fp_excp_el) { - gen_exception_insn(s, s->pc_curr, EXCP_UDEF, - syn_simd_access_trap(1, 0xe, false), s->fp_excp_el); - return 0; - } - if (!s->vfp_enabled) { - return 1; - } - - opr_sz = (1 + q) * 8; - if (fn_gvec_ptr) { - TCGv_ptr ptr; - if (ptr_is_env) { - ptr = cpu_env; - } else { - ptr = get_fpstatus_ptr(1); - } - tcg_gen_gvec_3_ptr(vfp_reg_offset(1, rd), off_rn, off_rm, ptr, - opr_sz, opr_sz, data, fn_gvec_ptr); - if (!ptr_is_env) { - tcg_temp_free_ptr(ptr); - } - } else { - tcg_gen_gvec_3_ool(vfp_reg_offset(1, rd), off_rn, off_rm, - opr_sz, opr_sz, data, fn_gvec); - } - return 0; -} - /* Advanced SIMD two registers and a scalar extension. * 31 24 23 22 20 16 12 11 10 9 8 3 0 * +-----------------+----+---+----+----+----+---+----+---+----+---------+----+ @@ -XXX,XX +XXX,XX @@ static void disas_arm_insn(DisasContext *s, unsigned int insn) } } } - } else if ((insn & 0x0e000a00) == 0x0c000800 - && arm_dc_feature(s, ARM_FEATURE_V8)) { - if (disas_neon_insn_3same_ext(s, insn)) { - goto illegal_op; - } - return; } else if ((insn & 0x0f000a00) == 0x0e000800 && arm_dc_feature(s, ARM_FEATURE_V8)) { if (disas_neon_insn_2reg_scalar_ext(s, insn)) { @@ -XXX,XX +XXX,XX @@ static void disas_thumb2_insn(DisasContext *s, uint32_t insn) } break; } - if ((insn & 0xfe000a00) == 0xfc000800 + if ((insn & 0xff000a00) == 0xfe000800 && arm_dc_feature(s, ARM_FEATURE_V8)) { /* The Thumb2 and ARM encodings are identical. */ - if (disas_neon_insn_3same_ext(s, insn)) { - goto illegal_op; - } - } else if ((insn & 0xff000a00) == 0xfe000800 - && arm_dc_feature(s, ARM_FEATURE_V8)) { - /* The Thumb2 and ARM encodings are identical. */ if (disas_neon_insn_2reg_scalar_ext(s, insn)) { goto illegal_op; } -- 2.20.1 Convert VCMLA (scalar) in the 2reg-scalar-ext group to decodetree. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-9-peter.maydell@linaro.org --- target/arm/neon-shared.decode | 5 +++++ target/arm/translate-neon.inc.c | 40 +++++++++++++++++++++++++++++++++ target/arm/translate.c | 26 +-------------------- 3 files changed, 46 insertions(+), 25 deletions(-) diff --git a/target/arm/neon-shared.decode b/target/arm/neon-shared.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-shared.decode +++ b/target/arm/neon-shared.decode @@ -XXX,XX +XXX,XX @@ VFML 1111 110 0 s:1 . 10 .... .... 1000 . 0 . 1 .... \ vm=%vm_sp vn=%vn_sp vd=%vd_dp q=0 VFML 1111 110 0 s:1 . 10 .... .... 1000 . 1 . 1 .... \ vm=%vm_dp vn=%vn_dp vd=%vd_dp q=1 + +VCMLA_scalar 1111 1110 0 . rot:2 .... .... 1000 . q:1 index:1 0 vm:4 \ + vn=%vn_dp vd=%vd_dp size=0 +VCMLA_scalar 1111 1110 1 . rot:2 .... .... 1000 . q:1 . 0 .... \ + vm=%vm_dp vn=%vn_dp vd=%vd_dp size=1 index=0 diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ static bool trans_VFML(DisasContext *s, arg_VFML *a) gen_helper_gvec_fmlal_a32); return true; } + +static bool trans_VCMLA_scalar(DisasContext *s, arg_VCMLA_scalar *a) +{ + gen_helper_gvec_3_ptr *fn_gvec_ptr; + int opr_sz; + TCGv_ptr fpst; + + if (!dc_isar_feature(aa32_vcma, s)) { + return false; + } + if (a->size == 0 && !dc_isar_feature(aa32_fp16_arith, s)) { + return false; + } + + /* UNDEF accesses to D16-D31 if they don't exist. */ + if (!dc_isar_feature(aa32_simd_r32, s) && + ((a->vd | a->vn | a->vm) & 0x10)) { + return false; + } + + if ((a->vd | a->vn) & a->q) { + return false; + } + + if (!vfp_access_check(s)) { + return true; + } + + fn_gvec_ptr = (a->size ? gen_helper_gvec_fcmlas_idx + : gen_helper_gvec_fcmlah_idx); + opr_sz = (1 + a->q) * 8; + fpst = get_fpstatus_ptr(1); + tcg_gen_gvec_3_ptr(vfp_reg_offset(1, a->vd), + vfp_reg_offset(1, a->vn), + vfp_reg_offset(1, a->vm), + fpst, opr_sz, opr_sz, + (a->index << 2) | a->rot, fn_gvec_ptr); + tcg_temp_free_ptr(fpst); + return true; +} diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static int disas_neon_insn_2reg_scalar_ext(DisasContext *s, uint32_t insn) bool is_long = false, q = extract32(insn, 6, 1); bool ptr_is_env = false; - if ((insn & 0xff000f10) == 0xfe000800) { - /* VCMLA (indexed) -- 1111 1110 S.RR .... .... 1000 ...0 .... */ - int rot = extract32(insn, 20, 2); - int size = extract32(insn, 23, 1); - int index; - - if (!dc_isar_feature(aa32_vcma, s)) { - return 1; - } - if (size == 0) { - if (!dc_isar_feature(aa32_fp16_arith, s)) { - return 1; - } - /* For fp16, rm is just Vm, and index is M. */ - rm = extract32(insn, 0, 4); - index = extract32(insn, 5, 1); - } else { - /* For fp32, rm is the usual M:Vm, and index is 0. */ - VFP_DREG_M(rm, insn); - index = 0; - } - data = (index << 2) | rot; - fn_gvec_ptr = (size ? gen_helper_gvec_fcmlas_idx - : gen_helper_gvec_fcmlah_idx); - } else if ((insn & 0xffb00f00) == 0xfe200d00) { + if ((insn & 0xffb00f00) == 0xfe200d00) { /* V[US]DOT -- 1111 1110 0.10 .... .... 1101 .Q.U .... */ int u = extract32(insn, 4, 1); -- 2.20.1 Convert the V[US]DOT (scalar) insns in the 2reg-scalar-ext group to decodetree. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-10-peter.maydell@linaro.org --- target/arm/neon-shared.decode | 3 +++ target/arm/translate-neon.inc.c | 35 +++++++++++++++++++++++++++++++++ target/arm/translate.c | 13 +----------- 3 files changed, 39 insertions(+), 12 deletions(-) diff --git a/target/arm/neon-shared.decode b/target/arm/neon-shared.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-shared.decode +++ b/target/arm/neon-shared.decode @@ -XXX,XX +XXX,XX @@ VCMLA_scalar 1111 1110 0 . rot:2 .... .... 1000 . q:1 index:1 0 vm:4 \ vn=%vn_dp vd=%vd_dp size=0 VCMLA_scalar 1111 1110 1 . rot:2 .... .... 1000 . q:1 . 0 .... \ vm=%vm_dp vn=%vn_dp vd=%vd_dp size=1 index=0 + +VDOT_scalar 1111 1110 0 . 10 .... .... 1101 . q:1 index:1 u:1 rm:4 \ + vm=%vm_dp vn=%vn_dp vd=%vd_dp diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ static bool trans_VCMLA_scalar(DisasContext *s, arg_VCMLA_scalar *a) tcg_temp_free_ptr(fpst); return true; } + +static bool trans_VDOT_scalar(DisasContext *s, arg_VDOT_scalar *a) +{ + gen_helper_gvec_3 *fn_gvec; + int opr_sz; + TCGv_ptr fpst; + + if (!dc_isar_feature(aa32_dp, s)) { + return false; + } + + /* UNDEF accesses to D16-D31 if they don't exist. */ + if (!dc_isar_feature(aa32_simd_r32, s) && + ((a->vd | a->vn) & 0x10)) { + return false; + } + + if ((a->vd | a->vn) & a->q) { + return false; + } + + if (!vfp_access_check(s)) { + return true; + } + + fn_gvec = a->u ? gen_helper_gvec_udot_idx_b : gen_helper_gvec_sdot_idx_b; + opr_sz = (1 + a->q) * 8; + fpst = get_fpstatus_ptr(1); + tcg_gen_gvec_3_ool(vfp_reg_offset(1, a->vd), + vfp_reg_offset(1, a->vn), + vfp_reg_offset(1, a->rm), + opr_sz, opr_sz, a->index, fn_gvec); + tcg_temp_free_ptr(fpst); + return true; +} diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static int disas_neon_insn_2reg_scalar_ext(DisasContext *s, uint32_t insn) bool is_long = false, q = extract32(insn, 6, 1); bool ptr_is_env = false; - if ((insn & 0xffb00f00) == 0xfe200d00) { - /* V[US]DOT -- 1111 1110 0.10 .... .... 1101 .Q.U .... */ - int u = extract32(insn, 4, 1); - - if (!dc_isar_feature(aa32_dp, s)) { - return 1; - } - fn_gvec = u ? gen_helper_gvec_udot_idx_b : gen_helper_gvec_sdot_idx_b; - /* rm is just Vm, and index is M. */ - data = extract32(insn, 5, 1); /* index */ - rm = extract32(insn, 0, 4); - } else if ((insn & 0xffa00f10) == 0xfe000810) { + if ((insn & 0xffa00f10) == 0xfe000810) { /* VFM[AS]L -- 1111 1110 0.0S .... .... 1000 .Q.1 .... */ int is_s = extract32(insn, 20, 1); int vm20 = extract32(insn, 0, 3); -- 2.20.1 Convert the VFM[AS]L (scalar) insns in the 2reg-scalar-ext group to decodetree. These are the last ones in the group so we can remove all the legacy decode for the group. Note that in disas_thumb2_insn() the parts of this encoding space where the decodetree decoder returns false will correctly be directed to illegal_op by the "(insn & (1 << 28))" check so they won't fall into disas_coproc_insn() by mistake. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-11-peter.maydell@linaro.org --- target/arm/neon-shared.decode | 7 +++ target/arm/translate-neon.inc.c | 32 ++++++++++ target/arm/translate.c | 107 +------------------------------- 3 files changed, 40 insertions(+), 106 deletions(-) diff --git a/target/arm/neon-shared.decode b/target/arm/neon-shared.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-shared.decode +++ b/target/arm/neon-shared.decode @@ -XXX,XX +XXX,XX @@ VCMLA_scalar 1111 1110 1 . rot:2 .... .... 1000 . q:1 . 0 .... \ VDOT_scalar 1111 1110 0 . 10 .... .... 1101 . q:1 index:1 u:1 rm:4 \ vm=%vm_dp vn=%vn_dp vd=%vd_dp + +%vfml_scalar_q0_rm 0:3 5:1 +%vfml_scalar_q1_index 5:1 3:1 +VFML_scalar 1111 1110 0 . 0 s:1 .... .... 1000 . 0 . 1 index:1 ... \ + rm=%vfml_scalar_q0_rm vn=%vn_sp vd=%vd_dp q=0 +VFML_scalar 1111 1110 0 . 0 s:1 .... .... 1000 . 1 . 1 . rm:3 \ + index=%vfml_scalar_q1_index vn=%vn_dp vd=%vd_dp q=1 diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ static bool trans_VDOT_scalar(DisasContext *s, arg_VDOT_scalar *a) tcg_temp_free_ptr(fpst); return true; } + +static bool trans_VFML_scalar(DisasContext *s, arg_VFML_scalar *a) +{ + int opr_sz; + + if (!dc_isar_feature(aa32_fhm, s)) { + return false; + } + + /* UNDEF accesses to D16-D31 if they don't exist. */ + if (!dc_isar_feature(aa32_simd_r32, s) && + ((a->vd & 0x10) || (a->q && (a->vn & 0x10)))) { + return false; + } + + if (a->vd & a->q) { + return false; + } + + if (!vfp_access_check(s)) { + return true; + } + + opr_sz = (1 + a->q) * 8; + tcg_gen_gvec_3_ptr(vfp_reg_offset(1, a->vd), + vfp_reg_offset(a->q, a->vn), + vfp_reg_offset(a->q, a->rm), + cpu_env, opr_sz, opr_sz, + (a->index << 2) | a->s, /* is_2 == 0 */ + gen_helper_gvec_fmlal_idx_a32); + return true; +} diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static int disas_dsp_insn(DisasContext *s, uint32_t insn) } #define VFP_REG_SHR(x, n) (((n) > 0) ? (x) >> (n) : (x) << -(n)) -#define VFP_SREG(insn, bigbit, smallbit) \ - ((VFP_REG_SHR(insn, bigbit - 1) & 0x1e) | (((insn) >> (smallbit)) & 1)) #define VFP_DREG(reg, insn, bigbit, smallbit) do { \ if (dc_isar_feature(aa32_simd_r32, s)) { \ reg = (((insn) >> (bigbit)) & 0x0f) \ @@ -XXX,XX +XXX,XX @@ static int disas_dsp_insn(DisasContext *s, uint32_t insn) reg = ((insn) >> (bigbit)) & 0x0f; \ }} while (0) -#define VFP_SREG_D(insn) VFP_SREG(insn, 12, 22) #define VFP_DREG_D(reg, insn) VFP_DREG(reg, insn, 12, 22) -#define VFP_SREG_N(insn) VFP_SREG(insn, 16, 7) #define VFP_DREG_N(reg, insn) VFP_DREG(reg, insn, 16, 7) -#define VFP_SREG_M(insn) VFP_SREG(insn, 0, 5) #define VFP_DREG_M(reg, insn) VFP_DREG(reg, insn, 0, 5) static void gen_neon_dup_low16(TCGv_i32 var) @@ -XXX,XX +XXX,XX @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) return 0; } -/* Advanced SIMD two registers and a scalar extension. - * 31 24 23 22 20 16 12 11 10 9 8 3 0 - * +-----------------+----+---+----+----+----+---+----+---+----+---------+----+ - * | 1 1 1 1 1 1 1 0 | o1 | D | o2 | Vn | Vd | 1 | o3 | 0 | o4 | N Q M U | Vm | - * +-----------------+----+---+----+----+----+---+----+---+----+---------+----+ - * - */ - -static int disas_neon_insn_2reg_scalar_ext(DisasContext *s, uint32_t insn) -{ - gen_helper_gvec_3 *fn_gvec = NULL; - gen_helper_gvec_3_ptr *fn_gvec_ptr = NULL; - int rd, rn, rm, opr_sz, data; - int off_rn, off_rm; - bool is_long = false, q = extract32(insn, 6, 1); - bool ptr_is_env = false; - - if ((insn & 0xffa00f10) == 0xfe000810) { - /* VFM[AS]L -- 1111 1110 0.0S .... .... 1000 .Q.1 .... */ - int is_s = extract32(insn, 20, 1); - int vm20 = extract32(insn, 0, 3); - int vm3 = extract32(insn, 3, 1); - int m = extract32(insn, 5, 1); - int index; - - if (!dc_isar_feature(aa32_fhm, s)) { - return 1; - } - if (q) { - rm = vm20; - index = m * 2 + vm3; - } else { - rm = vm20 * 2 + m; - index = vm3; - } - is_long = true; - data = (index << 2) | is_s; /* is_2 == 0 */ - fn_gvec_ptr = gen_helper_gvec_fmlal_idx_a32; - ptr_is_env = true; - } else { - return 1; - } - - VFP_DREG_D(rd, insn); - if (rd & q) { - return 1; - } - if (q || !is_long) { - VFP_DREG_N(rn, insn); - if (rn & q & !is_long) { - return 1; - } - off_rn = vfp_reg_offset(1, rn); - off_rm = vfp_reg_offset(1, rm); - } else { - rn = VFP_SREG_N(insn); - off_rn = vfp_reg_offset(0, rn); - off_rm = vfp_reg_offset(0, rm); - } - if (s->fp_excp_el) { - gen_exception_insn(s, s->pc_curr, EXCP_UDEF, - syn_simd_access_trap(1, 0xe, false), s->fp_excp_el); - return 0; - } - if (!s->vfp_enabled) { - return 1; - } - - opr_sz = (1 + q) * 8; - if (fn_gvec_ptr) { - TCGv_ptr ptr; - if (ptr_is_env) { - ptr = cpu_env; - } else { - ptr = get_fpstatus_ptr(1); - } - tcg_gen_gvec_3_ptr(vfp_reg_offset(1, rd), off_rn, off_rm, ptr, - opr_sz, opr_sz, data, fn_gvec_ptr); - if (!ptr_is_env) { - tcg_temp_free_ptr(ptr); - } - } else { - tcg_gen_gvec_3_ool(vfp_reg_offset(1, rd), off_rn, off_rm, - opr_sz, opr_sz, data, fn_gvec); - } - return 0; -} - static int disas_coproc_insn(DisasContext *s, uint32_t insn) { int cpnum, is64, crn, crm, opc1, opc2, isread, rt, rt2; @@ -XXX,XX +XXX,XX @@ static void disas_arm_insn(DisasContext *s, unsigned int insn) } } } - } else if ((insn & 0x0f000a00) == 0x0e000800 - && arm_dc_feature(s, ARM_FEATURE_V8)) { - if (disas_neon_insn_2reg_scalar_ext(s, insn)) { - goto illegal_op; - } - return; } goto illegal_op; } @@ -XXX,XX +XXX,XX @@ static void disas_thumb2_insn(DisasContext *s, uint32_t insn) } break; } - if ((insn & 0xff000a00) == 0xfe000800 - && arm_dc_feature(s, ARM_FEATURE_V8)) { - /* The Thumb2 and ARM encodings are identical. */ - if (disas_neon_insn_2reg_scalar_ext(s, insn)) { - goto illegal_op; - } - } else if (((insn >> 24) & 3) == 3) { + if (((insn >> 24) & 3) == 3) { /* Translate into the equivalent ARM encoding. */ insn = (insn & 0xe2ffffff) | ((insn & (1 << 28)) >> 4) | (1 << 28); if (disas_neon_data_insn(s, insn)) { -- 2.20.1 Convert the Neon "load/store multiple structures" insns to decodetree. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-12-peter.maydell@linaro.org --- target/arm/neon-ls.decode | 7 ++ target/arm/translate-neon.inc.c | 124 ++++++++++++++++++++++++++++++++ target/arm/translate.c | 91 +---------------------- 3 files changed, 133 insertions(+), 89 deletions(-) diff --git a/target/arm/neon-ls.decode b/target/arm/neon-ls.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-ls.decode +++ b/target/arm/neon-ls.decode @@ -XXX,XX +XXX,XX @@ # 0b1111_1001_xxx0_xxxx_xxxx_xxxx_xxxx_xxxx # This file works on the A32 encoding only; calling code for T32 has to # transform the insn into the A32 version first. + +%vd_dp 22:1 12:4 + +# Neon load/store multiple structures + +VLDST_multiple 1111 0100 0 . l:1 0 rn:4 .... itype:4 size:2 align:2 rm:4 \ + vd=%vd_dp diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ static bool trans_VFML_scalar(DisasContext *s, arg_VFML_scalar *a) gen_helper_gvec_fmlal_idx_a32); return true; } + +static struct { + int nregs; + int interleave; + int spacing; +} const neon_ls_element_type[11] = { + {1, 4, 1}, + {1, 4, 2}, + {4, 1, 1}, + {2, 2, 2}, + {1, 3, 1}, + {1, 3, 2}, + {3, 1, 1}, + {1, 1, 1}, + {1, 2, 1}, + {1, 2, 2}, + {2, 1, 1} +}; + +static void gen_neon_ldst_base_update(DisasContext *s, int rm, int rn, + int stride) +{ + if (rm != 15) { + TCGv_i32 base; + + base = load_reg(s, rn); + if (rm == 13) { + tcg_gen_addi_i32(base, base, stride); + } else { + TCGv_i32 index; + index = load_reg(s, rm); + tcg_gen_add_i32(base, base, index); + tcg_temp_free_i32(index); + } + store_reg(s, rn, base); + } +} + +static bool trans_VLDST_multiple(DisasContext *s, arg_VLDST_multiple *a) +{ + /* Neon load/store multiple structures */ + int nregs, interleave, spacing, reg, n; + MemOp endian = s->be_data; + int mmu_idx = get_mem_index(s); + int size = a->size; + TCGv_i64 tmp64; + TCGv_i32 addr, tmp; + + if (!arm_dc_feature(s, ARM_FEATURE_NEON)) { + return false; + } + + /* UNDEF accesses to D16-D31 if they don't exist */ + if (!dc_isar_feature(aa32_simd_r32, s) && (a->vd & 0x10)) { + return false; + } + if (a->itype > 10) { + return false; + } + /* Catch UNDEF cases for bad values of align field */ + switch (a->itype & 0xc) { + case 4: + if (a->align >= 2) { + return false; + } + break; + case 8: + if (a->align == 3) { + return false; + } + break; + default: + break; + } + nregs = neon_ls_element_type[a->itype].nregs; + interleave = neon_ls_element_type[a->itype].interleave; + spacing = neon_ls_element_type[a->itype].spacing; + if (size == 3 && (interleave | spacing) != 1) { + return false; + } + + if (!vfp_access_check(s)) { + return true; + } + + /* For our purposes, bytes are always little-endian. */ + if (size == 0) { + endian = MO_LE; + } + /* + * Consecutive little-endian elements from a single register + * can be promoted to a larger little-endian operation. + */ + if (interleave == 1 && endian == MO_LE) { + size = 3; + } + tmp64 = tcg_temp_new_i64(); + addr = tcg_temp_new_i32(); + tmp = tcg_const_i32(1 << size); + load_reg_var(s, addr, a->rn); + for (reg = 0; reg < nregs; reg++) { + for (n = 0; n < 8 >> size; n++) { + int xs; + for (xs = 0; xs < interleave; xs++) { + int tt = a->vd + reg + spacing * xs; + + if (a->l) { + gen_aa32_ld_i64(s, tmp64, addr, mmu_idx, endian | size); + neon_store_element64(tt, n, size, tmp64); + } else { + neon_load_element64(tmp64, tt, n, size); + gen_aa32_st_i64(s, tmp64, addr, mmu_idx, endian | size); + } + tcg_gen_add_i32(addr, addr, tmp); + } + } + } + tcg_temp_free_i32(addr); + tcg_temp_free_i32(tmp); + tcg_temp_free_i64(tmp64); + + gen_neon_ldst_base_update(s, a->rm, a->rn, nregs * interleave * 8); + return true; +} diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static void gen_neon_trn_u16(TCGv_i32 t0, TCGv_i32 t1) } -static struct { - int nregs; - int interleave; - int spacing; -} const neon_ls_element_type[11] = { - {1, 4, 1}, - {1, 4, 2}, - {4, 1, 1}, - {2, 2, 2}, - {1, 3, 1}, - {1, 3, 2}, - {3, 1, 1}, - {1, 1, 1}, - {1, 2, 1}, - {1, 2, 2}, - {2, 1, 1} -}; - /* Translate a NEON load/store element instruction. Return nonzero if the instruction is invalid. */ static int disas_neon_ls_insn(DisasContext *s, uint32_t insn) { int rd, rn, rm; - int op; int nregs; - int interleave; - int spacing; int stride; int size; int reg; int load; - int n; int vec_size; - int mmu_idx; - MemOp endian; TCGv_i32 addr; TCGv_i32 tmp; - TCGv_i32 tmp2; - TCGv_i64 tmp64; if (!arm_dc_feature(s, ARM_FEATURE_NEON)) { return 1; @@ -XXX,XX +XXX,XX @@ static int disas_neon_ls_insn(DisasContext *s, uint32_t insn) rn = (insn >> 16) & 0xf; rm = insn & 0xf; load = (insn & (1 << 21)) != 0; - endian = s->be_data; - mmu_idx = get_mem_index(s); if ((insn & (1 << 23)) == 0) { - /* Load store all elements. */ - op = (insn >> 8) & 0xf; - size = (insn >> 6) & 3; - if (op > 10) - return 1; - /* Catch UNDEF cases for bad values of align field */ - switch (op & 0xc) { - case 4: - if (((insn >> 5) & 1) == 1) { - return 1; - } - break; - case 8: - if (((insn >> 4) & 3) == 3) { - return 1; - } - break; - default: - break; - } - nregs = neon_ls_element_type[op].nregs; - interleave = neon_ls_element_type[op].interleave; - spacing = neon_ls_element_type[op].spacing; - if (size == 3 && (interleave | spacing) != 1) { - return 1; - } - /* For our purposes, bytes are always little-endian. */ - if (size == 0) { - endian = MO_LE; - } - /* Consecutive little-endian elements from a single register - * can be promoted to a larger little-endian operation. - */ - if (interleave == 1 && endian == MO_LE) { - size = 3; - } - tmp64 = tcg_temp_new_i64(); - addr = tcg_temp_new_i32(); - tmp2 = tcg_const_i32(1 << size); - load_reg_var(s, addr, rn); - for (reg = 0; reg < nregs; reg++) { - for (n = 0; n < 8 >> size; n++) { - int xs; - for (xs = 0; xs < interleave; xs++) { - int tt = rd + reg + spacing * xs; - - if (load) { - gen_aa32_ld_i64(s, tmp64, addr, mmu_idx, endian | size); - neon_store_element64(tt, n, size, tmp64); - } else { - neon_load_element64(tmp64, tt, n, size); - gen_aa32_st_i64(s, tmp64, addr, mmu_idx, endian | size); - } - tcg_gen_add_i32(addr, addr, tmp2); - } - } - } - tcg_temp_free_i32(addr); - tcg_temp_free_i32(tmp2); - tcg_temp_free_i64(tmp64); - stride = nregs * interleave * 8; + /* Load store all elements -- handled already by decodetree */ + return 1; } else { size = (insn >> 10) & 3; if (size == 3) { -- 2.20.1 Convert the Neon "load single structure to all lanes" insns to decodetree. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-13-peter.maydell@linaro.org --- target/arm/neon-ls.decode | 5 +++ target/arm/translate-neon.inc.c | 73 +++++++++++++++++++++++++++++++++ target/arm/translate.c | 55 +------------------------ 3 files changed, 80 insertions(+), 53 deletions(-) diff --git a/target/arm/neon-ls.decode b/target/arm/neon-ls.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-ls.decode +++ b/target/arm/neon-ls.decode @@ -XXX,XX +XXX,XX @@ VLDST_multiple 1111 0100 0 . l:1 0 rn:4 .... itype:4 size:2 align:2 rm:4 \ vd=%vd_dp + +# Neon load single element to all lanes + +VLD_all_lanes 1111 0100 1 . 1 0 rn:4 .... 11 n:2 size:2 t:1 a:1 rm:4 \ + vd=%vd_dp diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ static bool trans_VLDST_multiple(DisasContext *s, arg_VLDST_multiple *a) gen_neon_ldst_base_update(s, a->rm, a->rn, nregs * interleave * 8); return true; } + +static bool trans_VLD_all_lanes(DisasContext *s, arg_VLD_all_lanes *a) +{ + /* Neon load single structure to all lanes */ + int reg, stride, vec_size; + int vd = a->vd; + int size = a->size; + int nregs = a->n + 1; + TCGv_i32 addr, tmp; + + if (!arm_dc_feature(s, ARM_FEATURE_NEON)) { + return false; + } + + /* UNDEF accesses to D16-D31 if they don't exist */ + if (!dc_isar_feature(aa32_simd_r32, s) && (a->vd & 0x10)) { + return false; + } + + if (size == 3) { + if (nregs != 4 || a->a == 0) { + return false; + } + /* For VLD4 size == 3 a == 1 means 32 bits at 16 byte alignment */ + size = 2; + } + if (nregs == 1 && a->a == 1 && size == 0) { + return false; + } + if (nregs == 3 && a->a == 1) { + return false; + } + + if (!vfp_access_check(s)) { + return true; + } + + /* + * VLD1 to all lanes: T bit indicates how many Dregs to write. + * VLD2/3/4 to all lanes: T bit indicates register stride. + */ + stride = a->t ? 2 : 1; + vec_size = nregs == 1 ? stride * 8 : 8; + + tmp = tcg_temp_new_i32(); + addr = tcg_temp_new_i32(); + load_reg_var(s, addr, a->rn); + for (reg = 0; reg < nregs; reg++) { + gen_aa32_ld_i32(s, tmp, addr, get_mem_index(s), + s->be_data | size); + if ((vd & 1) && vec_size == 16) { + /* + * We cannot write 16 bytes at once because the + * destination is unaligned. + */ + tcg_gen_gvec_dup_i32(size, neon_reg_offset(vd, 0), + 8, 8, tmp); + tcg_gen_gvec_mov(0, neon_reg_offset(vd + 1, 0), + neon_reg_offset(vd, 0), 8, 8); + } else { + tcg_gen_gvec_dup_i32(size, neon_reg_offset(vd, 0), + vec_size, vec_size, tmp); + } + tcg_gen_addi_i32(addr, addr, 1 << size); + vd += stride; + } + tcg_temp_free_i32(tmp); + tcg_temp_free_i32(addr); + + gen_neon_ldst_base_update(s, a->rm, a->rn, (1 << size) * nregs); + + return true; +} diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static int disas_neon_ls_insn(DisasContext *s, uint32_t insn) int size; int reg; int load; - int vec_size; TCGv_i32 addr; TCGv_i32 tmp; @@ -XXX,XX +XXX,XX @@ static int disas_neon_ls_insn(DisasContext *s, uint32_t insn) } else { size = (insn >> 10) & 3; if (size == 3) { - /* Load single element to all lanes. */ - int a = (insn >> 4) & 1; - if (!load) { - return 1; - } - size = (insn >> 6) & 3; - nregs = ((insn >> 8) & 3) + 1; - - if (size == 3) { - if (nregs != 4 || a == 0) { - return 1; - } - /* For VLD4 size==3 a == 1 means 32 bits at 16 byte alignment */ - size = 2; - } - if (nregs == 1 && a == 1 && size == 0) { - return 1; - } - if (nregs == 3 && a == 1) { - return 1; - } - addr = tcg_temp_new_i32(); - load_reg_var(s, addr, rn); - - /* VLD1 to all lanes: bit 5 indicates how many Dregs to write. - * VLD2/3/4 to all lanes: bit 5 indicates register stride. - */ - stride = (insn & (1 << 5)) ? 2 : 1; - vec_size = nregs == 1 ? stride * 8 : 8; - - tmp = tcg_temp_new_i32(); - for (reg = 0; reg < nregs; reg++) { - gen_aa32_ld_i32(s, tmp, addr, get_mem_index(s), - s->be_data | size); - if ((rd & 1) && vec_size == 16) { - /* We cannot write 16 bytes at once because the - * destination is unaligned. - */ - tcg_gen_gvec_dup_i32(size, neon_reg_offset(rd, 0), - 8, 8, tmp); - tcg_gen_gvec_mov(0, neon_reg_offset(rd + 1, 0), - neon_reg_offset(rd, 0), 8, 8); - } else { - tcg_gen_gvec_dup_i32(size, neon_reg_offset(rd, 0), - vec_size, vec_size, tmp); - } - tcg_gen_addi_i32(addr, addr, 1 << size); - rd += stride; - } - tcg_temp_free_i32(tmp); - tcg_temp_free_i32(addr); - stride = (1 << size) * nregs; + /* Load single element to all lanes -- handled by decodetree */ + return 1; } else { /* Single element. */ int idx = (insn >> 4) & 0xf; -- 2.20.1 Convert the Neon "load/store single structure to one lane" insns to decodetree. As this is the last set of insns in the neon load/store group, we can remove the whole disas_neon_ls_insn() function. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-14-peter.maydell@linaro.org --- target/arm/neon-ls.decode | 11 +++ target/arm/translate-neon.inc.c | 89 +++++++++++++++++++ target/arm/translate.c | 147 -------------------------------- 3 files changed, 100 insertions(+), 147 deletions(-) diff --git a/target/arm/neon-ls.decode b/target/arm/neon-ls.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-ls.decode +++ b/target/arm/neon-ls.decode @@ -XXX,XX +XXX,XX @@ VLDST_multiple 1111 0100 0 . l:1 0 rn:4 .... itype:4 size:2 align:2 rm:4 \ VLD_all_lanes 1111 0100 1 . 1 0 rn:4 .... 11 n:2 size:2 t:1 a:1 rm:4 \ vd=%vd_dp + +# Neon load/store single structure to one lane +%imm1_5_p1 5:1 !function=plus1 +%imm1_6_p1 6:1 !function=plus1 + +VLDST_single 1111 0100 1 . l:1 0 rn:4 .... 00 n:2 reg_idx:3 align:1 rm:4 \ + vd=%vd_dp size=0 stride=1 +VLDST_single 1111 0100 1 . l:1 0 rn:4 .... 01 n:2 reg_idx:2 align:2 rm:4 \ + vd=%vd_dp size=1 stride=%imm1_5_p1 +VLDST_single 1111 0100 1 . l:1 0 rn:4 .... 10 n:2 reg_idx:1 align:3 rm:4 \ + vd=%vd_dp size=2 stride=%imm1_6_p1 diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ * It might be possible to convert it to a standalone .c file eventually. */ +static inline int plus1(DisasContext *s, int x) +{ + return x + 1; +} + /* Include the generated Neon decoder */ #include "decode-neon-dp.inc.c" #include "decode-neon-ls.inc.c" @@ -XXX,XX +XXX,XX @@ static bool trans_VLD_all_lanes(DisasContext *s, arg_VLD_all_lanes *a) return true; } + +static bool trans_VLDST_single(DisasContext *s, arg_VLDST_single *a) +{ + /* Neon load/store single structure to one lane */ + int reg; + int nregs = a->n + 1; + int vd = a->vd; + TCGv_i32 addr, tmp; + + if (!arm_dc_feature(s, ARM_FEATURE_NEON)) { + return false; + } + + /* UNDEF accesses to D16-D31 if they don't exist */ + if (!dc_isar_feature(aa32_simd_r32, s) && (a->vd & 0x10)) { + return false; + } + + /* Catch the UNDEF cases. This is unavoidably a bit messy. */ + switch (nregs) { + case 1: + if (((a->align & (1 << a->size)) != 0) || + (a->size == 2 && ((a->align & 3) == 1 || (a->align & 3) == 2))) { + return false; + } + break; + case 3: + if ((a->align & 1) != 0) { + return false; + } + /* fall through */ + case 2: + if (a->size == 2 && (a->align & 2) != 0) { + return false; + } + break; + case 4: + if ((a->size == 2) && ((a->align & 3) == 3)) { + return false; + } + break; + default: + abort(); + } + if ((vd + a->stride * (nregs - 1)) > 31) { + /* + * Attempts to write off the end of the register file are + * UNPREDICTABLE; we choose to UNDEF because otherwise we would + * access off the end of the array that holds the register data. + */ + return false; + } + + if (!vfp_access_check(s)) { + return true; + } + + tmp = tcg_temp_new_i32(); + addr = tcg_temp_new_i32(); + load_reg_var(s, addr, a->rn); + /* + * TODO: if we implemented alignment exceptions, we should check + * addr against the alignment encoded in a->align here. + */ + for (reg = 0; reg < nregs; reg++) { + if (a->l) { + gen_aa32_ld_i32(s, tmp, addr, get_mem_index(s), + s->be_data | a->size); + neon_store_element(vd, a->reg_idx, a->size, tmp); + } else { /* Store */ + neon_load_element(tmp, vd, a->reg_idx, a->size); + gen_aa32_st_i32(s, tmp, addr, get_mem_index(s), + s->be_data | a->size); + } + vd += a->stride; + tcg_gen_addi_i32(addr, addr, 1 << a->size); + } + tcg_temp_free_i32(addr); + tcg_temp_free_i32(tmp); + + gen_neon_ldst_base_update(s, a->rm, a->rn, (1 << a->size) * nregs); + + return true; +} diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static void gen_neon_trn_u16(TCGv_i32 t0, TCGv_i32 t1) tcg_temp_free_i32(rd); } - -/* Translate a NEON load/store element instruction. Return nonzero if the - instruction is invalid. */ -static int disas_neon_ls_insn(DisasContext *s, uint32_t insn) -{ - int rd, rn, rm; - int nregs; - int stride; - int size; - int reg; - int load; - TCGv_i32 addr; - TCGv_i32 tmp; - - if (!arm_dc_feature(s, ARM_FEATURE_NEON)) { - return 1; - } - - /* FIXME: this access check should not take precedence over UNDEF - * for invalid encodings; we will generate incorrect syndrome information - * for attempts to execute invalid vfp/neon encodings with FP disabled. - */ - if (s->fp_excp_el) { - gen_exception_insn(s, s->pc_curr, EXCP_UDEF, - syn_simd_access_trap(1, 0xe, false), s->fp_excp_el); - return 0; - } - - if (!s->vfp_enabled) - return 1; - VFP_DREG_D(rd, insn); - rn = (insn >> 16) & 0xf; - rm = insn & 0xf; - load = (insn & (1 << 21)) != 0; - if ((insn & (1 << 23)) == 0) { - /* Load store all elements -- handled already by decodetree */ - return 1; - } else { - size = (insn >> 10) & 3; - if (size == 3) { - /* Load single element to all lanes -- handled by decodetree */ - return 1; - } else { - /* Single element. */ - int idx = (insn >> 4) & 0xf; - int reg_idx; - switch (size) { - case 0: - reg_idx = (insn >> 5) & 7; - stride = 1; - break; - case 1: - reg_idx = (insn >> 6) & 3; - stride = (insn & (1 << 5)) ? 2 : 1; - break; - case 2: - reg_idx = (insn >> 7) & 1; - stride = (insn & (1 << 6)) ? 2 : 1; - break; - default: - abort(); - } - nregs = ((insn >> 8) & 3) + 1; - /* Catch the UNDEF cases. This is unavoidably a bit messy. */ - switch (nregs) { - case 1: - if (((idx & (1 << size)) != 0) || - (size == 2 && ((idx & 3) == 1 || (idx & 3) == 2))) { - return 1; - } - break; - case 3: - if ((idx & 1) != 0) { - return 1; - } - /* fall through */ - case 2: - if (size == 2 && (idx & 2) != 0) { - return 1; - } - break; - case 4: - if ((size == 2) && ((idx & 3) == 3)) { - return 1; - } - break; - default: - abort(); - } - if ((rd + stride * (nregs - 1)) > 31) { - /* Attempts to write off the end of the register file - * are UNPREDICTABLE; we choose to UNDEF because otherwise - * the neon_load_reg() would write off the end of the array. - */ - return 1; - } - tmp = tcg_temp_new_i32(); - addr = tcg_temp_new_i32(); - load_reg_var(s, addr, rn); - for (reg = 0; reg < nregs; reg++) { - if (load) { - gen_aa32_ld_i32(s, tmp, addr, get_mem_index(s), - s->be_data | size); - neon_store_element(rd, reg_idx, size, tmp); - } else { /* Store */ - neon_load_element(tmp, rd, reg_idx, size); - gen_aa32_st_i32(s, tmp, addr, get_mem_index(s), - s->be_data | size); - } - rd += stride; - tcg_gen_addi_i32(addr, addr, 1 << size); - } - tcg_temp_free_i32(addr); - tcg_temp_free_i32(tmp); - stride = nregs * (1 << size); - } - } - if (rm != 15) { - TCGv_i32 base; - - base = load_reg(s, rn); - if (rm == 13) { - tcg_gen_addi_i32(base, base, stride); - } else { - TCGv_i32 index; - index = load_reg(s, rm); - tcg_gen_add_i32(base, base, index); - tcg_temp_free_i32(index); - } - store_reg(s, rn, base); - } - return 0; -} - static inline void gen_neon_narrow(int size, TCGv_i32 dest, TCGv_i64 src) { switch (size) { @@ -XXX,XX +XXX,XX @@ static void disas_arm_insn(DisasContext *s, unsigned int insn) } return; } - if ((insn & 0x0f100000) == 0x04000000) { - /* NEON load/store. */ - if (disas_neon_ls_insn(s, insn)) { - goto illegal_op; - } - return; - } if ((insn & 0x0e000f00) == 0x0c000100) { if (arm_dc_feature(s, ARM_FEATURE_IWMMXT)) { /* iWMMXt register transfer. */ @@ -XXX,XX +XXX,XX @@ static void disas_thumb2_insn(DisasContext *s, uint32_t insn) } break; case 12: - if ((insn & 0x01100000) == 0x01000000) { - if (disas_neon_ls_insn(s, insn)) { - goto illegal_op; - } - break; - } goto illegal_op; default: illegal_op: -- 2.20.1 Convert the Neon 3-reg-same VADD and VSUB insns to decodetree. Note that we don't need the neon_3r_sizes[op] check here because all size values are OK for VADD and VSUB; we'll add this when we convert the first insn that has size restrictions. For this we need one of the GVecGen*Fn typedefs currently in translate-a64.h; move them all to translate.h as a block so they are visible to the 32-bit decoder. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-15-peter.maydell@linaro.org --- target/arm/translate-a64.h | 9 -------- target/arm/translate.h | 9 ++++++++ target/arm/neon-dp.decode | 17 +++++++++++++++ target/arm/translate-neon.inc.c | 38 +++++++++++++++++++++++++++++++++ target/arm/translate.c | 14 ++++-------- 5 files changed, 68 insertions(+), 19 deletions(-) diff --git a/target/arm/translate-a64.h b/target/arm/translate-a64.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a64.h +++ b/target/arm/translate-a64.h @@ -XXX,XX +XXX,XX @@ static inline int vec_full_reg_size(DisasContext *s) bool disas_sve(DisasContext *, uint32_t); -/* Note that the gvec expanders operate on offsets + sizes. */ -typedef void GVecGen2Fn(unsigned, uint32_t, uint32_t, uint32_t, uint32_t); -typedef void GVecGen2iFn(unsigned, uint32_t, uint32_t, int64_t, - uint32_t, uint32_t); -typedef void GVecGen3Fn(unsigned, uint32_t, uint32_t, - uint32_t, uint32_t, uint32_t); -typedef void GVecGen4Fn(unsigned, uint32_t, uint32_t, uint32_t, - uint32_t, uint32_t, uint32_t); - #endif /* TARGET_ARM_TRANSLATE_A64_H */ diff --git a/target/arm/translate.h b/target/arm/translate.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.h +++ b/target/arm/translate.h @@ -XXX,XX +XXX,XX @@ void gen_sshl_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b); #define dc_isar_feature(name, ctx) \ ({ DisasContext *ctx_ = (ctx); isar_feature_##name(ctx_->isar); }) +/* Note that the gvec expanders operate on offsets + sizes. */ +typedef void GVecGen2Fn(unsigned, uint32_t, uint32_t, uint32_t, uint32_t); +typedef void GVecGen2iFn(unsigned, uint32_t, uint32_t, int64_t, + uint32_t, uint32_t); +typedef void GVecGen3Fn(unsigned, uint32_t, uint32_t, + uint32_t, uint32_t, uint32_t); +typedef void GVecGen4Fn(unsigned, uint32_t, uint32_t, uint32_t, + uint32_t, uint32_t, uint32_t); + #endif /* TARGET_ARM_TRANSLATE_H */ diff --git a/target/arm/neon-dp.decode b/target/arm/neon-dp.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-dp.decode +++ b/target/arm/neon-dp.decode @@ -XXX,XX +XXX,XX @@ # # This file is processed by scripts/decodetree.py # +# VFP/Neon register fields; same as vfp.decode +%vm_dp 5:1 0:4 +%vn_dp 7:1 16:4 +%vd_dp 22:1 12:4 # Encodings for Neon data processing instructions where the T32 encoding # is a simple transformation of the A32 encoding. @@ -XXX,XX +XXX,XX @@ # 0b111p_1111_qqqq_qqqq_qqqq_qqqq_qqqq_qqqq # This file works on the A32 encoding only; calling code for T32 has to # transform the insn into the A32 version first. + +###################################################################### +# 3-reg-same grouping: +# 1111 001 U 0 D sz:2 Vn:4 Vd:4 opc:4 N Q M op Vm:4 +###################################################################### + +&3same vm vn vd q size + +@3same .... ... . . . size:2 .... .... .... . q:1 . . .... \ + &3same vm=%vm_dp vn=%vn_dp vd=%vd_dp + +VADD_3s 1111 001 0 0 . .. .... .... 1000 . . . 0 .... @3same +VSUB_3s 1111 001 1 0 . .. .... .... 1000 . . . 0 .... @3same diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ static bool trans_VLDST_single(DisasContext *s, arg_VLDST_single *a) return true; } + +static bool do_3same(DisasContext *s, arg_3same *a, GVecGen3Fn fn) +{ + int vec_size = a->q ? 16 : 8; + int rd_ofs = neon_reg_offset(a->vd, 0); + int rn_ofs = neon_reg_offset(a->vn, 0); + int rm_ofs = neon_reg_offset(a->vm, 0); + + if (!arm_dc_feature(s, ARM_FEATURE_NEON)) { + return false; + } + + /* UNDEF accesses to D16-D31 if they don't exist. */ + if (!dc_isar_feature(aa32_simd_r32, s) && + ((a->vd | a->vn | a->vm) & 0x10)) { + return false; + } + + if ((a->vn | a->vm | a->vd) & a->q) { + return false; + } + + if (!vfp_access_check(s)) { + return true; + } + + fn(a->size, rd_ofs, rn_ofs, rm_ofs, vec_size, vec_size); + return true; +} + +#define DO_3SAME(INSN, FUNC) \ + static bool trans_##INSN##_3s(DisasContext *s, arg_3same *a) \ + { \ + return do_3same(s, a, FUNC); \ + } + +DO_3SAME(VADD, tcg_gen_gvec_add) +DO_3SAME(VSUB, tcg_gen_gvec_sub) diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) } return 0; - case NEON_3R_VADD_VSUB: - if (u) { - tcg_gen_gvec_sub(size, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - } else { - tcg_gen_gvec_add(size, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - } - return 0; - case NEON_3R_VQADD: tcg_gen_gvec_4(rd_ofs, offsetof(CPUARMState, vfp.qc), rn_ofs, rm_ofs, vec_size, vec_size, @@ -XXX,XX +XXX,XX @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) tcg_gen_gvec_3(rd_ofs, rm_ofs, rn_ofs, vec_size, vec_size, u ? &ushl_op[size] : &sshl_op[size]); return 0; + + case NEON_3R_VADD_VSUB: + /* Already handled by decodetree */ + return 1; } if (size == 3) { -- 2.20.1 Convert the Neon logic ops in the 3-reg-same grouping to decodetree. Note that for the logic ops the 'size' field forms part of their decode and the actual operations are always bitwise. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-16-peter.maydell@linaro.org --- target/arm/neon-dp.decode | 12 +++++++++++ target/arm/translate-neon.inc.c | 19 +++++++++++++++++ target/arm/translate.c | 38 +-------------------------------- 3 files changed, 32 insertions(+), 37 deletions(-) diff --git a/target/arm/neon-dp.decode b/target/arm/neon-dp.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-dp.decode +++ b/target/arm/neon-dp.decode @@ -XXX,XX +XXX,XX @@ @3same .... ... . . . size:2 .... .... .... . q:1 . . .... \ &3same vm=%vm_dp vn=%vn_dp vd=%vd_dp +@3same_logic .... ... . . . .. .... .... .... . q:1 .. .... \ + &3same vm=%vm_dp vn=%vn_dp vd=%vd_dp size=0 + +VAND_3s 1111 001 0 0 . 00 .... .... 0001 ... 1 .... @3same_logic +VBIC_3s 1111 001 0 0 . 01 .... .... 0001 ... 1 .... @3same_logic +VORR_3s 1111 001 0 0 . 10 .... .... 0001 ... 1 .... @3same_logic +VORN_3s 1111 001 0 0 . 11 .... .... 0001 ... 1 .... @3same_logic +VEOR_3s 1111 001 1 0 . 00 .... .... 0001 ... 1 .... @3same_logic +VBSL_3s 1111 001 1 0 . 01 .... .... 0001 ... 1 .... @3same_logic +VBIT_3s 1111 001 1 0 . 10 .... .... 0001 ... 1 .... @3same_logic +VBIF_3s 1111 001 1 0 . 11 .... .... 0001 ... 1 .... @3same_logic + VADD_3s 1111 001 0 0 . .. .... .... 1000 . . . 0 .... @3same VSUB_3s 1111 001 1 0 . .. .... .... 1000 . . . 0 .... @3same diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ static bool do_3same(DisasContext *s, arg_3same *a, GVecGen3Fn fn) DO_3SAME(VADD, tcg_gen_gvec_add) DO_3SAME(VSUB, tcg_gen_gvec_sub) +DO_3SAME(VAND, tcg_gen_gvec_and) +DO_3SAME(VBIC, tcg_gen_gvec_andc) +DO_3SAME(VORR, tcg_gen_gvec_or) +DO_3SAME(VORN, tcg_gen_gvec_orc) +DO_3SAME(VEOR, tcg_gen_gvec_xor) + +/* These insns are all gvec_bitsel but with the inputs in various orders. */ +#define DO_3SAME_BITSEL(INSN, O1, O2, O3) \ + static void gen_##INSN##_3s(unsigned vece, uint32_t rd_ofs, \ + uint32_t rn_ofs, uint32_t rm_ofs, \ + uint32_t oprsz, uint32_t maxsz) \ + { \ + tcg_gen_gvec_bitsel(vece, rd_ofs, O1, O2, O3, oprsz, maxsz); \ + } \ + DO_3SAME(INSN, gen_##INSN##_3s) + +DO_3SAME_BITSEL(VBSL, rd_ofs, rn_ofs, rm_ofs) +DO_3SAME_BITSEL(VBIT, rm_ofs, rn_ofs, rd_ofs) +DO_3SAME_BITSEL(VBIF, rm_ofs, rd_ofs, rn_ofs) diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) } return 1; - case NEON_3R_LOGIC: /* Logic ops. */ - switch ((u << 2) | size) { - case 0: /* VAND */ - tcg_gen_gvec_and(0, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - break; - case 1: /* VBIC */ - tcg_gen_gvec_andc(0, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - break; - case 2: /* VORR */ - tcg_gen_gvec_or(0, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - break; - case 3: /* VORN */ - tcg_gen_gvec_orc(0, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - break; - case 4: /* VEOR */ - tcg_gen_gvec_xor(0, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - break; - case 5: /* VBSL */ - tcg_gen_gvec_bitsel(MO_8, rd_ofs, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - break; - case 6: /* VBIT */ - tcg_gen_gvec_bitsel(MO_8, rd_ofs, rm_ofs, rn_ofs, rd_ofs, - vec_size, vec_size); - break; - case 7: /* VBIF */ - tcg_gen_gvec_bitsel(MO_8, rd_ofs, rm_ofs, rd_ofs, rn_ofs, - vec_size, vec_size); - break; - } - return 0; - case NEON_3R_VQADD: tcg_gen_gvec_4(rd_ofs, offsetof(CPUARMState, vfp.qc), rn_ofs, rm_ofs, vec_size, vec_size, @@ -XXX,XX +XXX,XX @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) return 0; case NEON_3R_VADD_VSUB: + case NEON_3R_LOGIC: /* Already handled by decodetree */ return 1; } -- 2.20.1 Convert the Neon 3-reg-same VMAX and VMIN insns to decodetree. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-17-peter.maydell@linaro.org --- target/arm/neon-dp.decode | 5 +++++ target/arm/translate-neon.inc.c | 14 ++++++++++++++ target/arm/translate.c | 21 ++------------------- 3 files changed, 21 insertions(+), 19 deletions(-) diff --git a/target/arm/neon-dp.decode b/target/arm/neon-dp.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-dp.decode +++ b/target/arm/neon-dp.decode @@ -XXX,XX +XXX,XX @@ VBSL_3s 1111 001 1 0 . 01 .... .... 0001 ... 1 .... @3same_logic VBIT_3s 1111 001 1 0 . 10 .... .... 0001 ... 1 .... @3same_logic VBIF_3s 1111 001 1 0 . 11 .... .... 0001 ... 1 .... @3same_logic +VMAX_S_3s 1111 001 0 0 . .. .... .... 0110 . . . 0 .... @3same +VMAX_U_3s 1111 001 1 0 . .. .... .... 0110 . . . 0 .... @3same +VMIN_S_3s 1111 001 0 0 . .. .... .... 0110 . . . 1 .... @3same +VMIN_U_3s 1111 001 1 0 . .. .... .... 0110 . . . 1 .... @3same + VADD_3s 1111 001 0 0 . .. .... .... 1000 . . . 0 .... @3same VSUB_3s 1111 001 1 0 . .. .... .... 1000 . . . 0 .... @3same diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ DO_3SAME(VEOR, tcg_gen_gvec_xor) DO_3SAME_BITSEL(VBSL, rd_ofs, rn_ofs, rm_ofs) DO_3SAME_BITSEL(VBIT, rm_ofs, rn_ofs, rd_ofs) DO_3SAME_BITSEL(VBIF, rm_ofs, rd_ofs, rn_ofs) + +#define DO_3SAME_NO_SZ_3(INSN, FUNC) \ + static bool trans_##INSN##_3s(DisasContext *s, arg_3same *a) \ + { \ + if (a->size == 3) { \ + return false; \ + } \ + return do_3same(s, a, FUNC); \ + } + +DO_3SAME_NO_SZ_3(VMAX_S, tcg_gen_gvec_smax) +DO_3SAME_NO_SZ_3(VMAX_U, tcg_gen_gvec_umax) +DO_3SAME_NO_SZ_3(VMIN_S, tcg_gen_gvec_smin) +DO_3SAME_NO_SZ_3(VMIN_U, tcg_gen_gvec_umin) diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) rd_ofs, rn_ofs, rm_ofs, vec_size, vec_size); return 0; - case NEON_3R_VMAX: - if (u) { - tcg_gen_gvec_umax(size, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - } else { - tcg_gen_gvec_smax(size, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - } - return 0; - case NEON_3R_VMIN: - if (u) { - tcg_gen_gvec_umin(size, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - } else { - tcg_gen_gvec_smin(size, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - } - return 0; - case NEON_3R_VSHL: /* Note the operation is vshl vd,vm,vn */ tcg_gen_gvec_3(rd_ofs, rm_ofs, rn_ofs, vec_size, vec_size, @@ -XXX,XX +XXX,XX @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) case NEON_3R_VADD_VSUB: case NEON_3R_LOGIC: + case NEON_3R_VMAX: + case NEON_3R_VMIN: /* Already handled by decodetree */ return 1; } -- 2.20.1 Convert the Neon comparison ops in the 3-reg-same grouping to decodetree. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-18-peter.maydell@linaro.org --- target/arm/neon-dp.decode | 8 ++++++++ target/arm/translate-neon.inc.c | 22 ++++++++++++++++++++++ target/arm/translate.c | 23 +++-------------------- 3 files changed, 33 insertions(+), 20 deletions(-) diff --git a/target/arm/neon-dp.decode b/target/arm/neon-dp.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-dp.decode +++ b/target/arm/neon-dp.decode @@ -XXX,XX +XXX,XX @@ VBSL_3s 1111 001 1 0 . 01 .... .... 0001 ... 1 .... @3same_logic VBIT_3s 1111 001 1 0 . 10 .... .... 0001 ... 1 .... @3same_logic VBIF_3s 1111 001 1 0 . 11 .... .... 0001 ... 1 .... @3same_logic +VCGT_S_3s 1111 001 0 0 . .. .... .... 0011 . . . 0 .... @3same +VCGT_U_3s 1111 001 1 0 . .. .... .... 0011 . . . 0 .... @3same +VCGE_S_3s 1111 001 0 0 . .. .... .... 0011 . . . 1 .... @3same +VCGE_U_3s 1111 001 1 0 . .. .... .... 0011 . . . 1 .... @3same + VMAX_S_3s 1111 001 0 0 . .. .... .... 0110 . . . 0 .... @3same VMAX_U_3s 1111 001 1 0 . .. .... .... 0110 . . . 0 .... @3same VMIN_S_3s 1111 001 0 0 . .. .... .... 0110 . . . 1 .... @3same @@ -XXX,XX +XXX,XX @@ VMIN_U_3s 1111 001 1 0 . .. .... .... 0110 . . . 1 .... @3same VADD_3s 1111 001 0 0 . .. .... .... 1000 . . . 0 .... @3same VSUB_3s 1111 001 1 0 . .. .... .... 1000 . . . 0 .... @3same + +VTST_3s 1111 001 0 0 . .. .... .... 1000 . . . 1 .... @3same +VCEQ_3s 1111 001 1 0 . .. .... .... 1000 . . . 1 .... @3same diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ DO_3SAME_NO_SZ_3(VMAX_S, tcg_gen_gvec_smax) DO_3SAME_NO_SZ_3(VMAX_U, tcg_gen_gvec_umax) DO_3SAME_NO_SZ_3(VMIN_S, tcg_gen_gvec_smin) DO_3SAME_NO_SZ_3(VMIN_U, tcg_gen_gvec_umin) + +#define DO_3SAME_CMP(INSN, COND) \ + static void gen_##INSN##_3s(unsigned vece, uint32_t rd_ofs, \ + uint32_t rn_ofs, uint32_t rm_ofs, \ + uint32_t oprsz, uint32_t maxsz) \ + { \ + tcg_gen_gvec_cmp(COND, vece, rd_ofs, rn_ofs, rm_ofs, oprsz, maxsz); \ + } \ + DO_3SAME_NO_SZ_3(INSN, gen_##INSN##_3s) + +DO_3SAME_CMP(VCGT_S, TCG_COND_GT) +DO_3SAME_CMP(VCGT_U, TCG_COND_GTU) +DO_3SAME_CMP(VCGE_S, TCG_COND_GE) +DO_3SAME_CMP(VCGE_U, TCG_COND_GEU) +DO_3SAME_CMP(VCEQ, TCG_COND_EQ) + +static void gen_VTST_3s(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs, + uint32_t rm_ofs, uint32_t oprsz, uint32_t maxsz) +{ + tcg_gen_gvec_3(rd_ofs, rn_ofs, rm_ofs, oprsz, maxsz, &cmtst_op[vece]); +} +DO_3SAME_NO_SZ_3(VTST, gen_VTST_3s) diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) u ? &mls_op[size] : &mla_op[size]); return 0; - case NEON_3R_VTST_VCEQ: - if (u) { /* VCEQ */ - tcg_gen_gvec_cmp(TCG_COND_EQ, size, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - } else { /* VTST */ - tcg_gen_gvec_3(rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size, &cmtst_op[size]); - } - return 0; - - case NEON_3R_VCGT: - tcg_gen_gvec_cmp(u ? TCG_COND_GTU : TCG_COND_GT, size, - rd_ofs, rn_ofs, rm_ofs, vec_size, vec_size); - return 0; - - case NEON_3R_VCGE: - tcg_gen_gvec_cmp(u ? TCG_COND_GEU : TCG_COND_GE, size, - rd_ofs, rn_ofs, rm_ofs, vec_size, vec_size); - return 0; - case NEON_3R_VSHL: /* Note the operation is vshl vd,vm,vn */ tcg_gen_gvec_3(rd_ofs, rm_ofs, rn_ofs, vec_size, vec_size, @@ -XXX,XX +XXX,XX @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) case NEON_3R_LOGIC: case NEON_3R_VMAX: case NEON_3R_VMIN: + case NEON_3R_VTST_VCEQ: + case NEON_3R_VCGT: + case NEON_3R_VCGE: /* Already handled by decodetree */ return 1; } -- 2.20.1 Convert the Neon VQADD/VQSUB insns in the 3-reg-same grouping to decodetree. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-19-peter.maydell@linaro.org --- target/arm/neon-dp.decode | 6 ++++++ target/arm/translate-neon.inc.c | 15 +++++++++++++++ target/arm/translate.c | 14 ++------------ 3 files changed, 23 insertions(+), 12 deletions(-) diff --git a/target/arm/neon-dp.decode b/target/arm/neon-dp.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-dp.decode +++ b/target/arm/neon-dp.decode @@ -XXX,XX +XXX,XX @@ @3same .... ... . . . size:2 .... .... .... . q:1 . . .... \ &3same vm=%vm_dp vn=%vn_dp vd=%vd_dp +VQADD_S_3s 1111 001 0 0 . .. .... .... 0000 . . . 1 .... @3same +VQADD_U_3s 1111 001 1 0 . .. .... .... 0000 . . . 1 .... @3same + @3same_logic .... ... . . . .. .... .... .... . q:1 .. .... \ &3same vm=%vm_dp vn=%vn_dp vd=%vd_dp size=0 @@ -XXX,XX +XXX,XX @@ VBSL_3s 1111 001 1 0 . 01 .... .... 0001 ... 1 .... @3same_logic VBIT_3s 1111 001 1 0 . 10 .... .... 0001 ... 1 .... @3same_logic VBIF_3s 1111 001 1 0 . 11 .... .... 0001 ... 1 .... @3same_logic +VQSUB_S_3s 1111 001 0 0 . .. .... .... 0010 . . . 1 .... @3same +VQSUB_U_3s 1111 001 1 0 . .. .... .... 0010 . . . 1 .... @3same + VCGT_S_3s 1111 001 0 0 . .. .... .... 0011 . . . 0 .... @3same VCGT_U_3s 1111 001 1 0 . .. .... .... 0011 . . . 0 .... @3same VCGE_S_3s 1111 001 0 0 . .. .... .... 0011 . . . 1 .... @3same diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ static void gen_VTST_3s(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs, tcg_gen_gvec_3(rd_ofs, rn_ofs, rm_ofs, oprsz, maxsz, &cmtst_op[vece]); } DO_3SAME_NO_SZ_3(VTST, gen_VTST_3s) + +#define DO_3SAME_GVEC4(INSN, OPARRAY) \ + static void gen_##INSN##_3s(unsigned vece, uint32_t rd_ofs, \ + uint32_t rn_ofs, uint32_t rm_ofs, \ + uint32_t oprsz, uint32_t maxsz) \ + { \ + tcg_gen_gvec_4(rd_ofs, offsetof(CPUARMState, vfp.qc), \ + rn_ofs, rm_ofs, oprsz, maxsz, &OPARRAY[vece]); \ + } \ + DO_3SAME(INSN, gen_##INSN##_3s) + +DO_3SAME_GVEC4(VQADD_S, sqadd_op) +DO_3SAME_GVEC4(VQADD_U, uqadd_op) +DO_3SAME_GVEC4(VQSUB_S, sqsub_op) +DO_3SAME_GVEC4(VQSUB_U, uqsub_op) diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) } return 1; - case NEON_3R_VQADD: - tcg_gen_gvec_4(rd_ofs, offsetof(CPUARMState, vfp.qc), - rn_ofs, rm_ofs, vec_size, vec_size, - (u ? uqadd_op : sqadd_op) + size); - return 0; - - case NEON_3R_VQSUB: - tcg_gen_gvec_4(rd_ofs, offsetof(CPUARMState, vfp.qc), - rn_ofs, rm_ofs, vec_size, vec_size, - (u ? uqsub_op : sqsub_op) + size); - return 0; - case NEON_3R_VMUL: /* VMUL */ if (u) { /* Polynomial case allows only P8. */ @@ -XXX,XX +XXX,XX @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) case NEON_3R_VTST_VCEQ: case NEON_3R_VCGT: case NEON_3R_VCGE: + case NEON_3R_VQADD: + case NEON_3R_VQSUB: /* Already handled by decodetree */ return 1; } -- 2.20.1 Convert the Neon VMUL, VMLA, VMLS and VSHL insns in the 3-reg-same grouping to decodetree. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-20-peter.maydell@linaro.org --- target/arm/neon-dp.decode | 9 +++++++ target/arm/translate-neon.inc.c | 44 +++++++++++++++++++++++++++++++++ target/arm/translate.c | 28 +++------------------ 3 files changed, 56 insertions(+), 25 deletions(-) diff --git a/target/arm/neon-dp.decode b/target/arm/neon-dp.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-dp.decode +++ b/target/arm/neon-dp.decode @@ -XXX,XX +XXX,XX @@ VCGT_U_3s 1111 001 1 0 . .. .... .... 0011 . . . 0 .... @3same VCGE_S_3s 1111 001 0 0 . .. .... .... 0011 . . . 1 .... @3same VCGE_U_3s 1111 001 1 0 . .. .... .... 0011 . . . 1 .... @3same +VSHL_S_3s 1111 001 0 0 . .. .... .... 0100 . . . 0 .... @3same +VSHL_U_3s 1111 001 1 0 . .. .... .... 0100 . . . 0 .... @3same + VMAX_S_3s 1111 001 0 0 . .. .... .... 0110 . . . 0 .... @3same VMAX_U_3s 1111 001 1 0 . .. .... .... 0110 . . . 0 .... @3same VMIN_S_3s 1111 001 0 0 . .. .... .... 0110 . . . 1 .... @3same @@ -XXX,XX +XXX,XX @@ VSUB_3s 1111 001 1 0 . .. .... .... 1000 . . . 0 .... @3same VTST_3s 1111 001 0 0 . .. .... .... 1000 . . . 1 .... @3same VCEQ_3s 1111 001 1 0 . .. .... .... 1000 . . . 1 .... @3same + +VMLA_3s 1111 001 0 0 . .. .... .... 1001 . . . 0 .... @3same +VMLS_3s 1111 001 1 0 . .. .... .... 1001 . . . 0 .... @3same + +VMUL_3s 1111 001 0 0 . .. .... .... 1001 . . . 1 .... @3same +VMUL_p_3s 1111 001 1 0 . .. .... .... 1001 . . . 1 .... @3same diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ DO_3SAME_NO_SZ_3(VMAX_S, tcg_gen_gvec_smax) DO_3SAME_NO_SZ_3(VMAX_U, tcg_gen_gvec_umax) DO_3SAME_NO_SZ_3(VMIN_S, tcg_gen_gvec_smin) DO_3SAME_NO_SZ_3(VMIN_U, tcg_gen_gvec_umin) +DO_3SAME_NO_SZ_3(VMUL, tcg_gen_gvec_mul) #define DO_3SAME_CMP(INSN, COND) \ static void gen_##INSN##_3s(unsigned vece, uint32_t rd_ofs, \ @@ -XXX,XX +XXX,XX @@ DO_3SAME_GVEC4(VQADD_S, sqadd_op) DO_3SAME_GVEC4(VQADD_U, uqadd_op) DO_3SAME_GVEC4(VQSUB_S, sqsub_op) DO_3SAME_GVEC4(VQSUB_U, uqsub_op) + +static void gen_VMUL_p_3s(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs, + uint32_t rm_ofs, uint32_t oprsz, uint32_t maxsz) +{ + tcg_gen_gvec_3_ool(rd_ofs, rn_ofs, rm_ofs, oprsz, maxsz, + 0, gen_helper_gvec_pmul_b); +} + +static bool trans_VMUL_p_3s(DisasContext *s, arg_3same *a) +{ + if (a->size != 0) { + return false; + } + return do_3same(s, a, gen_VMUL_p_3s); +} + +#define DO_3SAME_GVEC3_NO_SZ_3(INSN, OPARRAY) \ + static void gen_##INSN##_3s(unsigned vece, uint32_t rd_ofs, \ + uint32_t rn_ofs, uint32_t rm_ofs, \ + uint32_t oprsz, uint32_t maxsz) \ + { \ + tcg_gen_gvec_3(rd_ofs, rn_ofs, rm_ofs, \ + oprsz, maxsz, &OPARRAY[vece]); \ + } \ + DO_3SAME_NO_SZ_3(INSN, gen_##INSN##_3s) + + +DO_3SAME_GVEC3_NO_SZ_3(VMLA, mla_op) +DO_3SAME_GVEC3_NO_SZ_3(VMLS, mls_op) + +#define DO_3SAME_GVEC3_SHIFT(INSN, OPARRAY) \ + static void gen_##INSN##_3s(unsigned vece, uint32_t rd_ofs, \ + uint32_t rn_ofs, uint32_t rm_ofs, \ + uint32_t oprsz, uint32_t maxsz) \ + { \ + /* Note the operation is vshl vd,vm,vn */ \ + tcg_gen_gvec_3(rd_ofs, rm_ofs, rn_ofs, \ + oprsz, maxsz, &OPARRAY[vece]); \ + } \ + DO_3SAME(INSN, gen_##INSN##_3s) + +DO_3SAME_GVEC3_SHIFT(VSHL_S, sshl_op) +DO_3SAME_GVEC3_SHIFT(VSHL_U, ushl_op) diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) } return 1; - case NEON_3R_VMUL: /* VMUL */ - if (u) { - /* Polynomial case allows only P8. */ - if (size != 0) { - return 1; - } - tcg_gen_gvec_3_ool(rd_ofs, rn_ofs, rm_ofs, vec_size, vec_size, - 0, gen_helper_gvec_pmul_b); - } else { - tcg_gen_gvec_mul(size, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - } - return 0; - - case NEON_3R_VML: /* VMLA, VMLS */ - tcg_gen_gvec_3(rd_ofs, rn_ofs, rm_ofs, vec_size, vec_size, - u ? &mls_op[size] : &mla_op[size]); - return 0; - - case NEON_3R_VSHL: - /* Note the operation is vshl vd,vm,vn */ - tcg_gen_gvec_3(rd_ofs, rm_ofs, rn_ofs, vec_size, vec_size, - u ? &ushl_op[size] : &sshl_op[size]); - return 0; - case NEON_3R_VADD_VSUB: case NEON_3R_LOGIC: case NEON_3R_VMAX: @@ -XXX,XX +XXX,XX @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) case NEON_3R_VCGE: case NEON_3R_VQADD: case NEON_3R_VQSUB: + case NEON_3R_VMUL: + case NEON_3R_VML: + case NEON_3R_VSHL: /* Already handled by decodetree */ return 1; } -- 2.20.1 We're going to want at least some of the NeonGen* typedefs for the refactored 32-bit Neon decoder, so move them all to translate.h since it makes more sense to keep them in one group. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-23-peter.maydell@linaro.org --- target/arm/translate.h | 17 +++++++++++++++++ target/arm/translate-a64.c | 17 ----------------- 2 files changed, 17 insertions(+), 17 deletions(-) diff --git a/target/arm/translate.h b/target/arm/translate.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.h +++ b/target/arm/translate.h @@ -XXX,XX +XXX,XX @@ typedef void GVecGen3Fn(unsigned, uint32_t, uint32_t, typedef void GVecGen4Fn(unsigned, uint32_t, uint32_t, uint32_t, uint32_t, uint32_t, uint32_t); +/* Function prototype for gen_ functions for calling Neon helpers */ +typedef void NeonGenOneOpEnvFn(TCGv_i32, TCGv_ptr, TCGv_i32); +typedef void NeonGenTwoOpFn(TCGv_i32, TCGv_i32, TCGv_i32); +typedef void NeonGenTwoOpEnvFn(TCGv_i32, TCGv_ptr, TCGv_i32, TCGv_i32); +typedef void NeonGenTwo64OpFn(TCGv_i64, TCGv_i64, TCGv_i64); +typedef void NeonGenTwo64OpEnvFn(TCGv_i64, TCGv_ptr, TCGv_i64, TCGv_i64); +typedef void NeonGenNarrowFn(TCGv_i32, TCGv_i64); +typedef void NeonGenNarrowEnvFn(TCGv_i32, TCGv_ptr, TCGv_i64); +typedef void NeonGenWidenFn(TCGv_i64, TCGv_i32); +typedef void NeonGenTwoSingleOPFn(TCGv_i32, TCGv_i32, TCGv_i32, TCGv_ptr); +typedef void NeonGenTwoDoubleOPFn(TCGv_i64, TCGv_i64, TCGv_i64, TCGv_ptr); +typedef void NeonGenOneOpFn(TCGv_i64, TCGv_i64); +typedef void CryptoTwoOpFn(TCGv_ptr, TCGv_ptr); +typedef void CryptoThreeOpIntFn(TCGv_ptr, TCGv_ptr, TCGv_i32); +typedef void CryptoThreeOpFn(TCGv_ptr, TCGv_ptr, TCGv_ptr); +typedef void AtomicThreeOpFn(TCGv_i64, TCGv_i64, TCGv_i64, TCGArg, MemOp); + #endif /* TARGET_ARM_TRANSLATE_H */ diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a64.c +++ b/target/arm/translate-a64.c @@ -XXX,XX +XXX,XX @@ typedef struct AArch64DecodeTable { AArch64DecodeFn *disas_fn; } AArch64DecodeTable; -/* Function prototype for gen_ functions for calling Neon helpers */ -typedef void NeonGenOneOpEnvFn(TCGv_i32, TCGv_ptr, TCGv_i32); -typedef void NeonGenTwoOpFn(TCGv_i32, TCGv_i32, TCGv_i32); -typedef void NeonGenTwoOpEnvFn(TCGv_i32, TCGv_ptr, TCGv_i32, TCGv_i32); -typedef void NeonGenTwo64OpFn(TCGv_i64, TCGv_i64, TCGv_i64); -typedef void NeonGenTwo64OpEnvFn(TCGv_i64, TCGv_ptr, TCGv_i64, TCGv_i64); -typedef void NeonGenNarrowFn(TCGv_i32, TCGv_i64); -typedef void NeonGenNarrowEnvFn(TCGv_i32, TCGv_ptr, TCGv_i64); -typedef void NeonGenWidenFn(TCGv_i64, TCGv_i32); -typedef void NeonGenTwoSingleOPFn(TCGv_i32, TCGv_i32, TCGv_i32, TCGv_ptr); -typedef void NeonGenTwoDoubleOPFn(TCGv_i64, TCGv_i64, TCGv_i64, TCGv_ptr); -typedef void NeonGenOneOpFn(TCGv_i64, TCGv_i64); -typedef void CryptoTwoOpFn(TCGv_ptr, TCGv_ptr); -typedef void CryptoThreeOpIntFn(TCGv_ptr, TCGv_ptr, TCGv_i32); -typedef void CryptoThreeOpFn(TCGv_ptr, TCGv_ptr, TCGv_ptr); -typedef void AtomicThreeOpFn(TCGv_i64, TCGv_i64, TCGv_i64, TCGArg, MemOp); - /* initialize TCG globals. */ void a64_translate_init(void) { -- 2.20.1 Hi; here's the latest arm pullreq. This is mostly patches from RTH, plus a couple of other more minor things. Switching to PCREL is the big one, hopefully should improve performance. thanks -- PMM The following changes since commit 214a8da23651f2472b296b3293e619fd58d9e212: Merge tag 'for-upstream' of https://gitlab.com/bonzini/qemu into staging (2022-10-18 11:14:31 -0400) are available in the Git repository at: https://git.linaro.org/people/pmaydell/qemu-arm.git tags/pull-target-arm-20221020 for you to fetch changes up to 5db899303799e49209016a93289b8694afa1449e: hw/ide/microdrive: Use device_cold_reset() for self-resets (2022-10-20 12:11:53 +0100) ---------------------------------------------------------------- target-arm queue: * Switch to TARGET_TB_PCREL * More pagetable-walk refactoring preparatory to HAFDBS * update the cortex-a15 MIDR to latest rev * hw/char/pl011: fix baud rate calculation * hw/ide/microdrive: Use device_cold_reset() for self-resets ---------------------------------------------------------------- Alex Bennée (1): target/arm: update the cortex-a15 MIDR to latest rev Baruch Siach (1): hw/char/pl011: fix baud rate calculation Peter Maydell (1): hw/ide/microdrive: Use device_cold_reset() for self-resets Richard Henderson (21): target/arm: Enable TARGET_PAGE_ENTRY_EXTRA target/arm: Use probe_access_full for MTE target/arm: Use probe_access_full for BTI target/arm: Add ARMMMUIdx_Phys_{S,NS} target/arm: Move ARMMMUIdx_Stage2 to a real tlb mmu_idx target/arm: Restrict tlb flush from vttbr_write to vmid change target/arm: Split out S1Translate type target/arm: Plumb debug into S1Translate target/arm: Move be test for regime into S1TranslateResult target/arm: Use softmmu tlbs for page table walking target/arm: Split out get_phys_addr_twostage target/arm: Use bool consistently for get_phys_addr subroutines target/arm: Introduce curr_insn_len target/arm: Change gen_goto_tb to work on displacements target/arm: Change gen_*set_pc_im to gen_*update_pc target/arm: Change gen_exception_insn* to work on displacements target/arm: Remove gen_exception_internal_insn pc argument target/arm: Change gen_jmp* to work on displacements target/arm: Introduce gen_pc_plus_diff for aarch64 target/arm: Introduce gen_pc_plus_diff for aarch32 target/arm: Enable TARGET_TB_PCREL target/arm/cpu-param.h | 17 +- target/arm/cpu.h | 47 ++-- target/arm/internals.h | 1 + target/arm/sve_ldst_internal.h | 1 + target/arm/translate-a32.h | 2 +- target/arm/translate.h | 66 ++++- hw/char/pl011.c | 2 +- hw/ide/microdrive.c | 8 +- target/arm/cpu.c | 23 +- target/arm/cpu_tcg.c | 4 +- target/arm/helper.c | 155 +++++++++--- target/arm/mte_helper.c | 62 ++--- target/arm/ptw.c | 535 +++++++++++++++++++++++++---------------- target/arm/sve_helper.c | 54 ++--- target/arm/tlb_helper.c | 24 +- target/arm/translate-a64.c | 220 ++++++++++------- target/arm/translate-m-nocp.c | 8 +- target/arm/translate-mve.c | 2 +- target/arm/translate-vfp.c | 10 +- target/arm/translate.c | 284 +++++++++++++--------- 20 files changed, 918 insertions(+), 607 deletions(-) From: Baruch Siach <baruch@tkos.co.il> The PL011 TRM says that "UARTIBRD = 0 is invalid and UARTFBRD is ignored when this is the case". But the code looks at FBRD for the invalid case. Fix this. Signed-off-by: Baruch Siach <baruch@tkos.co.il> Message-id: 1408f62a2e45665816527d4845ffde650957d5ab.1665051588.git.baruchs-c@neureality.ai Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- hw/char/pl011.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/hw/char/pl011.c b/hw/char/pl011.c index XXXXXXX..XXXXXXX 100644 --- a/hw/char/pl011.c +++ b/hw/char/pl011.c @@ -XXX,XX +XXX,XX @@ static unsigned int pl011_get_baudrate(const PL011State *s) { uint64_t clk; - if (s->fbrd == 0) { + if (s->ibrd == 0) { return 0; } -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> The CPUTLBEntryFull structure now stores the original pte attributes, as well as the physical address. Therefore, we no longer need a separate bit in MemTxAttrs, nor do we need to walk the tree of memory regions. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20221011031911.2408754-3-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/cpu.h | 1 - target/arm/sve_ldst_internal.h | 1 + target/arm/mte_helper.c | 62 ++++++++++------------------------ target/arm/sve_helper.c | 54 ++++++++++------------------- target/arm/tlb_helper.c | 4 --- 5 files changed, 36 insertions(+), 86 deletions(-) diff --git a/target/arm/cpu.h b/target/arm/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.h +++ b/target/arm/cpu.h @@ -XXX,XX +XXX,XX @@ static inline MemTxAttrs *typecheck_memtxattrs(MemTxAttrs *x) * generic target bits directly. */ #define arm_tlb_bti_gp(x) (typecheck_memtxattrs(x)->target_tlb_bit0) -#define arm_tlb_mte_tagged(x) (typecheck_memtxattrs(x)->target_tlb_bit1) /* * AArch64 usage of the PAGE_TARGET_* bits for linux-user. diff --git a/target/arm/sve_ldst_internal.h b/target/arm/sve_ldst_internal.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_ldst_internal.h +++ b/target/arm/sve_ldst_internal.h @@ -XXX,XX +XXX,XX @@ typedef struct { void *host; int flags; MemTxAttrs attrs; + bool tagged; } SVEHostPage; bool sve_probe_page(SVEHostPage *info, bool nofault, CPUARMState *env, diff --git a/target/arm/mte_helper.c b/target/arm/mte_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mte_helper.c +++ b/target/arm/mte_helper.c @@ -XXX,XX +XXX,XX @@ static uint8_t *allocation_tag_mem(CPUARMState *env, int ptr_mmu_idx, TARGET_PAGE_BITS - LOG2_TAG_GRANULE - 1); return tags + index; #else - uintptr_t index; CPUTLBEntryFull *full; + MemTxAttrs attrs; int in_page, flags; - ram_addr_t ptr_ra; hwaddr ptr_paddr, tag_paddr, xlat; MemoryRegion *mr; ARMASIdx tag_asi; @@ -XXX,XX +XXX,XX @@ static uint8_t *allocation_tag_mem(CPUARMState *env, int ptr_mmu_idx, * valid. Indicate to probe_access_flags no-fault, then assert that * we received a valid page. */ - flags = probe_access_flags(env, ptr, ptr_access, ptr_mmu_idx, - ra == 0, &host, ra); + flags = probe_access_full(env, ptr, ptr_access, ptr_mmu_idx, + ra == 0, &host, &full, ra); assert(!(flags & TLB_INVALID_MASK)); - /* - * Find the CPUTLBEntryFull for ptr. This *must* be present in the TLB - * because we just found the mapping. - * TODO: Perhaps there should be a cputlb helper that returns a - * matching tlb entry + iotlb entry. - */ - index = tlb_index(env, ptr_mmu_idx, ptr); -# ifdef CONFIG_DEBUG_TCG - { - CPUTLBEntry *entry = tlb_entry(env, ptr_mmu_idx, ptr); - target_ulong comparator = (ptr_access == MMU_DATA_LOAD - ? entry->addr_read - : tlb_addr_write(entry)); - g_assert(tlb_hit(comparator, ptr)); - } -# endif - full = &env_tlb(env)->d[ptr_mmu_idx].fulltlb[index]; - /* If the virtual page MemAttr != Tagged, access unchecked. */ - if (!arm_tlb_mte_tagged(&full->attrs)) { + if (full->pte_attrs != 0xf0) { return NULL; } @@ -XXX,XX +XXX,XX @@ static uint8_t *allocation_tag_mem(CPUARMState *env, int ptr_mmu_idx, return NULL; } + /* + * Remember these values across the second lookup below, + * which may invalidate this pointer via tlb resize. + */ + ptr_paddr = full->phys_addr; + attrs = full->attrs; + full = NULL; + /* * The Normal memory access can extend to the next page. E.g. a single * 8-byte access to the last byte of a page will check only the last @@ -XXX,XX +XXX,XX @@ static uint8_t *allocation_tag_mem(CPUARMState *env, int ptr_mmu_idx, */ in_page = -(ptr | TARGET_PAGE_MASK); if (unlikely(ptr_size > in_page)) { - void *ignore; - flags |= probe_access_flags(env, ptr + in_page, ptr_access, - ptr_mmu_idx, ra == 0, &ignore, ra); + flags |= probe_access_full(env, ptr + in_page, ptr_access, + ptr_mmu_idx, ra == 0, &host, &full, ra); assert(!(flags & TLB_INVALID_MASK)); } @@ -XXX,XX +XXX,XX @@ static uint8_t *allocation_tag_mem(CPUARMState *env, int ptr_mmu_idx, if (unlikely(flags & TLB_WATCHPOINT)) { int wp = ptr_access == MMU_DATA_LOAD ? BP_MEM_READ : BP_MEM_WRITE; assert(ra != 0); - cpu_check_watchpoint(env_cpu(env), ptr, ptr_size, - full->attrs, wp, ra); + cpu_check_watchpoint(env_cpu(env), ptr, ptr_size, attrs, wp, ra); } - /* - * Find the physical address within the normal mem space. - * The memory region lookup must succeed because TLB_MMIO was - * not set in the cputlb lookup above. - */ - mr = memory_region_from_host(host, &ptr_ra); - tcg_debug_assert(mr != NULL); - tcg_debug_assert(memory_region_is_ram(mr)); - ptr_paddr = ptr_ra; - do { - ptr_paddr += mr->addr; - mr = mr->container; - } while (mr); - /* Convert to the physical address in tag space. */ tag_paddr = ptr_paddr >> (LOG2_TAG_GRANULE + 1); /* Look up the address in tag space. */ - tag_asi = full->attrs.secure ? ARMASIdx_TagS : ARMASIdx_TagNS; + tag_asi = attrs.secure ? ARMASIdx_TagS : ARMASIdx_TagNS; tag_as = cpu_get_address_space(env_cpu(env), tag_asi); mr = address_space_translate(tag_as, tag_paddr, &xlat, NULL, - tag_access == MMU_DATA_STORE, - full->attrs); + tag_access == MMU_DATA_STORE, attrs); /* * Note that @mr will never be NULL. If there is nothing in the address diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ bool sve_probe_page(SVEHostPage *info, bool nofault, CPUARMState *env, */ addr = useronly_clean_ptr(addr); +#ifdef CONFIG_USER_ONLY flags = probe_access_flags(env, addr, access_type, mmu_idx, nofault, &info->host, retaddr); + memset(&info->attrs, 0, sizeof(info->attrs)); + /* Require both ANON and MTE; see allocation_tag_mem(). */ + info->tagged = (flags & PAGE_ANON) && (flags & PAGE_MTE); +#else + CPUTLBEntryFull *full; + flags = probe_access_full(env, addr, access_type, mmu_idx, nofault, + &info->host, &full, retaddr); + info->attrs = full->attrs; + info->tagged = full->pte_attrs == 0xf0; +#endif info->flags = flags; if (flags & TLB_INVALID_MASK) { @@ -XXX,XX +XXX,XX @@ bool sve_probe_page(SVEHostPage *info, bool nofault, CPUARMState *env, /* Ensure that info->host[] is relative to addr, not addr + mem_off. */ info->host -= mem_off; - -#ifdef CONFIG_USER_ONLY - memset(&info->attrs, 0, sizeof(info->attrs)); - /* Require both MAP_ANON and PROT_MTE -- see allocation_tag_mem. */ - arm_tlb_mte_tagged(&info->attrs) = - (flags & PAGE_ANON) && (flags & PAGE_MTE); -#else - /* - * Find the iotlbentry for addr and return the transaction attributes. - * This *must* be present in the TLB because we just found the mapping. - */ - { - uintptr_t index = tlb_index(env, mmu_idx, addr); - -# ifdef CONFIG_DEBUG_TCG - CPUTLBEntry *entry = tlb_entry(env, mmu_idx, addr); - target_ulong comparator = (access_type == MMU_DATA_LOAD - ? entry->addr_read - : tlb_addr_write(entry)); - g_assert(tlb_hit(comparator, addr)); -# endif - - CPUTLBEntryFull *full = &env_tlb(env)->d[mmu_idx].fulltlb[index]; - info->attrs = full->attrs; - } -#endif - return true; } @@ -XXX,XX +XXX,XX @@ void sve_cont_ldst_mte_check(SVEContLdSt *info, CPUARMState *env, intptr_t mem_off, reg_off, reg_last; /* Process the page only if MemAttr == Tagged. */ - if (arm_tlb_mte_tagged(&info->page[0].attrs)) { + if (info->page[0].tagged) { mem_off = info->mem_off_first[0]; reg_off = info->reg_off_first[0]; reg_last = info->reg_off_split; @@ -XXX,XX +XXX,XX @@ void sve_cont_ldst_mte_check(SVEContLdSt *info, CPUARMState *env, } mem_off = info->mem_off_first[1]; - if (mem_off >= 0 && arm_tlb_mte_tagged(&info->page[1].attrs)) { + if (mem_off >= 0 && info->page[1].tagged) { reg_off = info->reg_off_first[1]; reg_last = info->reg_off_last[1]; @@ -XXX,XX +XXX,XX @@ void sve_ldnfff1_r(CPUARMState *env, void *vg, const target_ulong addr, * Disable MTE checking if the Tagged bit is not set. Since TBI must * be set within MTEDESC for MTE, !mtedesc => !mte_active. */ - if (!arm_tlb_mte_tagged(&info.page[0].attrs)) { + if (!info.page[0].tagged) { mtedesc = 0; } @@ -XXX,XX +XXX,XX @@ void sve_ld1_z(CPUARMState *env, void *vd, uint64_t *vg, void *vm, cpu_check_watchpoint(env_cpu(env), addr, msize, info.attrs, BP_MEM_READ, retaddr); } - if (mtedesc && arm_tlb_mte_tagged(&info.attrs)) { + if (mtedesc && info.tagged) { mte_check(env, mtedesc, addr, retaddr); } if (unlikely(info.flags & TLB_MMIO)) { @@ -XXX,XX +XXX,XX @@ void sve_ld1_z(CPUARMState *env, void *vd, uint64_t *vg, void *vm, msize, info.attrs, BP_MEM_READ, retaddr); } - if (mtedesc && arm_tlb_mte_tagged(&info.attrs)) { + if (mtedesc && info.tagged) { mte_check(env, mtedesc, addr, retaddr); } tlb_fn(env, &scratch, reg_off, addr, retaddr); @@ -XXX,XX +XXX,XX @@ void sve_ldff1_z(CPUARMState *env, void *vd, uint64_t *vg, void *vm, (env_cpu(env), addr, msize) & BP_MEM_READ)) { goto fault; } - if (mtedesc && - arm_tlb_mte_tagged(&info.attrs) && - !mte_probe(env, mtedesc, addr)) { + if (mtedesc && info.tagged && !mte_probe(env, mtedesc, addr)) { goto fault; } @@ -XXX,XX +XXX,XX @@ void sve_st1_z(CPUARMState *env, void *vd, uint64_t *vg, void *vm, info.attrs, BP_MEM_WRITE, retaddr); } - if (mtedesc && arm_tlb_mte_tagged(&info.attrs)) { + if (mtedesc && info.tagged) { mte_check(env, mtedesc, addr, retaddr); } } diff --git a/target/arm/tlb_helper.c b/target/arm/tlb_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/tlb_helper.c +++ b/target/arm/tlb_helper.c @@ -XXX,XX +XXX,XX @@ bool arm_cpu_tlb_fill(CPUState *cs, vaddr address, int size, res.f.phys_addr &= TARGET_PAGE_MASK; address &= TARGET_PAGE_MASK; } - /* Notice and record tagged memory. */ - if (cpu_isar_feature(aa64_mte, cpu) && res.cacheattrs.attrs == 0xf0) { - arm_tlb_mte_tagged(&res.f.attrs) = true; - } res.f.pte_attrs = res.cacheattrs.attrs; res.f.shareability = res.cacheattrs.shareability; -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Add a field to TARGET_PAGE_ENTRY_EXTRA to hold the guarded bit. In is_guarded_page, use probe_access_full instead of just guessing that the tlb entry is still present. Also handles the FIXME about executing from device memory. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20221011031911.2408754-4-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/cpu-param.h | 9 +++++---- target/arm/cpu.h | 13 ------------- target/arm/internals.h | 1 + target/arm/ptw.c | 7 ++++--- target/arm/translate-a64.c | 21 ++++++++++----------- 5 files changed, 20 insertions(+), 31 deletions(-) diff --git a/target/arm/cpu-param.h b/target/arm/cpu-param.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu-param.h +++ b/target/arm/cpu-param.h @@ -XXX,XX +XXX,XX @@ * * For ARMMMUIdx_Stage2*, pte_attrs is the S2 descriptor bits [5:2]. * Otherwise, pte_attrs is the same as the MAIR_EL1 8-bit format. - * For shareability, as in the SH field of the VMSAv8-64 PTEs. + * For shareability and guarded, as in the SH and GP fields respectively + * of the VMSAv8-64 PTEs. */ # define TARGET_PAGE_ENTRY_EXTRA \ - uint8_t pte_attrs; \ - uint8_t shareability; - + uint8_t pte_attrs; \ + uint8_t shareability; \ + bool guarded; #endif #define NB_MMU_MODES 8 diff --git a/target/arm/cpu.h b/target/arm/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.h +++ b/target/arm/cpu.h @@ -XXX,XX +XXX,XX @@ static inline uint64_t *aa64_vfp_qreg(CPUARMState *env, unsigned regno) /* Shared between translate-sve.c and sve_helper.c. */ extern const uint64_t pred_esz_masks[5]; -/* Helper for the macros below, validating the argument type. */ -static inline MemTxAttrs *typecheck_memtxattrs(MemTxAttrs *x) -{ - return x; -} - -/* - * Lvalue macros for ARM TLB bits that we must cache in the TCG TLB. - * Using these should be a bit more self-documenting than using the - * generic target bits directly. - */ -#define arm_tlb_bti_gp(x) (typecheck_memtxattrs(x)->target_tlb_bit0) - /* * AArch64 usage of the PAGE_TARGET_* bits for linux-user. * Note that with the Linux kernel, PROT_MTE may not be cleared by mprotect diff --git a/target/arm/internals.h b/target/arm/internals.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/internals.h +++ b/target/arm/internals.h @@ -XXX,XX +XXX,XX @@ typedef struct ARMCacheAttrs { unsigned int attrs:8; unsigned int shareability:2; /* as in the SH field of the VMSAv8-64 PTEs */ bool is_s2_format:1; + bool guarded:1; /* guarded bit of the v8-64 PTE */ } ARMCacheAttrs; /* Fields that are valid upon success. */ diff --git a/target/arm/ptw.c b/target/arm/ptw.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/ptw.c +++ b/target/arm/ptw.c @@ -XXX,XX +XXX,XX @@ static bool get_phys_addr_lpae(CPUARMState *env, uint64_t address, */ result->f.attrs.secure = false; } - /* When in aarch64 mode, and BTI is enabled, remember GP in the IOTLB. */ - if (aarch64 && guarded && cpu_isar_feature(aa64_bti, cpu)) { - arm_tlb_bti_gp(&result->f.attrs) = true; + + /* When in aarch64 mode, and BTI is enabled, remember GP in the TLB. */ + if (aarch64 && cpu_isar_feature(aa64_bti, cpu)) { + result->f.guarded = guarded; } if (mmu_idx == ARMMMUIdx_Stage2 || mmu_idx == ARMMMUIdx_Stage2_S) { diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a64.c +++ b/target/arm/translate-a64.c @@ -XXX,XX +XXX,XX @@ static bool is_guarded_page(CPUARMState *env, DisasContext *s) #ifdef CONFIG_USER_ONLY return page_get_flags(addr) & PAGE_BTI; #else + CPUTLBEntryFull *full; + void *host; int mmu_idx = arm_to_core_mmu_idx(s->mmu_idx); - unsigned int index = tlb_index(env, mmu_idx, addr); - CPUTLBEntry *entry = tlb_entry(env, mmu_idx, addr); + int flags; /* * We test this immediately after reading an insn, which means - * that any normal page must be in the TLB. The only exception - * would be for executing from flash or device memory, which - * does not retain the TLB entry. - * - * FIXME: Assume false for those, for now. We could use - * arm_cpu_get_phys_page_attrs_debug to re-read the page - * table entry even for that case. + * that the TLB entry must be present and valid, and thus this + * access will never raise an exception. */ - return (tlb_hit(entry->addr_code, addr) && - arm_tlb_bti_gp(&env_tlb(env)->d[mmu_idx].fulltlb[index].attrs)); + flags = probe_access_full(env, addr, MMU_INST_FETCH, mmu_idx, + false, &host, &full, 0); + assert(!(flags & TLB_INVALID_MASK)); + + return full->guarded; #endif } -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Not yet used, but add mmu indexes for 1-1 mapping to physical addresses. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20221011031911.2408754-5-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/cpu-param.h | 2 +- target/arm/cpu.h | 7 ++++++- target/arm/ptw.c | 19 +++++++++++++++++-- 3 files changed, 24 insertions(+), 4 deletions(-) diff --git a/target/arm/cpu-param.h b/target/arm/cpu-param.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu-param.h +++ b/target/arm/cpu-param.h @@ -XXX,XX +XXX,XX @@ bool guarded; #endif -#define NB_MMU_MODES 8 +#define NB_MMU_MODES 10 #endif diff --git a/target/arm/cpu.h b/target/arm/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.h +++ b/target/arm/cpu.h @@ -XXX,XX +XXX,XX @@ bool write_cpustate_to_list(ARMCPU *cpu, bool kvm_sync); * EL2 EL2&0 +PAN * EL2 (aka NS PL2) * EL3 (aka S PL1) + * Physical (NS & S) * - * for a total of 8 different mmu_idx. + * for a total of 10 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 @@ -XXX,XX +XXX,XX @@ typedef enum ARMMMUIdx { ARMMMUIdx_E2 = 6 | ARM_MMU_IDX_A, ARMMMUIdx_E3 = 7 | ARM_MMU_IDX_A, + /* TLBs with 1-1 mapping to the physical address spaces. */ + ARMMMUIdx_Phys_NS = 8 | ARM_MMU_IDX_A, + ARMMMUIdx_Phys_S = 9 | 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. diff --git a/target/arm/ptw.c b/target/arm/ptw.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/ptw.c +++ b/target/arm/ptw.c @@ -XXX,XX +XXX,XX @@ static bool regime_translation_disabled(CPUARMState *env, ARMMMUIdx mmu_idx, case ARMMMUIdx_E3: break; + case ARMMMUIdx_Phys_NS: + case ARMMMUIdx_Phys_S: + /* No translation for physical address spaces. */ + return true; + default: g_assert_not_reached(); } @@ -XXX,XX +XXX,XX @@ static bool get_phys_addr_disabled(CPUARMState *env, target_ulong address, { uint8_t memattr = 0x00; /* Device nGnRnE */ uint8_t shareability = 0; /* non-sharable */ + int r_el; - if (mmu_idx != ARMMMUIdx_Stage2 && mmu_idx != ARMMMUIdx_Stage2_S) { - int r_el = regime_el(env, mmu_idx); + switch (mmu_idx) { + case ARMMMUIdx_Stage2: + case ARMMMUIdx_Stage2_S: + case ARMMMUIdx_Phys_NS: + case ARMMMUIdx_Phys_S: + break; + default: + r_el = regime_el(env, 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]; @@ -XXX,XX +XXX,XX @@ static bool get_phys_addr_disabled(CPUARMState *env, target_ulong address, shareability = 2; /* outer sharable */ } result->cacheattrs.is_s2_format = false; + break; } result->f.phys_addr = address; @@ -XXX,XX +XXX,XX @@ bool get_phys_addr(CPUARMState *env, target_ulong address, is_secure = arm_is_secure_below_el3(env); break; case ARMMMUIdx_Stage2: + case ARMMMUIdx_Phys_NS: case ARMMMUIdx_MPrivNegPri: case ARMMMUIdx_MUserNegPri: case ARMMMUIdx_MPriv: @@ -XXX,XX +XXX,XX @@ bool get_phys_addr(CPUARMState *env, target_ulong address, break; case ARMMMUIdx_E3: case ARMMMUIdx_Stage2_S: + case ARMMMUIdx_Phys_S: case ARMMMUIdx_MSPrivNegPri: case ARMMMUIdx_MSUserNegPri: case ARMMMUIdx_MSPriv: -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> We had been marking this ARM_MMU_IDX_NOTLB, move it to a real tlb. Flush the tlb when invalidating stage 1+2 translations. Re-use alle1_tlbmask() for other instances of EL1&0 + Stage2. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Message-id: 20221011031911.2408754-6-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/cpu-param.h | 2 +- target/arm/cpu.h | 23 ++++--- target/arm/helper.c | 151 ++++++++++++++++++++++++++++++----------- 3 files changed, 127 insertions(+), 49 deletions(-) diff --git a/target/arm/cpu-param.h b/target/arm/cpu-param.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu-param.h +++ b/target/arm/cpu-param.h @@ -XXX,XX +XXX,XX @@ bool guarded; #endif -#define NB_MMU_MODES 10 +#define NB_MMU_MODES 12 #endif diff --git a/target/arm/cpu.h b/target/arm/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.h +++ b/target/arm/cpu.h @@ -XXX,XX +XXX,XX @@ bool write_cpustate_to_list(ARMCPU *cpu, bool kvm_sync); * EL2 (aka NS PL2) * EL3 (aka S PL1) * Physical (NS & S) + * Stage2 (NS & S) * - * for a total of 10 different mmu_idx. + * for a total of 12 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 @@ -XXX,XX +XXX,XX @@ typedef enum ARMMMUIdx { ARMMMUIdx_Phys_NS = 8 | ARM_MMU_IDX_A, ARMMMUIdx_Phys_S = 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 = 10 | ARM_MMU_IDX_A, + ARMMMUIdx_Stage2_S = 11 | 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. @@ -XXX,XX +XXX,XX @@ 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, - /* - * Not allocated a TLB: used only for second stage of an S12 page - * table walk, or for descriptor loads during first stage of an S1 - * page table walk. Note that if we ever want to have a TLB for this - * then various TLB flush insns which currently are no-ops or flush - * only stage 1 MMU indexes will need to change to flush stage 2. - */ - ARMMMUIdx_Stage2 = 3 | ARM_MMU_IDX_NOTLB, - ARMMMUIdx_Stage2_S = 4 | ARM_MMU_IDX_NOTLB, /* * M-profile. @@ -XXX,XX +XXX,XX @@ typedef enum ARMMMUIdxBit { TO_CORE_BIT(E20_2), TO_CORE_BIT(E20_2_PAN), TO_CORE_BIT(E3), + TO_CORE_BIT(Stage2), + TO_CORE_BIT(Stage2_S), TO_CORE_BIT(MUser), TO_CORE_BIT(MPriv), diff --git a/target/arm/helper.c b/target/arm/helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper.c +++ b/target/arm/helper.c @@ -XXX,XX +XXX,XX @@ static void contextidr_write(CPUARMState *env, const ARMCPRegInfo *ri, raw_write(env, ri, value); } +static int alle1_tlbmask(CPUARMState *env) +{ + /* + * Note that the 'ALL' scope must invalidate both stage 1 and + * stage 2 translations, whereas most other scopes only invalidate + * stage 1 translations. + */ + return (ARMMMUIdxBit_E10_1 | + ARMMMUIdxBit_E10_1_PAN | + ARMMMUIdxBit_E10_0 | + ARMMMUIdxBit_Stage2 | + ARMMMUIdxBit_Stage2_S); +} + + /* IS variants of TLB operations must affect all cores */ static void tlbiall_is_write(CPUARMState *env, const ARMCPRegInfo *ri, uint64_t value) @@ -XXX,XX +XXX,XX @@ static void tlbiall_nsnh_write(CPUARMState *env, const ARMCPRegInfo *ri, { CPUState *cs = env_cpu(env); - tlb_flush_by_mmuidx(cs, - ARMMMUIdxBit_E10_1 | - ARMMMUIdxBit_E10_1_PAN | - ARMMMUIdxBit_E10_0); + tlb_flush_by_mmuidx(cs, alle1_tlbmask(env)); } static void tlbiall_nsnh_is_write(CPUARMState *env, const ARMCPRegInfo *ri, @@ -XXX,XX +XXX,XX @@ static void tlbiall_nsnh_is_write(CPUARMState *env, const ARMCPRegInfo *ri, { CPUState *cs = env_cpu(env); - tlb_flush_by_mmuidx_all_cpus_synced(cs, - ARMMMUIdxBit_E10_1 | - ARMMMUIdxBit_E10_1_PAN | - ARMMMUIdxBit_E10_0); + tlb_flush_by_mmuidx_all_cpus_synced(cs, alle1_tlbmask(env)); } @@ -XXX,XX +XXX,XX @@ static void tlbimva_hyp_is_write(CPUARMState *env, const ARMCPRegInfo *ri, ARMMMUIdxBit_E2); } +static void tlbiipas2_hyp_write(CPUARMState *env, const ARMCPRegInfo *ri, + uint64_t value) +{ + CPUState *cs = env_cpu(env); + uint64_t pageaddr = (value & MAKE_64BIT_MASK(0, 28)) << 12; + + tlb_flush_page_by_mmuidx(cs, pageaddr, ARMMMUIdxBit_Stage2); +} + +static void tlbiipas2is_hyp_write(CPUARMState *env, const ARMCPRegInfo *ri, + uint64_t value) +{ + CPUState *cs = env_cpu(env); + uint64_t pageaddr = (value & MAKE_64BIT_MASK(0, 28)) << 12; + + tlb_flush_page_by_mmuidx_all_cpus_synced(cs, pageaddr, ARMMMUIdxBit_Stage2); +} + static const ARMCPRegInfo cp_reginfo[] = { /* Define the secure and non-secure FCSE identifier CP registers * separately because there is no secure bank in V8 (no _EL3). This allows @@ -XXX,XX +XXX,XX @@ static void vttbr_write(CPUARMState *env, const ARMCPRegInfo *ri, /* * A change in VMID to the stage2 page table (Stage2) invalidates - * the combined stage 1&2 tlbs (EL10_1 and EL10_0). + * the stage2 and combined stage 1&2 tlbs (EL10_1 and EL10_0). */ if (raw_read(env, ri) != value) { - uint16_t mask = ARMMMUIdxBit_E10_1 | - ARMMMUIdxBit_E10_1_PAN | - ARMMMUIdxBit_E10_0; - tlb_flush_by_mmuidx(cs, mask); + tlb_flush_by_mmuidx(cs, alle1_tlbmask(env)); raw_write(env, ri, value); } } @@ -XXX,XX +XXX,XX @@ static void tlbi_aa64_vmalle1_write(CPUARMState *env, const ARMCPRegInfo *ri, } } -static int alle1_tlbmask(CPUARMState *env) -{ - /* - * Note that the 'ALL' scope must invalidate both stage 1 and - * stage 2 translations, whereas most other scopes only invalidate - * stage 1 translations. - */ - return (ARMMMUIdxBit_E10_1 | - ARMMMUIdxBit_E10_1_PAN | - ARMMMUIdxBit_E10_0); -} - static int e2_tlbmask(CPUARMState *env) { return (ARMMMUIdxBit_E20_0 | @@ -XXX,XX +XXX,XX @@ static void tlbi_aa64_vae3is_write(CPUARMState *env, const ARMCPRegInfo *ri, ARMMMUIdxBit_E3, bits); } +static int ipas2e1_tlbmask(CPUARMState *env, int64_t value) +{ + /* + * The MSB of value is the NS field, which only applies if SEL2 + * is implemented and SCR_EL3.NS is not set (i.e. in secure mode). + */ + return (value >= 0 + && cpu_isar_feature(aa64_sel2, env_archcpu(env)) + && arm_is_secure_below_el3(env) + ? ARMMMUIdxBit_Stage2_S + : ARMMMUIdxBit_Stage2); +} + +static void tlbi_aa64_ipas2e1_write(CPUARMState *env, const ARMCPRegInfo *ri, + uint64_t value) +{ + CPUState *cs = env_cpu(env); + int mask = ipas2e1_tlbmask(env, value); + uint64_t pageaddr = sextract64(value << 12, 0, 56); + + if (tlb_force_broadcast(env)) { + tlb_flush_page_by_mmuidx_all_cpus_synced(cs, pageaddr, mask); + } else { + tlb_flush_page_by_mmuidx(cs, pageaddr, mask); + } +} + +static void tlbi_aa64_ipas2e1is_write(CPUARMState *env, const ARMCPRegInfo *ri, + uint64_t value) +{ + CPUState *cs = env_cpu(env); + int mask = ipas2e1_tlbmask(env, value); + uint64_t pageaddr = sextract64(value << 12, 0, 56); + + tlb_flush_page_by_mmuidx_all_cpus_synced(cs, pageaddr, mask); +} + #ifdef TARGET_AARCH64 typedef struct { uint64_t base; @@ -XXX,XX +XXX,XX @@ static void tlbi_aa64_rvae3is_write(CPUARMState *env, do_rvae_write(env, value, ARMMMUIdxBit_E3, true); } + +static void tlbi_aa64_ripas2e1_write(CPUARMState *env, const ARMCPRegInfo *ri, + uint64_t value) +{ + do_rvae_write(env, value, ipas2e1_tlbmask(env, value), + tlb_force_broadcast(env)); +} + +static void tlbi_aa64_ripas2e1is_write(CPUARMState *env, + const ARMCPRegInfo *ri, + uint64_t value) +{ + do_rvae_write(env, value, ipas2e1_tlbmask(env, value), true); +} #endif static CPAccessResult aa64_zva_access(CPUARMState *env, const ARMCPRegInfo *ri, @@ -XXX,XX +XXX,XX @@ static const ARMCPRegInfo v8_cp_reginfo[] = { .writefn = tlbi_aa64_vae1_write }, { .name = "TLBI_IPAS2E1IS", .state = ARM_CP_STATE_AA64, .opc0 = 1, .opc1 = 4, .crn = 8, .crm = 0, .opc2 = 1, - .access = PL2_W, .type = ARM_CP_NOP }, + .access = PL2_W, .type = ARM_CP_NO_RAW, + .writefn = tlbi_aa64_ipas2e1is_write }, { .name = "TLBI_IPAS2LE1IS", .state = ARM_CP_STATE_AA64, .opc0 = 1, .opc1 = 4, .crn = 8, .crm = 0, .opc2 = 5, - .access = PL2_W, .type = ARM_CP_NOP }, + .access = PL2_W, .type = ARM_CP_NO_RAW, + .writefn = tlbi_aa64_ipas2e1is_write }, { .name = "TLBI_ALLE1IS", .state = ARM_CP_STATE_AA64, .opc0 = 1, .opc1 = 4, .crn = 8, .crm = 3, .opc2 = 4, .access = PL2_W, .type = ARM_CP_NO_RAW, @@ -XXX,XX +XXX,XX @@ static const ARMCPRegInfo v8_cp_reginfo[] = { .writefn = tlbi_aa64_alle1is_write }, { .name = "TLBI_IPAS2E1", .state = ARM_CP_STATE_AA64, .opc0 = 1, .opc1 = 4, .crn = 8, .crm = 4, .opc2 = 1, - .access = PL2_W, .type = ARM_CP_NOP }, + .access = PL2_W, .type = ARM_CP_NO_RAW, + .writefn = tlbi_aa64_ipas2e1_write }, { .name = "TLBI_IPAS2LE1", .state = ARM_CP_STATE_AA64, .opc0 = 1, .opc1 = 4, .crn = 8, .crm = 4, .opc2 = 5, - .access = PL2_W, .type = ARM_CP_NOP }, + .access = PL2_W, .type = ARM_CP_NO_RAW, + .writefn = tlbi_aa64_ipas2e1_write }, { .name = "TLBI_ALLE1", .state = ARM_CP_STATE_AA64, .opc0 = 1, .opc1 = 4, .crn = 8, .crm = 7, .opc2 = 4, .access = PL2_W, .type = ARM_CP_NO_RAW, @@ -XXX,XX +XXX,XX @@ static const ARMCPRegInfo v8_cp_reginfo[] = { .writefn = tlbimva_hyp_is_write }, { .name = "TLBIIPAS2", .cp = 15, .opc1 = 4, .crn = 8, .crm = 4, .opc2 = 1, - .type = ARM_CP_NOP, .access = PL2_W }, + .type = ARM_CP_NO_RAW, .access = PL2_W, + .writefn = tlbiipas2_hyp_write }, { .name = "TLBIIPAS2IS", .cp = 15, .opc1 = 4, .crn = 8, .crm = 0, .opc2 = 1, - .type = ARM_CP_NOP, .access = PL2_W }, + .type = ARM_CP_NO_RAW, .access = PL2_W, + .writefn = tlbiipas2is_hyp_write }, { .name = "TLBIIPAS2L", .cp = 15, .opc1 = 4, .crn = 8, .crm = 4, .opc2 = 5, - .type = ARM_CP_NOP, .access = PL2_W }, + .type = ARM_CP_NO_RAW, .access = PL2_W, + .writefn = tlbiipas2_hyp_write }, { .name = "TLBIIPAS2LIS", .cp = 15, .opc1 = 4, .crn = 8, .crm = 0, .opc2 = 5, - .type = ARM_CP_NOP, .access = PL2_W }, + .type = ARM_CP_NO_RAW, .access = PL2_W, + .writefn = tlbiipas2is_hyp_write }, /* 32 bit cache operations */ { .name = "ICIALLUIS", .cp = 15, .opc1 = 0, .crn = 7, .crm = 1, .opc2 = 0, .type = ARM_CP_NOP, .access = PL1_W, .accessfn = aa64_cacheop_pou_access }, @@ -XXX,XX +XXX,XX @@ static const ARMCPRegInfo tlbirange_reginfo[] = { .writefn = tlbi_aa64_rvae1_write }, { .name = "TLBI_RIPAS2E1IS", .state = ARM_CP_STATE_AA64, .opc0 = 1, .opc1 = 4, .crn = 8, .crm = 0, .opc2 = 2, - .access = PL2_W, .type = ARM_CP_NOP }, + .access = PL2_W, .type = ARM_CP_NO_RAW, + .writefn = tlbi_aa64_ripas2e1is_write }, { .name = "TLBI_RIPAS2LE1IS", .state = ARM_CP_STATE_AA64, .opc0 = 1, .opc1 = 4, .crn = 8, .crm = 0, .opc2 = 6, - .access = PL2_W, .type = ARM_CP_NOP }, + .access = PL2_W, .type = ARM_CP_NO_RAW, + .writefn = tlbi_aa64_ripas2e1is_write }, { .name = "TLBI_RVAE2IS", .state = ARM_CP_STATE_AA64, .opc0 = 1, .opc1 = 4, .crn = 8, .crm = 2, .opc2 = 1, .access = PL2_W, .type = ARM_CP_NO_RAW | ARM_CP_EL3_NO_EL2_UNDEF, @@ -XXX,XX +XXX,XX @@ static const ARMCPRegInfo tlbirange_reginfo[] = { .writefn = tlbi_aa64_rvae2is_write }, { .name = "TLBI_RIPAS2E1", .state = ARM_CP_STATE_AA64, .opc0 = 1, .opc1 = 4, .crn = 8, .crm = 4, .opc2 = 2, - .access = PL2_W, .type = ARM_CP_NOP }, - { .name = "TLBI_RIPAS2LE1", .state = ARM_CP_STATE_AA64, + .access = PL2_W, .type = ARM_CP_NO_RAW, + .writefn = tlbi_aa64_ripas2e1_write }, + { .name = "TLBI_RIPAS2LE1", .state = ARM_CP_STATE_AA64, .opc0 = 1, .opc1 = 4, .crn = 8, .crm = 4, .opc2 = 6, - .access = PL2_W, .type = ARM_CP_NOP }, + .access = PL2_W, .type = ARM_CP_NO_RAW, + .writefn = tlbi_aa64_ripas2e1_write }, { .name = "TLBI_RVAE2OS", .state = ARM_CP_STATE_AA64, .opc0 = 1, .opc1 = 4, .crn = 8, .crm = 5, .opc2 = 1, .access = PL2_W, .type = ARM_CP_NO_RAW | ARM_CP_EL3_NO_EL2_UNDEF, -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Compare only the VMID field when considering whether we need to flush. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Message-id: 20221011031911.2408754-7-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/target/arm/helper.c b/target/arm/helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper.c +++ b/target/arm/helper.c @@ -XXX,XX +XXX,XX @@ static void vttbr_write(CPUARMState *env, const ARMCPRegInfo *ri, * A change in VMID to the stage2 page table (Stage2) invalidates * the stage2 and combined stage 1&2 tlbs (EL10_1 and EL10_0). */ - if (raw_read(env, ri) != value) { + if (extract64(raw_read(env, ri) ^ value, 48, 16) != 0) { tlb_flush_by_mmuidx(cs, alle1_tlbmask(env)); - raw_write(env, ri, value); } + raw_write(env, ri, value); } static const ARMCPRegInfo vmsa_pmsa_cp_reginfo[] = { -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Consolidate most of the inputs and outputs of S1_ptw_translate into a single structure. Plumb this through arm_ld*_ptw from the controlling get_phys_addr_* routine. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Message-id: 20221011031911.2408754-8-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/ptw.c | 140 ++++++++++++++++++++++++++--------------------- 1 file changed, 79 insertions(+), 61 deletions(-) diff --git a/target/arm/ptw.c b/target/arm/ptw.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/ptw.c +++ b/target/arm/ptw.c @@ -XXX,XX +XXX,XX @@ #include "idau.h" -static bool get_phys_addr_lpae(CPUARMState *env, uint64_t address, - MMUAccessType access_type, ARMMMUIdx mmu_idx, - bool is_secure, bool s1_is_el0, +typedef struct S1Translate { + ARMMMUIdx in_mmu_idx; + bool in_secure; + bool out_secure; + hwaddr out_phys; +} S1Translate; + +static bool get_phys_addr_lpae(CPUARMState *env, S1Translate *ptw, + uint64_t address, + MMUAccessType access_type, bool s1_is_el0, GetPhysAddrResult *result, ARMMMUFaultInfo *fi) __attribute__((nonnull)); @@ -XXX,XX +XXX,XX @@ static bool ptw_attrs_are_device(uint64_t hcr, ARMCacheAttrs cacheattrs) } /* Translate a S1 pagetable walk through S2 if needed. */ -static hwaddr S1_ptw_translate(CPUARMState *env, ARMMMUIdx mmu_idx, - hwaddr addr, bool *is_secure_ptr, - ARMMMUFaultInfo *fi) +static bool S1_ptw_translate(CPUARMState *env, S1Translate *ptw, + hwaddr addr, ARMMMUFaultInfo *fi) { - bool is_secure = *is_secure_ptr; + bool is_secure = ptw->in_secure; ARMMMUIdx s2_mmu_idx = is_secure ? ARMMMUIdx_Stage2_S : ARMMMUIdx_Stage2; - if (arm_mmu_idx_is_stage1_of_2(mmu_idx) && + if (arm_mmu_idx_is_stage1_of_2(ptw->in_mmu_idx) && !regime_translation_disabled(env, s2_mmu_idx, is_secure)) { GetPhysAddrResult s2 = {}; + S1Translate s2ptw = { + .in_mmu_idx = s2_mmu_idx, + .in_secure = is_secure, + }; uint64_t hcr; int ret; - ret = get_phys_addr_lpae(env, addr, MMU_DATA_LOAD, s2_mmu_idx, - is_secure, false, &s2, fi); + ret = get_phys_addr_lpae(env, &s2ptw, addr, MMU_DATA_LOAD, + false, &s2, fi); if (ret) { assert(fi->type != ARMFault_None); fi->s2addr = addr; fi->stage2 = true; fi->s1ptw = true; fi->s1ns = !is_secure; - return ~0; + return false; } hcr = arm_hcr_el2_eff_secstate(env, is_secure); @@ -XXX,XX +XXX,XX @@ static hwaddr S1_ptw_translate(CPUARMState *env, ARMMMUIdx mmu_idx, fi->stage2 = true; fi->s1ptw = true; fi->s1ns = !is_secure; - return ~0; + return false; } if (arm_is_secure_below_el3(env)) { @@ -XXX,XX +XXX,XX @@ static hwaddr S1_ptw_translate(CPUARMState *env, ARMMMUIdx mmu_idx, } else { is_secure = !(env->cp15.vtcr_el2 & VTCR_NSW); } - *is_secure_ptr = is_secure; } else { assert(!is_secure); } addr = s2.f.phys_addr; } - return addr; + + ptw->out_secure = is_secure; + ptw->out_phys = addr; + return true; } /* All loads done in the course of a page table walk go through here. */ -static uint32_t arm_ldl_ptw(CPUARMState *env, hwaddr addr, bool is_secure, - ARMMMUIdx mmu_idx, ARMMMUFaultInfo *fi) +static uint32_t arm_ldl_ptw(CPUARMState *env, S1Translate *ptw, hwaddr addr, + ARMMMUFaultInfo *fi) { CPUState *cs = env_cpu(env); MemTxAttrs attrs = {}; @@ -XXX,XX +XXX,XX @@ static uint32_t arm_ldl_ptw(CPUARMState *env, hwaddr addr, bool is_secure, AddressSpace *as; uint32_t data; - addr = S1_ptw_translate(env, mmu_idx, addr, &is_secure, fi); - attrs.secure = is_secure; - as = arm_addressspace(cs, attrs); - if (fi->s1ptw) { + if (!S1_ptw_translate(env, ptw, addr, fi)) { return 0; } - if (regime_translation_big_endian(env, mmu_idx)) { + addr = ptw->out_phys; + attrs.secure = ptw->out_secure; + as = arm_addressspace(cs, attrs); + if (regime_translation_big_endian(env, ptw->in_mmu_idx)) { data = address_space_ldl_be(as, addr, attrs, &result); } else { data = address_space_ldl_le(as, addr, attrs, &result); @@ -XXX,XX +XXX,XX @@ static uint32_t arm_ldl_ptw(CPUARMState *env, hwaddr addr, bool is_secure, return 0; } -static uint64_t arm_ldq_ptw(CPUARMState *env, hwaddr addr, bool is_secure, - ARMMMUIdx mmu_idx, ARMMMUFaultInfo *fi) +static uint64_t arm_ldq_ptw(CPUARMState *env, S1Translate *ptw, hwaddr addr, + ARMMMUFaultInfo *fi) { CPUState *cs = env_cpu(env); MemTxAttrs attrs = {}; @@ -XXX,XX +XXX,XX @@ static uint64_t arm_ldq_ptw(CPUARMState *env, hwaddr addr, bool is_secure, AddressSpace *as; uint64_t data; - addr = S1_ptw_translate(env, mmu_idx, addr, &is_secure, fi); - attrs.secure = is_secure; - as = arm_addressspace(cs, attrs); - if (fi->s1ptw) { + if (!S1_ptw_translate(env, ptw, addr, fi)) { return 0; } - if (regime_translation_big_endian(env, mmu_idx)) { + addr = ptw->out_phys; + attrs.secure = ptw->out_secure; + as = arm_addressspace(cs, attrs); + if (regime_translation_big_endian(env, ptw->in_mmu_idx)) { data = address_space_ldq_be(as, addr, attrs, &result); } else { data = address_space_ldq_le(as, addr, attrs, &result); @@ -XXX,XX +XXX,XX @@ 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)); } -static bool get_phys_addr_v5(CPUARMState *env, uint32_t address, - MMUAccessType access_type, ARMMMUIdx mmu_idx, - bool is_secure, GetPhysAddrResult *result, - ARMMMUFaultInfo *fi) +static bool get_phys_addr_v5(CPUARMState *env, S1Translate *ptw, + uint32_t address, MMUAccessType access_type, + GetPhysAddrResult *result, ARMMMUFaultInfo *fi) { int level = 1; uint32_t table; @@ -XXX,XX +XXX,XX @@ static bool get_phys_addr_v5(CPUARMState *env, uint32_t address, /* Pagetable walk. */ /* Lookup l1 descriptor. */ - if (!get_level1_table_address(env, mmu_idx, &table, address)) { + if (!get_level1_table_address(env, ptw->in_mmu_idx, &table, address)) { /* Section translation fault if page walk is disabled by PD0 or PD1 */ fi->type = ARMFault_Translation; goto do_fault; } - desc = arm_ldl_ptw(env, table, is_secure, mmu_idx, fi); + desc = arm_ldl_ptw(env, ptw, table, fi); if (fi->type != ARMFault_None) { goto do_fault; } type = (desc & 3); domain = (desc >> 5) & 0x0f; - if (regime_el(env, mmu_idx) == 1) { + if (regime_el(env, ptw->in_mmu_idx) == 1) { dacr = env->cp15.dacr_ns; } else { dacr = env->cp15.dacr_s; @@ -XXX,XX +XXX,XX @@ static bool get_phys_addr_v5(CPUARMState *env, uint32_t address, /* Fine pagetable. */ table = (desc & 0xfffff000) | ((address >> 8) & 0xffc); } - desc = arm_ldl_ptw(env, table, is_secure, mmu_idx, fi); + desc = arm_ldl_ptw(env, ptw, table, fi); if (fi->type != ARMFault_None) { goto do_fault; } @@ -XXX,XX +XXX,XX @@ static bool get_phys_addr_v5(CPUARMState *env, uint32_t address, g_assert_not_reached(); } } - result->f.prot = ap_to_rw_prot(env, mmu_idx, ap, domain_prot); + result->f.prot = ap_to_rw_prot(env, ptw->in_mmu_idx, ap, domain_prot); result->f.prot |= result->f.prot ? PAGE_EXEC : 0; if (!(result->f.prot & (1 << access_type))) { /* Access permission fault. */ @@ -XXX,XX +XXX,XX @@ do_fault: return true; } -static bool get_phys_addr_v6(CPUARMState *env, uint32_t address, - MMUAccessType access_type, ARMMMUIdx mmu_idx, - bool is_secure, GetPhysAddrResult *result, - ARMMMUFaultInfo *fi) +static bool get_phys_addr_v6(CPUARMState *env, S1Translate *ptw, + uint32_t address, MMUAccessType access_type, + GetPhysAddrResult *result, ARMMMUFaultInfo *fi) { ARMCPU *cpu = env_archcpu(env); + ARMMMUIdx mmu_idx = ptw->in_mmu_idx; int level = 1; uint32_t table; uint32_t desc; @@ -XXX,XX +XXX,XX @@ static bool get_phys_addr_v6(CPUARMState *env, uint32_t address, fi->type = ARMFault_Translation; goto do_fault; } - desc = arm_ldl_ptw(env, table, is_secure, mmu_idx, fi); + desc = arm_ldl_ptw(env, ptw, table, fi); if (fi->type != ARMFault_None) { goto do_fault; } @@ -XXX,XX +XXX,XX @@ static bool get_phys_addr_v6(CPUARMState *env, uint32_t address, ns = extract32(desc, 3, 1); /* Lookup l2 entry. */ table = (desc & 0xfffffc00) | ((address >> 10) & 0x3fc); - desc = arm_ldl_ptw(env, table, is_secure, mmu_idx, fi); + desc = arm_ldl_ptw(env, ptw, table, fi); if (fi->type != ARMFault_None) { goto do_fault; } @@ -XXX,XX +XXX,XX @@ static bool check_s2_mmu_setup(ARMCPU *cpu, bool is_aa64, int level, * the WnR bit is never set (the caller must do this). * * @env: CPUARMState + * @ptw: Current and next stage parameters for the walk. * @address: virtual address to get physical address for * @access_type: MMU_DATA_LOAD, MMU_DATA_STORE or MMU_INST_FETCH - * @mmu_idx: MMU index indicating required translation regime - * @s1_is_el0: if @mmu_idx is ARMMMUIdx_Stage2 (so this is a stage 2 page - * table walk), must be true if this is stage 2 of a stage 1+2 + * @s1_is_el0: if @ptw->in_mmu_idx is ARMMMUIdx_Stage2 + * (so this is a stage 2 page table walk), + * must be true if this is stage 2 of a stage 1+2 * walk for an EL0 access. If @mmu_idx is anything else, * @s1_is_el0 is ignored. * @result: set on translation success, * @fi: set to fault info if the translation fails */ -static bool get_phys_addr_lpae(CPUARMState *env, uint64_t address, - MMUAccessType access_type, ARMMMUIdx mmu_idx, - bool is_secure, bool s1_is_el0, +static bool get_phys_addr_lpae(CPUARMState *env, S1Translate *ptw, + uint64_t address, + MMUAccessType access_type, bool s1_is_el0, GetPhysAddrResult *result, ARMMMUFaultInfo *fi) { ARMCPU *cpu = env_archcpu(env); + ARMMMUIdx mmu_idx = ptw->in_mmu_idx; + bool is_secure = ptw->in_secure; /* Read an LPAE long-descriptor translation table. */ ARMFaultType fault_type = ARMFault_Translation; uint32_t level; @@ -XXX,XX +XXX,XX @@ static bool get_phys_addr_lpae(CPUARMState *env, uint64_t address, descaddr |= (address >> (stride * (4 - level))) & indexmask; descaddr &= ~7ULL; nstable = extract32(tableattrs, 4, 1); - descriptor = arm_ldq_ptw(env, descaddr, !nstable, mmu_idx, fi); + ptw->in_secure = !nstable; + descriptor = arm_ldq_ptw(env, ptw, descaddr, fi); if (fi->type != ARMFault_None) { goto do_fault; } @@ -XXX,XX +XXX,XX @@ bool get_phys_addr_with_secure(CPUARMState *env, target_ulong address, ARMMMUFaultInfo *fi) { ARMMMUIdx s1_mmu_idx = stage_1_mmu_idx(mmu_idx); + S1Translate ptw; if (mmu_idx != s1_mmu_idx) { /* @@ -XXX,XX +XXX,XX @@ bool get_phys_addr_with_secure(CPUARMState *env, target_ulong address, int ret; bool ipa_secure, s2walk_secure; ARMCacheAttrs cacheattrs1; - ARMMMUIdx s2_mmu_idx; bool is_el0; uint64_t hcr; @@ -XXX,XX +XXX,XX @@ bool get_phys_addr_with_secure(CPUARMState *env, target_ulong address, s2walk_secure = false; } - s2_mmu_idx = (s2walk_secure - ? ARMMMUIdx_Stage2_S : ARMMMUIdx_Stage2); + ptw.in_mmu_idx = + s2walk_secure ? ARMMMUIdx_Stage2_S : ARMMMUIdx_Stage2; + ptw.in_secure = s2walk_secure; is_el0 = mmu_idx == ARMMMUIdx_E10_0; /* @@ -XXX,XX +XXX,XX @@ bool get_phys_addr_with_secure(CPUARMState *env, target_ulong address, cacheattrs1 = result->cacheattrs; memset(result, 0, sizeof(*result)); - ret = get_phys_addr_lpae(env, ipa, access_type, s2_mmu_idx, - s2walk_secure, is_el0, result, fi); + ret = get_phys_addr_lpae(env, &ptw, ipa, access_type, + is_el0, result, fi); fi->s2addr = ipa; /* Combine the S1 and S2 perms. */ @@ -XXX,XX +XXX,XX @@ bool get_phys_addr_with_secure(CPUARMState *env, target_ulong address, return get_phys_addr_disabled(env, address, access_type, mmu_idx, is_secure, result, fi); } + + ptw.in_mmu_idx = mmu_idx; + ptw.in_secure = is_secure; + if (regime_using_lpae_format(env, mmu_idx)) { - return get_phys_addr_lpae(env, address, access_type, mmu_idx, - is_secure, false, result, fi); + return get_phys_addr_lpae(env, &ptw, address, access_type, false, + result, fi); } else if (regime_sctlr(env, mmu_idx) & SCTLR_XP) { - return get_phys_addr_v6(env, address, access_type, mmu_idx, - is_secure, result, fi); + return get_phys_addr_v6(env, &ptw, address, access_type, result, fi); } else { - return get_phys_addr_v5(env, address, access_type, mmu_idx, - is_secure, result, fi); + return get_phys_addr_v5(env, &ptw, address, access_type, result, fi); } } -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Before using softmmu page tables for the ptw, plumb down a debug parameter so that we can query page table entries from gdbstub without modifying cpu state. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20221011031911.2408754-9-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/ptw.c | 55 ++++++++++++++++++++++++++++++++---------------- 1 file changed, 37 insertions(+), 18 deletions(-) diff --git a/target/arm/ptw.c b/target/arm/ptw.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/ptw.c +++ b/target/arm/ptw.c @@ -XXX,XX +XXX,XX @@ typedef struct S1Translate { ARMMMUIdx in_mmu_idx; bool in_secure; + bool in_debug; bool out_secure; hwaddr out_phys; } S1Translate; @@ -XXX,XX +XXX,XX @@ static bool S1_ptw_translate(CPUARMState *env, S1Translate *ptw, S1Translate s2ptw = { .in_mmu_idx = s2_mmu_idx, .in_secure = is_secure, + .in_debug = ptw->in_debug, }; uint64_t hcr; int ret; @@ -XXX,XX +XXX,XX @@ static bool get_phys_addr_disabled(CPUARMState *env, target_ulong address, return 0; } -bool get_phys_addr_with_secure(CPUARMState *env, target_ulong address, - MMUAccessType access_type, ARMMMUIdx mmu_idx, - bool is_secure, GetPhysAddrResult *result, - ARMMMUFaultInfo *fi) +static bool get_phys_addr_with_struct(CPUARMState *env, S1Translate *ptw, + target_ulong address, + MMUAccessType access_type, + GetPhysAddrResult *result, + ARMMMUFaultInfo *fi) { + ARMMMUIdx mmu_idx = ptw->in_mmu_idx; ARMMMUIdx s1_mmu_idx = stage_1_mmu_idx(mmu_idx); - S1Translate ptw; + bool is_secure = ptw->in_secure; if (mmu_idx != s1_mmu_idx) { /* @@ -XXX,XX +XXX,XX @@ bool get_phys_addr_with_secure(CPUARMState *env, target_ulong address, bool is_el0; uint64_t hcr; - ret = get_phys_addr_with_secure(env, address, access_type, - s1_mmu_idx, is_secure, result, fi); + ptw->in_mmu_idx = s1_mmu_idx; + ret = get_phys_addr_with_struct(env, ptw, address, access_type, + result, fi); /* If S1 fails or S2 is disabled, return early. */ if (ret || regime_translation_disabled(env, ARMMMUIdx_Stage2, @@ -XXX,XX +XXX,XX @@ bool get_phys_addr_with_secure(CPUARMState *env, target_ulong address, s2walk_secure = false; } - ptw.in_mmu_idx = + ptw->in_mmu_idx = s2walk_secure ? ARMMMUIdx_Stage2_S : ARMMMUIdx_Stage2; - ptw.in_secure = s2walk_secure; + ptw->in_secure = s2walk_secure; is_el0 = mmu_idx == ARMMMUIdx_E10_0; /* @@ -XXX,XX +XXX,XX @@ bool get_phys_addr_with_secure(CPUARMState *env, target_ulong address, cacheattrs1 = result->cacheattrs; memset(result, 0, sizeof(*result)); - ret = get_phys_addr_lpae(env, &ptw, ipa, access_type, + ret = get_phys_addr_lpae(env, ptw, ipa, access_type, is_el0, result, fi); fi->s2addr = ipa; @@ -XXX,XX +XXX,XX @@ bool get_phys_addr_with_secure(CPUARMState *env, target_ulong address, is_secure, result, fi); } - ptw.in_mmu_idx = mmu_idx; - ptw.in_secure = is_secure; - if (regime_using_lpae_format(env, mmu_idx)) { - return get_phys_addr_lpae(env, &ptw, address, access_type, false, + return get_phys_addr_lpae(env, ptw, address, access_type, false, result, fi); } else if (regime_sctlr(env, mmu_idx) & SCTLR_XP) { - return get_phys_addr_v6(env, &ptw, address, access_type, result, fi); + return get_phys_addr_v6(env, ptw, address, access_type, result, fi); } else { - return get_phys_addr_v5(env, &ptw, address, access_type, result, fi); + return get_phys_addr_v5(env, ptw, address, access_type, result, fi); } } +bool get_phys_addr_with_secure(CPUARMState *env, target_ulong address, + MMUAccessType access_type, ARMMMUIdx mmu_idx, + bool is_secure, GetPhysAddrResult *result, + ARMMMUFaultInfo *fi) +{ + S1Translate ptw = { + .in_mmu_idx = mmu_idx, + .in_secure = is_secure, + }; + return get_phys_addr_with_struct(env, &ptw, address, access_type, + result, fi); +} + bool get_phys_addr(CPUARMState *env, target_ulong address, MMUAccessType access_type, ARMMMUIdx mmu_idx, GetPhysAddrResult *result, ARMMMUFaultInfo *fi) @@ -XXX,XX +XXX,XX @@ hwaddr arm_cpu_get_phys_page_attrs_debug(CPUState *cs, vaddr addr, { ARMCPU *cpu = ARM_CPU(cs); CPUARMState *env = &cpu->env; + S1Translate ptw = { + .in_mmu_idx = arm_mmu_idx(env), + .in_secure = arm_is_secure(env), + .in_debug = true, + }; GetPhysAddrResult res = {}; ARMMMUFaultInfo fi = {}; - ARMMMUIdx mmu_idx = arm_mmu_idx(env); bool ret; - ret = get_phys_addr(env, addr, MMU_DATA_LOAD, mmu_idx, &res, &fi); + ret = get_phys_addr_with_struct(env, &ptw, addr, MMU_DATA_LOAD, &res, &fi); *attrs = res.f.attrs; if (ret) { -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Hoist this test out of arm_ld[lq]_ptw into S1_ptw_translate. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20221011031911.2408754-10-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/ptw.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/target/arm/ptw.c b/target/arm/ptw.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/ptw.c +++ b/target/arm/ptw.c @@ -XXX,XX +XXX,XX @@ typedef struct S1Translate { bool in_secure; bool in_debug; bool out_secure; + bool out_be; hwaddr out_phys; } S1Translate; @@ -XXX,XX +XXX,XX @@ static bool S1_ptw_translate(CPUARMState *env, S1Translate *ptw, ptw->out_secure = is_secure; ptw->out_phys = addr; + ptw->out_be = regime_translation_big_endian(env, ptw->in_mmu_idx); return true; } @@ -XXX,XX +XXX,XX @@ static uint32_t arm_ldl_ptw(CPUARMState *env, S1Translate *ptw, hwaddr addr, addr = ptw->out_phys; attrs.secure = ptw->out_secure; as = arm_addressspace(cs, attrs); - if (regime_translation_big_endian(env, ptw->in_mmu_idx)) { + if (ptw->out_be) { data = address_space_ldl_be(as, addr, attrs, &result); } else { data = address_space_ldl_le(as, addr, attrs, &result); @@ -XXX,XX +XXX,XX @@ static uint64_t arm_ldq_ptw(CPUARMState *env, S1Translate *ptw, hwaddr addr, addr = ptw->out_phys; attrs.secure = ptw->out_secure; as = arm_addressspace(cs, attrs); - if (regime_translation_big_endian(env, ptw->in_mmu_idx)) { + if (ptw->out_be) { data = address_space_ldq_be(as, addr, attrs, &result); } else { data = address_space_ldq_le(as, addr, attrs, &result); -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> So far, limit the change to S1_ptw_translate, arm_ldl_ptw, and arm_ldq_ptw. Use probe_access_full to find the host address, and if so use a host load. If the probe fails, we've got our fault info already. On the off chance that page tables are not in RAM, continue to use the address_space_ld* functions. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20221011031911.2408754-11-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/cpu.h | 5 + target/arm/ptw.c | 196 +++++++++++++++++++++++++--------------- target/arm/tlb_helper.c | 17 +++- 3 files changed, 144 insertions(+), 74 deletions(-) diff --git a/target/arm/cpu.h b/target/arm/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.h +++ b/target/arm/cpu.h @@ -XXX,XX +XXX,XX @@ typedef struct CPUARMTBFlags { target_ulong flags2; } CPUARMTBFlags; +typedef struct ARMMMUFaultInfo ARMMMUFaultInfo; + typedef struct CPUArchState { /* Regs for current mode. */ uint32_t regs[16]; @@ -XXX,XX +XXX,XX @@ typedef struct CPUArchState { struct CPUBreakpoint *cpu_breakpoint[16]; struct CPUWatchpoint *cpu_watchpoint[16]; + /* Optional fault info across tlb lookup. */ + ARMMMUFaultInfo *tlb_fi; + /* Fields up to this point are cleared by a CPU reset */ struct {} end_reset_fields; diff --git a/target/arm/ptw.c b/target/arm/ptw.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/ptw.c +++ b/target/arm/ptw.c @@ -XXX,XX +XXX,XX @@ #include "qemu/osdep.h" #include "qemu/log.h" #include "qemu/range.h" +#include "exec/exec-all.h" #include "cpu.h" #include "internals.h" #include "idau.h" @@ -XXX,XX +XXX,XX @@ typedef struct S1Translate { bool out_secure; bool out_be; hwaddr out_phys; + void *out_host; } S1Translate; static bool get_phys_addr_lpae(CPUARMState *env, S1Translate *ptw, @@ -XXX,XX +XXX,XX @@ static bool regime_translation_disabled(CPUARMState *env, ARMMMUIdx mmu_idx, return (regime_sctlr(env, mmu_idx) & SCTLR_M) == 0; } -static bool ptw_attrs_are_device(uint64_t hcr, ARMCacheAttrs cacheattrs) +static bool S2_attrs_are_device(uint64_t hcr, uint8_t attrs) { /* * For an S1 page table walk, the stage 1 attributes are always @@ -XXX,XX +XXX,XX @@ static bool ptw_attrs_are_device(uint64_t hcr, ARMCacheAttrs cacheattrs) * With HCR_EL2.FWB == 1 this is when descriptor bit [4] is 0, ie * when cacheattrs.attrs bit [2] is 0. */ - assert(cacheattrs.is_s2_format); if (hcr & HCR_FWB) { - return (cacheattrs.attrs & 0x4) == 0; + return (attrs & 0x4) == 0; } else { - return (cacheattrs.attrs & 0xc) == 0; + return (attrs & 0xc) == 0; } } @@ -XXX,XX +XXX,XX @@ static bool S1_ptw_translate(CPUARMState *env, S1Translate *ptw, hwaddr addr, ARMMMUFaultInfo *fi) { bool is_secure = ptw->in_secure; + ARMMMUIdx mmu_idx = ptw->in_mmu_idx; ARMMMUIdx s2_mmu_idx = is_secure ? ARMMMUIdx_Stage2_S : ARMMMUIdx_Stage2; + bool s2_phys = false; + uint8_t pte_attrs; + bool pte_secure; - if (arm_mmu_idx_is_stage1_of_2(ptw->in_mmu_idx) && - !regime_translation_disabled(env, s2_mmu_idx, is_secure)) { - GetPhysAddrResult s2 = {}; - S1Translate s2ptw = { - .in_mmu_idx = s2_mmu_idx, - .in_secure = is_secure, - .in_debug = ptw->in_debug, - }; - uint64_t hcr; - int ret; + if (!arm_mmu_idx_is_stage1_of_2(mmu_idx) + || regime_translation_disabled(env, s2_mmu_idx, is_secure)) { + s2_mmu_idx = is_secure ? ARMMMUIdx_Phys_S : ARMMMUIdx_Phys_NS; + s2_phys = true; + } - ret = get_phys_addr_lpae(env, &s2ptw, addr, MMU_DATA_LOAD, - false, &s2, fi); - if (ret) { - assert(fi->type != ARMFault_None); - fi->s2addr = addr; - fi->stage2 = true; - fi->s1ptw = true; - fi->s1ns = !is_secure; - return false; + if (unlikely(ptw->in_debug)) { + /* + * From gdbstub, do not use softmmu so that we don't modify the + * state of the cpu at all, including softmmu tlb contents. + */ + if (s2_phys) { + ptw->out_phys = addr; + pte_attrs = 0; + pte_secure = is_secure; + } else { + S1Translate s2ptw = { + .in_mmu_idx = s2_mmu_idx, + .in_secure = is_secure, + .in_debug = true, + }; + GetPhysAddrResult s2 = { }; + if (!get_phys_addr_lpae(env, &s2ptw, addr, MMU_DATA_LOAD, + false, &s2, fi)) { + goto fail; + } + ptw->out_phys = s2.f.phys_addr; + pte_attrs = s2.cacheattrs.attrs; + pte_secure = s2.f.attrs.secure; } + ptw->out_host = NULL; + } else { + CPUTLBEntryFull *full; + int flags; - hcr = arm_hcr_el2_eff_secstate(env, is_secure); - if ((hcr & HCR_PTW) && ptw_attrs_are_device(hcr, s2.cacheattrs)) { + env->tlb_fi = fi; + flags = probe_access_full(env, addr, MMU_DATA_LOAD, + arm_to_core_mmu_idx(s2_mmu_idx), + true, &ptw->out_host, &full, 0); + env->tlb_fi = NULL; + + if (unlikely(flags & TLB_INVALID_MASK)) { + goto fail; + } + ptw->out_phys = full->phys_addr; + pte_attrs = full->pte_attrs; + pte_secure = full->attrs.secure; + } + + if (!s2_phys) { + uint64_t hcr = arm_hcr_el2_eff_secstate(env, is_secure); + + if ((hcr & HCR_PTW) && S2_attrs_are_device(hcr, pte_attrs)) { /* * PTW set and S1 walk touched S2 Device memory: * generate Permission fault. @@ -XXX,XX +XXX,XX @@ static bool S1_ptw_translate(CPUARMState *env, S1Translate *ptw, fi->s1ns = !is_secure; return false; } - - if (arm_is_secure_below_el3(env)) { - /* Check if page table walk is to secure or non-secure PA space. */ - if (is_secure) { - is_secure = !(env->cp15.vstcr_el2 & VSTCR_SW); - } else { - is_secure = !(env->cp15.vtcr_el2 & VTCR_NSW); - } - } else { - assert(!is_secure); - } - - addr = s2.f.phys_addr; } - ptw->out_secure = is_secure; - ptw->out_phys = addr; - ptw->out_be = regime_translation_big_endian(env, ptw->in_mmu_idx); + /* Check if page table walk is to secure or non-secure PA space. */ + ptw->out_secure = (is_secure + && !(pte_secure + ? env->cp15.vstcr_el2 & VSTCR_SW + : env->cp15.vtcr_el2 & VTCR_NSW)); + ptw->out_be = regime_translation_big_endian(env, mmu_idx); return true; + + fail: + assert(fi->type != ARMFault_None); + fi->s2addr = addr; + fi->stage2 = true; + fi->s1ptw = true; + fi->s1ns = !is_secure; + return false; } /* All loads done in the course of a page table walk go through here. */ @@ -XXX,XX +XXX,XX @@ static uint32_t arm_ldl_ptw(CPUARMState *env, S1Translate *ptw, hwaddr addr, ARMMMUFaultInfo *fi) { CPUState *cs = env_cpu(env); - MemTxAttrs attrs = {}; - MemTxResult result = MEMTX_OK; - AddressSpace *as; uint32_t data; if (!S1_ptw_translate(env, ptw, addr, fi)) { + /* Failure. */ + assert(fi->s1ptw); return 0; } - addr = ptw->out_phys; - attrs.secure = ptw->out_secure; - as = arm_addressspace(cs, attrs); - if (ptw->out_be) { - data = address_space_ldl_be(as, addr, attrs, &result); + + if (likely(ptw->out_host)) { + /* Page tables are in RAM, and we have the host address. */ + if (ptw->out_be) { + data = ldl_be_p(ptw->out_host); + } else { + data = ldl_le_p(ptw->out_host); + } } else { - data = address_space_ldl_le(as, addr, attrs, &result); + /* Page tables are in MMIO. */ + MemTxAttrs attrs = { .secure = ptw->out_secure }; + AddressSpace *as = arm_addressspace(cs, attrs); + MemTxResult result = MEMTX_OK; + + if (ptw->out_be) { + data = address_space_ldl_be(as, ptw->out_phys, attrs, &result); + } else { + data = address_space_ldl_le(as, ptw->out_phys, attrs, &result); + } + if (unlikely(result != MEMTX_OK)) { + fi->type = ARMFault_SyncExternalOnWalk; + fi->ea = arm_extabort_type(result); + return 0; + } } - if (result == MEMTX_OK) { - return data; - } - fi->type = ARMFault_SyncExternalOnWalk; - fi->ea = arm_extabort_type(result); - return 0; + return data; } static uint64_t arm_ldq_ptw(CPUARMState *env, S1Translate *ptw, hwaddr addr, ARMMMUFaultInfo *fi) { CPUState *cs = env_cpu(env); - MemTxAttrs attrs = {}; - MemTxResult result = MEMTX_OK; - AddressSpace *as; uint64_t data; if (!S1_ptw_translate(env, ptw, addr, fi)) { + /* Failure. */ + assert(fi->s1ptw); return 0; } - addr = ptw->out_phys; - attrs.secure = ptw->out_secure; - as = arm_addressspace(cs, attrs); - if (ptw->out_be) { - data = address_space_ldq_be(as, addr, attrs, &result); + + if (likely(ptw->out_host)) { + /* Page tables are in RAM, and we have the host address. */ + if (ptw->out_be) { + data = ldq_be_p(ptw->out_host); + } else { + data = ldq_le_p(ptw->out_host); + } } else { - data = address_space_ldq_le(as, addr, attrs, &result); + /* Page tables are in MMIO. */ + MemTxAttrs attrs = { .secure = ptw->out_secure }; + AddressSpace *as = arm_addressspace(cs, attrs); + MemTxResult result = MEMTX_OK; + + if (ptw->out_be) { + data = address_space_ldq_be(as, ptw->out_phys, attrs, &result); + } else { + data = address_space_ldq_le(as, ptw->out_phys, attrs, &result); + } + if (unlikely(result != MEMTX_OK)) { + fi->type = ARMFault_SyncExternalOnWalk; + fi->ea = arm_extabort_type(result); + return 0; + } } - if (result == MEMTX_OK) { - return data; - } - fi->type = ARMFault_SyncExternalOnWalk; - fi->ea = arm_extabort_type(result); - return 0; + return data; } static bool get_level1_table_address(CPUARMState *env, ARMMMUIdx mmu_idx, diff --git a/target/arm/tlb_helper.c b/target/arm/tlb_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/tlb_helper.c +++ b/target/arm/tlb_helper.c @@ -XXX,XX +XXX,XX @@ bool arm_cpu_tlb_fill(CPUState *cs, vaddr address, int size, bool probe, uintptr_t retaddr) { ARMCPU *cpu = ARM_CPU(cs); - ARMMMUFaultInfo fi = {}; GetPhysAddrResult res = {}; + ARMMMUFaultInfo local_fi, *fi; int ret; + /* + * Allow S1_ptw_translate to see any fault generated here. + * Since this may recurse, read and clear. + */ + fi = cpu->env.tlb_fi; + if (fi) { + cpu->env.tlb_fi = NULL; + } else { + fi = memset(&local_fi, 0, sizeof(local_fi)); + } + /* * Walk the page table and (if the mapping exists) add the page * to the TLB. On success, return true. Otherwise, if probing, @@ -XXX,XX +XXX,XX @@ bool arm_cpu_tlb_fill(CPUState *cs, vaddr address, int size, */ ret = get_phys_addr(&cpu->env, address, access_type, core_to_arm_mmu_idx(&cpu->env, mmu_idx), - &res, &fi); + &res, fi); if (likely(!ret)) { /* * Map a single [sub]page. Regions smaller than our declared @@ -XXX,XX +XXX,XX @@ bool arm_cpu_tlb_fill(CPUState *cs, vaddr address, int size, } else { /* now we have a real cpu fault */ cpu_restore_state(cs, retaddr, true); - arm_deliver_fault(cpu, address, access_type, mmu_idx, &fi); + arm_deliver_fault(cpu, address, access_type, mmu_idx, fi); } } #else -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20221011031911.2408754-12-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/ptw.c | 191 +++++++++++++++++++++++++---------------------- 1 file changed, 100 insertions(+), 91 deletions(-) diff --git a/target/arm/ptw.c b/target/arm/ptw.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/ptw.c +++ b/target/arm/ptw.c @@ -XXX,XX +XXX,XX @@ static bool get_phys_addr_lpae(CPUARMState *env, S1Translate *ptw, GetPhysAddrResult *result, ARMMMUFaultInfo *fi) __attribute__((nonnull)); +static bool get_phys_addr_with_struct(CPUARMState *env, S1Translate *ptw, + target_ulong address, + MMUAccessType access_type, + GetPhysAddrResult *result, + ARMMMUFaultInfo *fi) + __attribute__((nonnull)); + /* This mapping is common between ID_AA64MMFR0.PARANGE and TCR_ELx.{I}PS. */ static const uint8_t pamax_map[] = { [0] = 32, @@ -XXX,XX +XXX,XX @@ static bool get_phys_addr_disabled(CPUARMState *env, target_ulong address, return 0; } +static bool get_phys_addr_twostage(CPUARMState *env, S1Translate *ptw, + target_ulong address, + MMUAccessType access_type, + GetPhysAddrResult *result, + ARMMMUFaultInfo *fi) +{ + hwaddr ipa; + int s1_prot; + int ret; + bool is_secure = ptw->in_secure; + bool ipa_secure, s2walk_secure; + ARMCacheAttrs cacheattrs1; + bool is_el0; + uint64_t hcr; + + ret = get_phys_addr_with_struct(env, ptw, address, access_type, result, fi); + + /* If S1 fails or S2 is disabled, return early. */ + if (ret || regime_translation_disabled(env, ARMMMUIdx_Stage2, is_secure)) { + return ret; + } + + ipa = result->f.phys_addr; + ipa_secure = result->f.attrs.secure; + if (is_secure) { + /* Select TCR based on the NS bit from the S1 walk. */ + s2walk_secure = !(ipa_secure + ? env->cp15.vstcr_el2 & VSTCR_SW + : env->cp15.vtcr_el2 & VTCR_NSW); + } else { + assert(!ipa_secure); + s2walk_secure = false; + } + + is_el0 = ptw->in_mmu_idx == ARMMMUIdx_Stage1_E0; + ptw->in_mmu_idx = s2walk_secure ? ARMMMUIdx_Stage2_S : ARMMMUIdx_Stage2; + ptw->in_secure = s2walk_secure; + + /* + * S1 is done, now do S2 translation. + * Save the stage1 results so that we may merge prot and cacheattrs later. + */ + s1_prot = result->f.prot; + cacheattrs1 = result->cacheattrs; + memset(result, 0, sizeof(*result)); + + ret = get_phys_addr_lpae(env, ptw, ipa, access_type, is_el0, result, fi); + fi->s2addr = ipa; + + /* Combine the S1 and S2 perms. */ + result->f.prot &= s1_prot; + + /* If S2 fails, return early. */ + if (ret) { + return ret; + } + + /* Combine the S1 and S2 cache attributes. */ + hcr = arm_hcr_el2_eff_secstate(env, is_secure); + if (hcr & HCR_DC) { + /* + * HCR.DC forces the first stage attributes to + * Normal Non-Shareable, + * Inner Write-Back Read-Allocate Write-Allocate, + * Outer Write-Back Read-Allocate Write-Allocate. + * Do not overwrite Tagged within attrs. + */ + if (cacheattrs1.attrs != 0xf0) { + cacheattrs1.attrs = 0xff; + } + cacheattrs1.shareability = 0; + } + result->cacheattrs = combine_cacheattrs(hcr, cacheattrs1, + result->cacheattrs); + + /* + * Check if IPA translates to secure or non-secure PA space. + * Note that VSTCR overrides VTCR and {N}SW overrides {N}SA. + */ + result->f.attrs.secure = + (is_secure + && !(env->cp15.vstcr_el2 & (VSTCR_SA | VSTCR_SW)) + && (ipa_secure + || !(env->cp15.vtcr_el2 & (VTCR_NSA | VTCR_NSW)))); + + return 0; +} + static bool get_phys_addr_with_struct(CPUARMState *env, S1Translate *ptw, target_ulong address, MMUAccessType access_type, @@ -XXX,XX +XXX,XX @@ static bool get_phys_addr_with_struct(CPUARMState *env, S1Translate *ptw, if (mmu_idx != s1_mmu_idx) { /* * Call ourselves recursively to do the stage 1 and then stage 2 - * translations if mmu_idx is a two-stage regime. + * translations if mmu_idx is a two-stage regime, and EL2 present. + * Otherwise, a stage1+stage2 translation is just stage 1. */ + ptw->in_mmu_idx = mmu_idx = s1_mmu_idx; if (arm_feature(env, ARM_FEATURE_EL2)) { - hwaddr ipa; - int s1_prot; - int ret; - bool ipa_secure, s2walk_secure; - ARMCacheAttrs cacheattrs1; - bool is_el0; - uint64_t hcr; - - ptw->in_mmu_idx = s1_mmu_idx; - ret = get_phys_addr_with_struct(env, ptw, address, access_type, - result, fi); - - /* If S1 fails or S2 is disabled, return early. */ - if (ret || regime_translation_disabled(env, ARMMMUIdx_Stage2, - is_secure)) { - return ret; - } - - ipa = result->f.phys_addr; - ipa_secure = result->f.attrs.secure; - if (is_secure) { - /* Select TCR based on the NS bit from the S1 walk. */ - s2walk_secure = !(ipa_secure - ? env->cp15.vstcr_el2 & VSTCR_SW - : env->cp15.vtcr_el2 & VTCR_NSW); - } else { - assert(!ipa_secure); - s2walk_secure = false; - } - - ptw->in_mmu_idx = - s2walk_secure ? ARMMMUIdx_Stage2_S : ARMMMUIdx_Stage2; - ptw->in_secure = s2walk_secure; - is_el0 = mmu_idx == ARMMMUIdx_E10_0; - - /* - * S1 is done, now do S2 translation. - * Save the stage1 results so that we may merge - * prot and cacheattrs later. - */ - s1_prot = result->f.prot; - cacheattrs1 = result->cacheattrs; - memset(result, 0, sizeof(*result)); - - ret = get_phys_addr_lpae(env, ptw, ipa, access_type, - is_el0, result, fi); - fi->s2addr = ipa; - - /* Combine the S1 and S2 perms. */ - result->f.prot &= s1_prot; - - /* If S2 fails, return early. */ - if (ret) { - return ret; - } - - /* Combine the S1 and S2 cache attributes. */ - hcr = arm_hcr_el2_eff_secstate(env, is_secure); - if (hcr & HCR_DC) { - /* - * HCR.DC forces the first stage attributes to - * Normal Non-Shareable, - * Inner Write-Back Read-Allocate Write-Allocate, - * Outer Write-Back Read-Allocate Write-Allocate. - * Do not overwrite Tagged within attrs. - */ - if (cacheattrs1.attrs != 0xf0) { - cacheattrs1.attrs = 0xff; - } - cacheattrs1.shareability = 0; - } - result->cacheattrs = combine_cacheattrs(hcr, cacheattrs1, - result->cacheattrs); - - /* - * Check if IPA translates to secure or non-secure PA space. - * Note that VSTCR overrides VTCR and {N}SW overrides {N}SA. - */ - result->f.attrs.secure = - (is_secure - && !(env->cp15.vstcr_el2 & (VSTCR_SA | VSTCR_SW)) - && (ipa_secure - || !(env->cp15.vtcr_el2 & (VTCR_NSA | VTCR_NSW)))); - - return 0; - } else { - /* - * For non-EL2 CPUs a stage1+stage2 translation is just stage 1. - */ - mmu_idx = stage_1_mmu_idx(mmu_idx); + return get_phys_addr_twostage(env, ptw, address, access_type, + result, fi); } } -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> The return type of the functions is already bool, but in a few instances we used an integer type with the return statement. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20221011031911.2408754-13-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/ptw.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) diff --git a/target/arm/ptw.c b/target/arm/ptw.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/ptw.c +++ b/target/arm/ptw.c @@ -XXX,XX +XXX,XX @@ static bool get_phys_addr_disabled(CPUARMState *env, target_ulong address, result->f.lg_page_size = TARGET_PAGE_BITS; result->cacheattrs.shareability = shareability; result->cacheattrs.attrs = memattr; - return 0; + return false; } static bool get_phys_addr_twostage(CPUARMState *env, S1Translate *ptw, @@ -XXX,XX +XXX,XX @@ static bool get_phys_addr_twostage(CPUARMState *env, S1Translate *ptw, { hwaddr ipa; int s1_prot; - int ret; bool is_secure = ptw->in_secure; - bool ipa_secure, s2walk_secure; + bool ret, ipa_secure, s2walk_secure; ARMCacheAttrs cacheattrs1; bool is_el0; uint64_t hcr; @@ -XXX,XX +XXX,XX @@ static bool get_phys_addr_twostage(CPUARMState *env, S1Translate *ptw, && (ipa_secure || !(env->cp15.vtcr_el2 & (VTCR_NSA | VTCR_NSW)))); - return 0; + return false; } static bool get_phys_addr_with_struct(CPUARMState *env, S1Translate *ptw, -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> A simple helper to retrieve the length of the current insn. Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20221020030641.2066807-2-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate.h | 5 +++++ target/arm/translate-vfp.c | 2 +- target/arm/translate.c | 5 ++--- 3 files changed, 8 insertions(+), 4 deletions(-) diff --git a/target/arm/translate.h b/target/arm/translate.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.h +++ b/target/arm/translate.h @@ -XXX,XX +XXX,XX @@ static inline void disas_set_insn_syndrome(DisasContext *s, uint32_t syn) s->insn_start = NULL; } +static inline int curr_insn_len(DisasContext *s) +{ + return s->base.pc_next - s->pc_curr; +} + /* is_jmp field values */ #define DISAS_JUMP DISAS_TARGET_0 /* only pc was modified dynamically */ /* CPU state was modified dynamically; exit to main loop for interrupts. */ diff --git a/target/arm/translate-vfp.c b/target/arm/translate-vfp.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-vfp.c +++ b/target/arm/translate-vfp.c @@ -XXX,XX +XXX,XX @@ static bool vfp_access_check_a(DisasContext *s, bool ignore_vfp_enabled) if (s->sme_trap_nonstreaming) { gen_exception_insn(s, s->pc_curr, EXCP_UDEF, syn_smetrap(SME_ET_Streaming, - s->base.pc_next - s->pc_curr == 2)); + curr_insn_len(s) == 2)); return false; } diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static ISSInfo make_issinfo(DisasContext *s, int rd, bool p, bool w) /* ISS not valid if writeback */ if (p && !w) { ret = rd; - if (s->base.pc_next - s->pc_curr == 2) { + if (curr_insn_len(s) == 2) { ret |= ISSIs16Bit; } } else { @@ -XXX,XX +XXX,XX @@ static void arm_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu) /* nothing more to generate */ break; case DISAS_WFI: - gen_helper_wfi(cpu_env, - tcg_constant_i32(dc->base.pc_next - dc->pc_curr)); + gen_helper_wfi(cpu_env, tcg_constant_i32(curr_insn_len(dc))); /* * The helper doesn't necessarily throw an exception, but we * must go back to the main loop to check for interrupts anyway. -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> In preparation for TARGET_TB_PCREL, reduce reliance on absolute values. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20221020030641.2066807-3-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-a64.c | 40 ++++++++++++++++++++------------------ target/arm/translate.c | 10 ++++++---- 2 files changed, 27 insertions(+), 23 deletions(-) diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a64.c +++ b/target/arm/translate-a64.c @@ -XXX,XX +XXX,XX @@ static inline bool use_goto_tb(DisasContext *s, uint64_t dest) return translator_use_goto_tb(&s->base, dest); } -static inline void gen_goto_tb(DisasContext *s, int n, uint64_t dest) +static void gen_goto_tb(DisasContext *s, int n, int64_t diff) { + uint64_t dest = s->pc_curr + diff; + if (use_goto_tb(s, dest)) { tcg_gen_goto_tb(n); gen_a64_set_pc_im(dest); @@ -XXX,XX +XXX,XX @@ static inline AArch64DecodeFn *lookup_disas_fn(const AArch64DecodeTable *table, */ static void disas_uncond_b_imm(DisasContext *s, uint32_t insn) { - uint64_t addr = s->pc_curr + sextract32(insn, 0, 26) * 4; + int64_t diff = sextract32(insn, 0, 26) * 4; if (insn & (1U << 31)) { /* BL Branch with link */ @@ -XXX,XX +XXX,XX @@ static void disas_uncond_b_imm(DisasContext *s, uint32_t insn) /* B Branch / BL Branch with link */ reset_btype(s); - gen_goto_tb(s, 0, addr); + gen_goto_tb(s, 0, diff); } /* Compare and branch (immediate) @@ -XXX,XX +XXX,XX @@ static void disas_uncond_b_imm(DisasContext *s, uint32_t insn) static void disas_comp_b_imm(DisasContext *s, uint32_t insn) { unsigned int sf, op, rt; - uint64_t addr; + int64_t diff; TCGLabel *label_match; TCGv_i64 tcg_cmp; sf = extract32(insn, 31, 1); op = extract32(insn, 24, 1); /* 0: CBZ; 1: CBNZ */ rt = extract32(insn, 0, 5); - addr = s->pc_curr + sextract32(insn, 5, 19) * 4; + diff = sextract32(insn, 5, 19) * 4; tcg_cmp = read_cpu_reg(s, rt, sf); label_match = gen_new_label(); @@ -XXX,XX +XXX,XX @@ static void disas_comp_b_imm(DisasContext *s, uint32_t insn) tcg_gen_brcondi_i64(op ? TCG_COND_NE : TCG_COND_EQ, tcg_cmp, 0, label_match); - gen_goto_tb(s, 0, s->base.pc_next); + gen_goto_tb(s, 0, 4); gen_set_label(label_match); - gen_goto_tb(s, 1, addr); + gen_goto_tb(s, 1, diff); } /* Test and branch (immediate) @@ -XXX,XX +XXX,XX @@ static void disas_comp_b_imm(DisasContext *s, uint32_t insn) static void disas_test_b_imm(DisasContext *s, uint32_t insn) { unsigned int bit_pos, op, rt; - uint64_t addr; + int64_t diff; TCGLabel *label_match; TCGv_i64 tcg_cmp; bit_pos = (extract32(insn, 31, 1) << 5) | extract32(insn, 19, 5); op = extract32(insn, 24, 1); /* 0: TBZ; 1: TBNZ */ - addr = s->pc_curr + sextract32(insn, 5, 14) * 4; + diff = sextract32(insn, 5, 14) * 4; rt = extract32(insn, 0, 5); tcg_cmp = tcg_temp_new_i64(); @@ -XXX,XX +XXX,XX @@ static void disas_test_b_imm(DisasContext *s, uint32_t insn) tcg_gen_brcondi_i64(op ? TCG_COND_NE : TCG_COND_EQ, tcg_cmp, 0, label_match); tcg_temp_free_i64(tcg_cmp); - gen_goto_tb(s, 0, s->base.pc_next); + gen_goto_tb(s, 0, 4); gen_set_label(label_match); - gen_goto_tb(s, 1, addr); + gen_goto_tb(s, 1, diff); } /* Conditional branch (immediate) @@ -XXX,XX +XXX,XX @@ static void disas_test_b_imm(DisasContext *s, uint32_t insn) static void disas_cond_b_imm(DisasContext *s, uint32_t insn) { unsigned int cond; - uint64_t addr; + int64_t diff; if ((insn & (1 << 4)) || (insn & (1 << 24))) { unallocated_encoding(s); return; } - addr = s->pc_curr + sextract32(insn, 5, 19) * 4; + diff = sextract32(insn, 5, 19) * 4; cond = extract32(insn, 0, 4); reset_btype(s); @@ -XXX,XX +XXX,XX @@ static void disas_cond_b_imm(DisasContext *s, uint32_t insn) /* genuinely conditional branches */ TCGLabel *label_match = gen_new_label(); arm_gen_test_cc(cond, label_match); - gen_goto_tb(s, 0, s->base.pc_next); + gen_goto_tb(s, 0, 4); gen_set_label(label_match); - gen_goto_tb(s, 1, addr); + gen_goto_tb(s, 1, diff); } else { /* 0xe and 0xf are both "always" conditions */ - gen_goto_tb(s, 0, addr); + gen_goto_tb(s, 0, diff); } } @@ -XXX,XX +XXX,XX @@ static void handle_sync(DisasContext *s, uint32_t insn, * any pending interrupts immediately. */ reset_btype(s); - gen_goto_tb(s, 0, s->base.pc_next); + gen_goto_tb(s, 0, 4); return; case 7: /* SB */ @@ -XXX,XX +XXX,XX @@ static void handle_sync(DisasContext *s, uint32_t insn, * MB and end the TB instead. */ tcg_gen_mb(TCG_MO_ALL | TCG_BAR_SC); - gen_goto_tb(s, 0, s->base.pc_next); + gen_goto_tb(s, 0, 4); return; default: @@ -XXX,XX +XXX,XX @@ static void aarch64_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu) switch (dc->base.is_jmp) { case DISAS_NEXT: case DISAS_TOO_MANY: - gen_goto_tb(dc, 1, dc->base.pc_next); + gen_goto_tb(dc, 1, 4); break; default: case DISAS_UPDATE_EXIT: diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static void gen_goto_ptr(void) * cpu_loop_exec. Any live exit_requests will be processed as we * enter the next TB. */ -static void gen_goto_tb(DisasContext *s, int n, target_ulong dest) +static void gen_goto_tb(DisasContext *s, int n, int diff) { + target_ulong dest = s->pc_curr + diff; + if (translator_use_goto_tb(&s->base, dest)) { tcg_gen_goto_tb(n); gen_set_pc_im(s, dest); @@ -XXX,XX +XXX,XX @@ static inline void gen_jmp_tb(DisasContext *s, uint32_t dest, int tbno) * gen_jmp(); * on the second call to gen_jmp(). */ - gen_goto_tb(s, tbno, dest); + gen_goto_tb(s, tbno, dest - s->pc_curr); break; case DISAS_UPDATE_NOCHAIN: case DISAS_UPDATE_EXIT: @@ -XXX,XX +XXX,XX @@ static void arm_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu) switch (dc->base.is_jmp) { case DISAS_NEXT: case DISAS_TOO_MANY: - gen_goto_tb(dc, 1, dc->base.pc_next); + gen_goto_tb(dc, 1, curr_insn_len(dc)); break; case DISAS_UPDATE_NOCHAIN: gen_set_pc_im(dc, dc->base.pc_next); @@ -XXX,XX +XXX,XX @@ static void arm_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu) gen_set_pc_im(dc, dc->base.pc_next); gen_singlestep_exception(dc); } else { - gen_goto_tb(dc, 1, dc->base.pc_next); + gen_goto_tb(dc, 1, curr_insn_len(dc)); } } } -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> In preparation for TARGET_TB_PCREL, reduce reliance on absolute values by passing in pc difference. Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20221020030641.2066807-4-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-a32.h | 2 +- target/arm/translate.h | 6 ++-- target/arm/translate-a64.c | 32 +++++++++--------- target/arm/translate-vfp.c | 2 +- target/arm/translate.c | 68 ++++++++++++++++++++------------------ 5 files changed, 56 insertions(+), 54 deletions(-) diff --git a/target/arm/translate-a32.h b/target/arm/translate-a32.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a32.h +++ b/target/arm/translate-a32.h @@ -XXX,XX +XXX,XX @@ void write_neon_element64(TCGv_i64 src, int reg, int ele, MemOp memop); TCGv_i32 add_reg_for_lit(DisasContext *s, int reg, int ofs); void gen_set_cpsr(TCGv_i32 var, uint32_t mask); void gen_set_condexec(DisasContext *s); -void gen_set_pc_im(DisasContext *s, target_ulong val); +void gen_update_pc(DisasContext *s, target_long diff); void gen_lookup_tb(DisasContext *s); long vfp_reg_offset(bool dp, unsigned reg); long neon_full_reg_offset(unsigned reg); diff --git a/target/arm/translate.h b/target/arm/translate.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.h +++ b/target/arm/translate.h @@ -XXX,XX +XXX,XX @@ static inline int curr_insn_len(DisasContext *s) * For instructions which want an immediate exit to the main loop, as opposed * to attempting to use lookup_and_goto_ptr. Unlike DISAS_UPDATE_EXIT, this * doesn't write the PC on exiting the translation loop so you need to ensure - * something (gen_a64_set_pc_im or runtime helper) has done so before we reach + * something (gen_a64_update_pc or runtime helper) has done so before we reach * return from cpu_tb_exec. */ #define DISAS_EXIT DISAS_TARGET_9 @@ -XXX,XX +XXX,XX @@ static inline int curr_insn_len(DisasContext *s) #ifdef TARGET_AARCH64 void a64_translate_init(void); -void gen_a64_set_pc_im(uint64_t val); +void gen_a64_update_pc(DisasContext *s, target_long diff); extern const TranslatorOps aarch64_translator_ops; #else static inline void a64_translate_init(void) { } -static inline void gen_a64_set_pc_im(uint64_t val) +static inline void gen_a64_update_pc(DisasContext *s, target_long diff) { } #endif diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a64.c +++ b/target/arm/translate-a64.c @@ -XXX,XX +XXX,XX @@ static void reset_btype(DisasContext *s) } } -void gen_a64_set_pc_im(uint64_t val) +void gen_a64_update_pc(DisasContext *s, target_long diff) { - tcg_gen_movi_i64(cpu_pc, val); + tcg_gen_movi_i64(cpu_pc, s->pc_curr + diff); } /* @@ -XXX,XX +XXX,XX @@ static void gen_exception_internal(int excp) static void gen_exception_internal_insn(DisasContext *s, uint64_t pc, int excp) { - gen_a64_set_pc_im(pc); + gen_a64_update_pc(s, pc - s->pc_curr); gen_exception_internal(excp); s->base.is_jmp = DISAS_NORETURN; } static void gen_exception_bkpt_insn(DisasContext *s, uint32_t syndrome) { - gen_a64_set_pc_im(s->pc_curr); + gen_a64_update_pc(s, 0); gen_helper_exception_bkpt_insn(cpu_env, tcg_constant_i32(syndrome)); s->base.is_jmp = DISAS_NORETURN; } @@ -XXX,XX +XXX,XX @@ static void gen_goto_tb(DisasContext *s, int n, int64_t diff) if (use_goto_tb(s, dest)) { tcg_gen_goto_tb(n); - gen_a64_set_pc_im(dest); + gen_a64_update_pc(s, diff); tcg_gen_exit_tb(s->base.tb, n); s->base.is_jmp = DISAS_NORETURN; } else { - gen_a64_set_pc_im(dest); + gen_a64_update_pc(s, diff); if (s->ss_active) { gen_step_complete_exception(s); } else { @@ -XXX,XX +XXX,XX @@ static void handle_sys(DisasContext *s, uint32_t insn, bool isread, uint32_t syndrome; syndrome = syn_aa64_sysregtrap(op0, op1, op2, crn, crm, rt, isread); - gen_a64_set_pc_im(s->pc_curr); + gen_a64_update_pc(s, 0); gen_helper_access_check_cp_reg(cpu_env, tcg_constant_ptr(ri), tcg_constant_i32(syndrome), @@ -XXX,XX +XXX,XX @@ static void handle_sys(DisasContext *s, uint32_t insn, bool isread, * The readfn or writefn might raise an exception; * synchronize the CPU state in case it does. */ - gen_a64_set_pc_im(s->pc_curr); + gen_a64_update_pc(s, 0); } /* Handle special cases first */ @@ -XXX,XX +XXX,XX @@ static void disas_exc(DisasContext *s, uint32_t insn) /* The pre HVC helper handles cases when HVC gets trapped * as an undefined insn by runtime configuration. */ - gen_a64_set_pc_im(s->pc_curr); + gen_a64_update_pc(s, 0); gen_helper_pre_hvc(cpu_env); gen_ss_advance(s); gen_exception_insn_el(s, s->base.pc_next, EXCP_HVC, @@ -XXX,XX +XXX,XX @@ static void disas_exc(DisasContext *s, uint32_t insn) unallocated_encoding(s); break; } - gen_a64_set_pc_im(s->pc_curr); + gen_a64_update_pc(s, 0); gen_helper_pre_smc(cpu_env, tcg_constant_i32(syn_aa64_smc(imm16))); gen_ss_advance(s); gen_exception_insn_el(s, s->base.pc_next, EXCP_SMC, @@ -XXX,XX +XXX,XX @@ static void aarch64_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu) */ switch (dc->base.is_jmp) { default: - gen_a64_set_pc_im(dc->base.pc_next); + gen_a64_update_pc(dc, 4); /* fall through */ case DISAS_EXIT: case DISAS_JUMP: @@ -XXX,XX +XXX,XX @@ static void aarch64_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu) break; default: case DISAS_UPDATE_EXIT: - gen_a64_set_pc_im(dc->base.pc_next); + gen_a64_update_pc(dc, 4); /* fall through */ case DISAS_EXIT: tcg_gen_exit_tb(NULL, 0); break; case DISAS_UPDATE_NOCHAIN: - gen_a64_set_pc_im(dc->base.pc_next); + gen_a64_update_pc(dc, 4); /* fall through */ case DISAS_JUMP: tcg_gen_lookup_and_goto_ptr(); @@ -XXX,XX +XXX,XX @@ static void aarch64_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu) case DISAS_SWI: break; case DISAS_WFE: - gen_a64_set_pc_im(dc->base.pc_next); + gen_a64_update_pc(dc, 4); gen_helper_wfe(cpu_env); break; case DISAS_YIELD: - gen_a64_set_pc_im(dc->base.pc_next); + gen_a64_update_pc(dc, 4); gen_helper_yield(cpu_env); break; case DISAS_WFI: @@ -XXX,XX +XXX,XX @@ static void aarch64_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu) * This is a special case because we don't want to just halt * the CPU if trying to debug across a WFI. */ - gen_a64_set_pc_im(dc->base.pc_next); + gen_a64_update_pc(dc, 4); gen_helper_wfi(cpu_env, tcg_constant_i32(4)); /* * The helper doesn't necessarily throw an exception, but we diff --git a/target/arm/translate-vfp.c b/target/arm/translate-vfp.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-vfp.c +++ b/target/arm/translate-vfp.c @@ -XXX,XX +XXX,XX @@ static bool trans_VMSR_VMRS(DisasContext *s, arg_VMSR_VMRS *a) case ARM_VFP_FPSID: if (s->current_el == 1) { gen_set_condexec(s); - gen_set_pc_im(s, s->pc_curr); + gen_update_pc(s, 0); gen_helper_check_hcr_el2_trap(cpu_env, tcg_constant_i32(a->rt), tcg_constant_i32(a->reg)); diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ void gen_set_condexec(DisasContext *s) } } -void gen_set_pc_im(DisasContext *s, target_ulong val) +void gen_update_pc(DisasContext *s, target_long diff) { - tcg_gen_movi_i32(cpu_R[15], val); + tcg_gen_movi_i32(cpu_R[15], s->pc_curr + diff); } /* Set PC and Thumb state from var. var is marked as dead. */ @@ -XXX,XX +XXX,XX @@ static inline void gen_bxns(DisasContext *s, int rm) /* The bxns helper may raise an EXCEPTION_EXIT exception, so in theory * we need to sync state before calling it, but: - * - we don't need to do gen_set_pc_im() because the bxns helper will + * - we don't need to do gen_update_pc() because the bxns helper will * always set the PC itself * - we don't need to do gen_set_condexec() because BXNS is UNPREDICTABLE * unless it's outside an IT block or the last insn in an IT block, @@ -XXX,XX +XXX,XX @@ static inline void gen_blxns(DisasContext *s, int rm) * We do however need to set the PC, because the blxns helper reads it. * The blxns helper may throw an exception. */ - gen_set_pc_im(s, s->base.pc_next); + gen_update_pc(s, curr_insn_len(s)); gen_helper_v7m_blxns(cpu_env, var); tcg_temp_free_i32(var); s->base.is_jmp = DISAS_EXIT; @@ -XXX,XX +XXX,XX @@ static inline void gen_hvc(DisasContext *s, int imm16) * as an undefined insn by runtime configuration (ie before * the insn really executes). */ - gen_set_pc_im(s, s->pc_curr); + gen_update_pc(s, 0); gen_helper_pre_hvc(cpu_env); /* Otherwise we will treat this as a real exception which * happens after execution of the insn. (The distinction matters @@ -XXX,XX +XXX,XX @@ static inline void gen_hvc(DisasContext *s, int imm16) * for single stepping.) */ s->svc_imm = imm16; - gen_set_pc_im(s, s->base.pc_next); + gen_update_pc(s, curr_insn_len(s)); s->base.is_jmp = DISAS_HVC; } @@ -XXX,XX +XXX,XX @@ static inline void gen_smc(DisasContext *s) /* As with HVC, we may take an exception either before or after * the insn executes. */ - gen_set_pc_im(s, s->pc_curr); + gen_update_pc(s, 0); gen_helper_pre_smc(cpu_env, tcg_constant_i32(syn_aa32_smc())); - gen_set_pc_im(s, s->base.pc_next); + gen_update_pc(s, curr_insn_len(s)); s->base.is_jmp = DISAS_SMC; } static void gen_exception_internal_insn(DisasContext *s, uint32_t pc, int excp) { gen_set_condexec(s); - gen_set_pc_im(s, pc); + gen_update_pc(s, pc - s->pc_curr); gen_exception_internal(excp); s->base.is_jmp = DISAS_NORETURN; } @@ -XXX,XX +XXX,XX @@ static void gen_exception_insn_el_v(DisasContext *s, uint64_t pc, int excp, uint32_t syn, TCGv_i32 tcg_el) { if (s->aarch64) { - gen_a64_set_pc_im(pc); + gen_a64_update_pc(s, pc - s->pc_curr); } else { gen_set_condexec(s); - gen_set_pc_im(s, pc); + gen_update_pc(s, pc - s->pc_curr); } gen_exception_el_v(excp, syn, tcg_el); s->base.is_jmp = DISAS_NORETURN; @@ -XXX,XX +XXX,XX @@ void gen_exception_insn_el(DisasContext *s, uint64_t pc, int excp, void gen_exception_insn(DisasContext *s, uint64_t pc, int excp, uint32_t syn) { if (s->aarch64) { - gen_a64_set_pc_im(pc); + gen_a64_update_pc(s, pc - s->pc_curr); } else { gen_set_condexec(s); - gen_set_pc_im(s, pc); + gen_update_pc(s, pc - s->pc_curr); } gen_exception(excp, syn); s->base.is_jmp = DISAS_NORETURN; @@ -XXX,XX +XXX,XX @@ void gen_exception_insn(DisasContext *s, uint64_t pc, int excp, uint32_t syn) static void gen_exception_bkpt_insn(DisasContext *s, uint32_t syn) { gen_set_condexec(s); - gen_set_pc_im(s, s->pc_curr); + gen_update_pc(s, 0); gen_helper_exception_bkpt_insn(cpu_env, tcg_constant_i32(syn)); s->base.is_jmp = DISAS_NORETURN; } @@ -XXX,XX +XXX,XX @@ static void gen_goto_tb(DisasContext *s, int n, int diff) if (translator_use_goto_tb(&s->base, dest)) { tcg_gen_goto_tb(n); - gen_set_pc_im(s, dest); + gen_update_pc(s, diff); tcg_gen_exit_tb(s->base.tb, n); } else { - gen_set_pc_im(s, dest); + gen_update_pc(s, diff); gen_goto_ptr(); } s->base.is_jmp = DISAS_NORETURN; @@ -XXX,XX +XXX,XX @@ static void gen_goto_tb(DisasContext *s, int n, int diff) /* Jump, specifying which TB number to use if we gen_goto_tb() */ static inline void gen_jmp_tb(DisasContext *s, uint32_t dest, int tbno) { + int diff = dest - s->pc_curr; + if (unlikely(s->ss_active)) { /* An indirect jump so that we still trigger the debug exception. */ - gen_set_pc_im(s, dest); + gen_update_pc(s, diff); s->base.is_jmp = DISAS_JUMP; return; } @@ -XXX,XX +XXX,XX @@ static inline void gen_jmp_tb(DisasContext *s, uint32_t dest, int tbno) * gen_jmp(); * on the second call to gen_jmp(). */ - gen_goto_tb(s, tbno, dest - s->pc_curr); + gen_goto_tb(s, tbno, diff); break; case DISAS_UPDATE_NOCHAIN: case DISAS_UPDATE_EXIT: @@ -XXX,XX +XXX,XX @@ static inline void gen_jmp_tb(DisasContext *s, uint32_t dest, int tbno) * Avoid using goto_tb so we really do exit back to the main loop * and don't chain to another TB. */ - gen_set_pc_im(s, dest); + gen_update_pc(s, diff); gen_goto_ptr(); s->base.is_jmp = DISAS_NORETURN; break; @@ -XXX,XX +XXX,XX @@ static void gen_msr_banked(DisasContext *s, int r, int sysm, int rn) /* Sync state because msr_banked() can raise exceptions */ gen_set_condexec(s); - gen_set_pc_im(s, s->pc_curr); + gen_update_pc(s, 0); tcg_reg = load_reg(s, rn); gen_helper_msr_banked(cpu_env, tcg_reg, tcg_constant_i32(tgtmode), @@ -XXX,XX +XXX,XX @@ static void gen_mrs_banked(DisasContext *s, int r, int sysm, int rn) /* Sync state because mrs_banked() can raise exceptions */ gen_set_condexec(s); - gen_set_pc_im(s, s->pc_curr); + gen_update_pc(s, 0); tcg_reg = tcg_temp_new_i32(); gen_helper_mrs_banked(tcg_reg, cpu_env, tcg_constant_i32(tgtmode), @@ -XXX,XX +XXX,XX @@ static void do_coproc_insn(DisasContext *s, int cpnum, int is64, } gen_set_condexec(s); - gen_set_pc_im(s, s->pc_curr); + gen_update_pc(s, 0); gen_helper_access_check_cp_reg(cpu_env, tcg_constant_ptr(ri), tcg_constant_i32(syndrome), @@ -XXX,XX +XXX,XX @@ static void do_coproc_insn(DisasContext *s, int cpnum, int is64, * synchronize the CPU state in case it does. */ gen_set_condexec(s); - gen_set_pc_im(s, s->pc_curr); + gen_update_pc(s, 0); } /* Handle special cases first */ @@ -XXX,XX +XXX,XX @@ static void do_coproc_insn(DisasContext *s, int cpnum, int is64, unallocated_encoding(s); return; } - gen_set_pc_im(s, s->base.pc_next); + gen_update_pc(s, curr_insn_len(s)); s->base.is_jmp = DISAS_WFI; return; default: @@ -XXX,XX +XXX,XX @@ static void gen_srs(DisasContext *s, addr = tcg_temp_new_i32(); /* get_r13_banked() will raise an exception if called from System mode */ gen_set_condexec(s); - gen_set_pc_im(s, s->pc_curr); + gen_update_pc(s, 0); gen_helper_get_r13_banked(addr, cpu_env, tcg_constant_i32(mode)); switch (amode) { case 0: /* DA */ @@ -XXX,XX +XXX,XX @@ static bool trans_YIELD(DisasContext *s, arg_YIELD *a) * scheduling of other vCPUs. */ if (!(tb_cflags(s->base.tb) & CF_PARALLEL)) { - gen_set_pc_im(s, s->base.pc_next); + gen_update_pc(s, curr_insn_len(s)); s->base.is_jmp = DISAS_YIELD; } return true; @@ -XXX,XX +XXX,XX @@ static bool trans_WFE(DisasContext *s, arg_WFE *a) * implemented so we can't sleep like WFI does. */ if (!(tb_cflags(s->base.tb) & CF_PARALLEL)) { - gen_set_pc_im(s, s->base.pc_next); + gen_update_pc(s, curr_insn_len(s)); s->base.is_jmp = DISAS_WFE; } return true; @@ -XXX,XX +XXX,XX @@ static bool trans_WFE(DisasContext *s, arg_WFE *a) static bool trans_WFI(DisasContext *s, arg_WFI *a) { /* For WFI, halt the vCPU until an IRQ. */ - gen_set_pc_im(s, s->base.pc_next); + gen_update_pc(s, curr_insn_len(s)); s->base.is_jmp = DISAS_WFI; return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_SVC(DisasContext *s, arg_SVC *a) (a->imm == semihost_imm)) { gen_exception_internal_insn(s, s->pc_curr, EXCP_SEMIHOST); } else { - gen_set_pc_im(s, s->base.pc_next); + gen_update_pc(s, curr_insn_len(s)); s->svc_imm = a->imm; s->base.is_jmp = DISAS_SWI; } @@ -XXX,XX +XXX,XX @@ static void arm_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu) case DISAS_TOO_MANY: case DISAS_UPDATE_EXIT: case DISAS_UPDATE_NOCHAIN: - gen_set_pc_im(dc, dc->base.pc_next); + gen_update_pc(dc, curr_insn_len(dc)); /* fall through */ default: /* FIXME: Single stepping a WFI insn will not halt the CPU. */ @@ -XXX,XX +XXX,XX @@ static void arm_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu) gen_goto_tb(dc, 1, curr_insn_len(dc)); break; case DISAS_UPDATE_NOCHAIN: - gen_set_pc_im(dc, dc->base.pc_next); + gen_update_pc(dc, curr_insn_len(dc)); /* fall through */ case DISAS_JUMP: gen_goto_ptr(); break; case DISAS_UPDATE_EXIT: - gen_set_pc_im(dc, dc->base.pc_next); + gen_update_pc(dc, curr_insn_len(dc)); /* fall through */ default: /* indicate that the hash table must be used to find the next TB */ @@ -XXX,XX +XXX,XX @@ static void arm_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu) gen_set_label(dc->condlabel); gen_set_condexec(dc); if (unlikely(dc->ss_active)) { - gen_set_pc_im(dc, dc->base.pc_next); + gen_update_pc(dc, curr_insn_len(dc)); gen_singlestep_exception(dc); } else { gen_goto_tb(dc, 1, curr_insn_len(dc)); -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> In preparation for TARGET_TB_PCREL, reduce reliance on absolute values. Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20221020030641.2066807-5-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate.h | 5 +++-- target/arm/translate-a64.c | 28 ++++++++++------------- target/arm/translate-m-nocp.c | 6 ++--- target/arm/translate-mve.c | 2 +- target/arm/translate-vfp.c | 6 ++--- target/arm/translate.c | 42 +++++++++++++++++------------------ 6 files changed, 43 insertions(+), 46 deletions(-) diff --git a/target/arm/translate.h b/target/arm/translate.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.h +++ b/target/arm/translate.h @@ -XXX,XX +XXX,XX @@ void arm_jump_cc(DisasCompare *cmp, TCGLabel *label); void arm_gen_test_cc(int cc, TCGLabel *label); MemOp pow2_align(unsigned i); void unallocated_encoding(DisasContext *s); -void gen_exception_insn_el(DisasContext *s, uint64_t pc, int excp, +void gen_exception_insn_el(DisasContext *s, target_long pc_diff, int excp, uint32_t syn, uint32_t target_el); -void gen_exception_insn(DisasContext *s, uint64_t pc, int excp, uint32_t syn); +void gen_exception_insn(DisasContext *s, target_long pc_diff, + int excp, uint32_t syn); /* Return state of Alternate Half-precision flag, caller frees result */ static inline TCGv_i32 get_ahp_flag(void) diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a64.c +++ b/target/arm/translate-a64.c @@ -XXX,XX +XXX,XX @@ static bool fp_access_check_only(DisasContext *s) assert(!s->fp_access_checked); s->fp_access_checked = true; - gen_exception_insn_el(s, s->pc_curr, EXCP_UDEF, + gen_exception_insn_el(s, 0, EXCP_UDEF, syn_fp_access_trap(1, 0xe, false, 0), s->fp_excp_el); return false; @@ -XXX,XX +XXX,XX @@ static bool fp_access_check(DisasContext *s) return false; } if (s->sme_trap_nonstreaming && s->is_nonstreaming) { - gen_exception_insn(s, s->pc_curr, EXCP_UDEF, + gen_exception_insn(s, 0, EXCP_UDEF, syn_smetrap(SME_ET_Streaming, false)); return false; } @@ -XXX,XX +XXX,XX @@ bool sve_access_check(DisasContext *s) goto fail_exit; } } else if (s->sve_excp_el) { - gen_exception_insn_el(s, s->pc_curr, EXCP_UDEF, + gen_exception_insn_el(s, 0, EXCP_UDEF, syn_sve_access_trap(), s->sve_excp_el); goto fail_exit; } @@ -XXX,XX +XXX,XX @@ bool sve_access_check(DisasContext *s) static bool sme_access_check(DisasContext *s) { if (s->sme_excp_el) { - gen_exception_insn_el(s, s->pc_curr, EXCP_UDEF, + gen_exception_insn_el(s, 0, EXCP_UDEF, syn_smetrap(SME_ET_AccessTrap, false), s->sme_excp_el); return false; @@ -XXX,XX +XXX,XX @@ bool sme_enabled_check_with_svcr(DisasContext *s, unsigned req) return false; } if (FIELD_EX64(req, SVCR, SM) && !s->pstate_sm) { - gen_exception_insn(s, s->pc_curr, EXCP_UDEF, + gen_exception_insn(s, 0, EXCP_UDEF, syn_smetrap(SME_ET_NotStreaming, false)); return false; } if (FIELD_EX64(req, SVCR, ZA) && !s->pstate_za) { - gen_exception_insn(s, s->pc_curr, EXCP_UDEF, + gen_exception_insn(s, 0, EXCP_UDEF, syn_smetrap(SME_ET_InactiveZA, false)); return false; } @@ -XXX,XX +XXX,XX @@ static void gen_sysreg_undef(DisasContext *s, bool isread, } else { syndrome = syn_uncategorized(); } - gen_exception_insn(s, s->pc_curr, EXCP_UDEF, syndrome); + gen_exception_insn(s, 0, EXCP_UDEF, syndrome); } /* MRS - move from system register @@ -XXX,XX +XXX,XX @@ static void disas_exc(DisasContext *s, uint32_t insn) switch (op2_ll) { case 1: /* SVC */ gen_ss_advance(s); - gen_exception_insn(s, s->base.pc_next, EXCP_SWI, - syn_aa64_svc(imm16)); + gen_exception_insn(s, 4, EXCP_SWI, syn_aa64_svc(imm16)); break; case 2: /* HVC */ if (s->current_el == 0) { @@ -XXX,XX +XXX,XX @@ static void disas_exc(DisasContext *s, uint32_t insn) gen_a64_update_pc(s, 0); gen_helper_pre_hvc(cpu_env); gen_ss_advance(s); - gen_exception_insn_el(s, s->base.pc_next, EXCP_HVC, - syn_aa64_hvc(imm16), 2); + gen_exception_insn_el(s, 4, EXCP_HVC, syn_aa64_hvc(imm16), 2); break; case 3: /* SMC */ if (s->current_el == 0) { @@ -XXX,XX +XXX,XX @@ static void disas_exc(DisasContext *s, uint32_t insn) gen_a64_update_pc(s, 0); gen_helper_pre_smc(cpu_env, tcg_constant_i32(syn_aa64_smc(imm16))); gen_ss_advance(s); - gen_exception_insn_el(s, s->base.pc_next, EXCP_SMC, - syn_aa64_smc(imm16), 3); + gen_exception_insn_el(s, 4, EXCP_SMC, syn_aa64_smc(imm16), 3); break; default: unallocated_encoding(s); @@ -XXX,XX +XXX,XX @@ static void aarch64_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu) * Illegal execution state. This has priority over BTI * exceptions, but comes after instruction abort exceptions. */ - gen_exception_insn(s, s->pc_curr, EXCP_UDEF, syn_illegalstate()); + gen_exception_insn(s, 0, EXCP_UDEF, syn_illegalstate()); return; } @@ -XXX,XX +XXX,XX @@ static void aarch64_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu) if (s->btype != 0 && s->guarded_page && !btype_destination_ok(insn, s->bt, s->btype)) { - gen_exception_insn(s, s->pc_curr, EXCP_UDEF, - syn_btitrap(s->btype)); + gen_exception_insn(s, 0, EXCP_UDEF, syn_btitrap(s->btype)); return; } } else { diff --git a/target/arm/translate-m-nocp.c b/target/arm/translate-m-nocp.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-m-nocp.c +++ b/target/arm/translate-m-nocp.c @@ -XXX,XX +XXX,XX @@ static bool trans_VSCCLRM(DisasContext *s, arg_VSCCLRM *a) tcg_gen_brcondi_i32(TCG_COND_EQ, sfpa, 0, s->condlabel); if (s->fp_excp_el != 0) { - gen_exception_insn_el(s, s->pc_curr, EXCP_NOCP, + gen_exception_insn_el(s, 0, EXCP_NOCP, syn_uncategorized(), s->fp_excp_el); return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_NOCP(DisasContext *s, arg_nocp *a) } if (a->cp != 10) { - gen_exception_insn(s, s->pc_curr, EXCP_NOCP, syn_uncategorized()); + gen_exception_insn(s, 0, EXCP_NOCP, syn_uncategorized()); return true; } if (s->fp_excp_el != 0) { - gen_exception_insn_el(s, s->pc_curr, EXCP_NOCP, + gen_exception_insn_el(s, 0, EXCP_NOCP, syn_uncategorized(), s->fp_excp_el); return true; } diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ bool mve_eci_check(DisasContext *s) return true; default: /* Reserved value: INVSTATE UsageFault */ - gen_exception_insn(s, s->pc_curr, EXCP_INVSTATE, syn_uncategorized()); + gen_exception_insn(s, 0, EXCP_INVSTATE, syn_uncategorized()); return false; } } diff --git a/target/arm/translate-vfp.c b/target/arm/translate-vfp.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-vfp.c +++ b/target/arm/translate-vfp.c @@ -XXX,XX +XXX,XX @@ static bool vfp_access_check_a(DisasContext *s, bool ignore_vfp_enabled) int coproc = arm_dc_feature(s, ARM_FEATURE_V8) ? 0 : 0xa; uint32_t syn = syn_fp_access_trap(1, 0xe, false, coproc); - gen_exception_insn_el(s, s->pc_curr, EXCP_UDEF, syn, s->fp_excp_el); + gen_exception_insn_el(s, 0, EXCP_UDEF, syn, s->fp_excp_el); return false; } @@ -XXX,XX +XXX,XX @@ static bool vfp_access_check_a(DisasContext *s, bool ignore_vfp_enabled) * appear to be any insns which touch VFP which are allowed. */ if (s->sme_trap_nonstreaming) { - gen_exception_insn(s, s->pc_curr, EXCP_UDEF, + gen_exception_insn(s, 0, EXCP_UDEF, syn_smetrap(SME_ET_Streaming, curr_insn_len(s) == 2)); return false; @@ -XXX,XX +XXX,XX @@ bool vfp_access_check_m(DisasContext *s, bool skip_context_update) * the encoding space handled by the patterns in m-nocp.decode, * and for them we may need to raise NOCP here. */ - gen_exception_insn_el(s, s->pc_curr, EXCP_NOCP, + gen_exception_insn_el(s, 0, EXCP_NOCP, syn_uncategorized(), s->fp_excp_el); return false; } diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static void gen_exception(int excp, uint32_t syndrome) tcg_constant_i32(syndrome)); } -static void gen_exception_insn_el_v(DisasContext *s, uint64_t pc, int excp, - uint32_t syn, TCGv_i32 tcg_el) +static void gen_exception_insn_el_v(DisasContext *s, target_long pc_diff, + int excp, uint32_t syn, TCGv_i32 tcg_el) { if (s->aarch64) { - gen_a64_update_pc(s, pc - s->pc_curr); + gen_a64_update_pc(s, pc_diff); } else { gen_set_condexec(s); - gen_update_pc(s, pc - s->pc_curr); + gen_update_pc(s, pc_diff); } gen_exception_el_v(excp, syn, tcg_el); s->base.is_jmp = DISAS_NORETURN; } -void gen_exception_insn_el(DisasContext *s, uint64_t pc, int excp, +void gen_exception_insn_el(DisasContext *s, target_long pc_diff, int excp, uint32_t syn, uint32_t target_el) { - gen_exception_insn_el_v(s, pc, excp, syn, tcg_constant_i32(target_el)); + gen_exception_insn_el_v(s, pc_diff, excp, syn, + tcg_constant_i32(target_el)); } -void gen_exception_insn(DisasContext *s, uint64_t pc, int excp, uint32_t syn) +void gen_exception_insn(DisasContext *s, target_long pc_diff, + int excp, uint32_t syn) { if (s->aarch64) { - gen_a64_update_pc(s, pc - s->pc_curr); + gen_a64_update_pc(s, pc_diff); } else { gen_set_condexec(s); - gen_update_pc(s, pc - s->pc_curr); + gen_update_pc(s, pc_diff); } gen_exception(excp, syn); s->base.is_jmp = DISAS_NORETURN; @@ -XXX,XX +XXX,XX @@ static void gen_exception_bkpt_insn(DisasContext *s, uint32_t syn) void unallocated_encoding(DisasContext *s) { /* Unallocated and reserved encodings are uncategorized */ - gen_exception_insn(s, s->pc_curr, EXCP_UDEF, syn_uncategorized()); + gen_exception_insn(s, 0, EXCP_UDEF, syn_uncategorized()); } /* Force a TB lookup after an instruction that changes the CPU state. */ @@ -XXX,XX +XXX,XX @@ static bool msr_banked_access_decode(DisasContext *s, int r, int sysm, int rn, tcg_el = tcg_constant_i32(3); } - gen_exception_insn_el_v(s, s->pc_curr, EXCP_UDEF, + gen_exception_insn_el_v(s, 0, EXCP_UDEF, syn_uncategorized(), tcg_el); tcg_temp_free_i32(tcg_el); return false; @@ -XXX,XX +XXX,XX @@ static bool msr_banked_access_decode(DisasContext *s, int r, int sysm, int rn, undef: /* If we get here then some access check did not pass */ - gen_exception_insn(s, s->pc_curr, EXCP_UDEF, syn_uncategorized()); + gen_exception_insn(s, 0, EXCP_UDEF, syn_uncategorized()); return false; } @@ -XXX,XX +XXX,XX @@ static void gen_srs(DisasContext *s, * For the UNPREDICTABLE cases we choose to UNDEF. */ if (s->current_el == 1 && !s->ns && mode == ARM_CPU_MODE_MON) { - gen_exception_insn_el(s, s->pc_curr, EXCP_UDEF, - syn_uncategorized(), 3); + gen_exception_insn_el(s, 0, EXCP_UDEF, syn_uncategorized(), 3); return; } @@ -XXX,XX +XXX,XX @@ static bool trans_WLS(DisasContext *s, arg_WLS *a) * Do the check-and-raise-exception by hand. */ if (s->fp_excp_el) { - gen_exception_insn_el(s, s->pc_curr, EXCP_NOCP, + gen_exception_insn_el(s, 0, EXCP_NOCP, syn_uncategorized(), s->fp_excp_el); return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_LE(DisasContext *s, arg_LE *a) tmp = load_cpu_field(v7m.ltpsize); tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, 4, skipexc); tcg_temp_free_i32(tmp); - gen_exception_insn(s, s->pc_curr, EXCP_INVSTATE, syn_uncategorized()); + gen_exception_insn(s, 0, EXCP_INVSTATE, syn_uncategorized()); gen_set_label(skipexc); } @@ -XXX,XX +XXX,XX @@ static void disas_arm_insn(DisasContext *s, unsigned int insn) * UsageFault exception. */ if (arm_dc_feature(s, ARM_FEATURE_M)) { - gen_exception_insn(s, s->pc_curr, EXCP_INVSTATE, syn_uncategorized()); + gen_exception_insn(s, 0, EXCP_INVSTATE, syn_uncategorized()); return; } @@ -XXX,XX +XXX,XX @@ static void disas_arm_insn(DisasContext *s, unsigned int insn) * Illegal execution state. This has priority over BTI * exceptions, but comes after instruction abort exceptions. */ - gen_exception_insn(s, s->pc_curr, EXCP_UDEF, syn_illegalstate()); + gen_exception_insn(s, 0, EXCP_UDEF, syn_illegalstate()); return; } @@ -XXX,XX +XXX,XX @@ static void thumb_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu) * Illegal execution state. This has priority over BTI * exceptions, but comes after instruction abort exceptions. */ - gen_exception_insn(dc, dc->pc_curr, EXCP_UDEF, syn_illegalstate()); + gen_exception_insn(dc, 0, EXCP_UDEF, syn_illegalstate()); return; } @@ -XXX,XX +XXX,XX @@ static void thumb_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu) */ tcg_remove_ops_after(dc->insn_eci_rewind); dc->condjmp = 0; - gen_exception_insn(dc, dc->pc_curr, EXCP_INVSTATE, - syn_uncategorized()); + gen_exception_insn(dc, 0, EXCP_INVSTATE, syn_uncategorized()); } arm_post_translate_insn(dc); -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> In preparation for TARGET_TB_PCREL, reduce reliance on absolute values. Since we always pass dc->pc_curr, fold the arithmetic to zero displacement. Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20221020030641.2066807-6-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-a64.c | 6 +++--- target/arm/translate.c | 10 +++++----- 2 files changed, 8 insertions(+), 8 deletions(-) diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a64.c +++ b/target/arm/translate-a64.c @@ -XXX,XX +XXX,XX @@ static void gen_exception_internal(int excp) gen_helper_exception_internal(cpu_env, tcg_constant_i32(excp)); } -static void gen_exception_internal_insn(DisasContext *s, uint64_t pc, int excp) +static void gen_exception_internal_insn(DisasContext *s, int excp) { - gen_a64_update_pc(s, pc - s->pc_curr); + gen_a64_update_pc(s, 0); gen_exception_internal(excp); s->base.is_jmp = DISAS_NORETURN; } @@ -XXX,XX +XXX,XX @@ static void disas_exc(DisasContext *s, uint32_t insn) * Secondly, "HLT 0xf000" is the A64 semihosting syscall instruction. */ if (semihosting_enabled(s->current_el == 0) && imm16 == 0xf000) { - gen_exception_internal_insn(s, s->pc_curr, EXCP_SEMIHOST); + gen_exception_internal_insn(s, EXCP_SEMIHOST); } else { unallocated_encoding(s); } diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static inline void gen_smc(DisasContext *s) s->base.is_jmp = DISAS_SMC; } -static void gen_exception_internal_insn(DisasContext *s, uint32_t pc, int excp) +static void gen_exception_internal_insn(DisasContext *s, int excp) { gen_set_condexec(s); - gen_update_pc(s, pc - s->pc_curr); + gen_update_pc(s, 0); gen_exception_internal(excp); s->base.is_jmp = DISAS_NORETURN; } @@ -XXX,XX +XXX,XX @@ static inline void gen_hlt(DisasContext *s, int imm) */ if (semihosting_enabled(s->current_el != 0) && (imm == (s->thumb ? 0x3c : 0xf000))) { - gen_exception_internal_insn(s, s->pc_curr, EXCP_SEMIHOST); + gen_exception_internal_insn(s, EXCP_SEMIHOST); return; } @@ -XXX,XX +XXX,XX @@ static bool trans_BKPT(DisasContext *s, arg_BKPT *a) if (arm_dc_feature(s, ARM_FEATURE_M) && semihosting_enabled(s->current_el == 0) && (a->imm == 0xab)) { - gen_exception_internal_insn(s, s->pc_curr, EXCP_SEMIHOST); + gen_exception_internal_insn(s, EXCP_SEMIHOST); } else { gen_exception_bkpt_insn(s, syn_aa32_bkpt(a->imm, false)); } @@ -XXX,XX +XXX,XX @@ static bool trans_SVC(DisasContext *s, arg_SVC *a) if (!arm_dc_feature(s, ARM_FEATURE_M) && semihosting_enabled(s->current_el == 0) && (a->imm == semihost_imm)) { - gen_exception_internal_insn(s, s->pc_curr, EXCP_SEMIHOST); + gen_exception_internal_insn(s, EXCP_SEMIHOST); } else { gen_update_pc(s, curr_insn_len(s)); s->svc_imm = a->imm; -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> In preparation for TARGET_TB_PCREL, reduce reliance on absolute values. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20221020030641.2066807-7-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate.c | 37 +++++++++++++++++++++---------------- 1 file changed, 21 insertions(+), 16 deletions(-) diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static uint32_t read_pc(DisasContext *s) return s->pc_curr + (s->thumb ? 4 : 8); } +/* The pc_curr difference for an architectural jump. */ +static target_long jmp_diff(DisasContext *s, target_long diff) +{ + return diff + (s->thumb ? 4 : 8); +} + /* Set a variable to the value of a CPU register. */ void load_reg_var(DisasContext *s, TCGv_i32 var, int reg) { @@ -XXX,XX +XXX,XX @@ static void gen_goto_ptr(void) * cpu_loop_exec. Any live exit_requests will be processed as we * enter the next TB. */ -static void gen_goto_tb(DisasContext *s, int n, int diff) +static void gen_goto_tb(DisasContext *s, int n, target_long diff) { target_ulong dest = s->pc_curr + diff; @@ -XXX,XX +XXX,XX @@ static void gen_goto_tb(DisasContext *s, int n, int diff) } /* Jump, specifying which TB number to use if we gen_goto_tb() */ -static inline void gen_jmp_tb(DisasContext *s, uint32_t dest, int tbno) +static void gen_jmp_tb(DisasContext *s, target_long diff, int tbno) { - int diff = dest - s->pc_curr; - if (unlikely(s->ss_active)) { /* An indirect jump so that we still trigger the debug exception. */ gen_update_pc(s, diff); @@ -XXX,XX +XXX,XX @@ static inline void gen_jmp_tb(DisasContext *s, uint32_t dest, int tbno) } } -static inline void gen_jmp(DisasContext *s, uint32_t dest) +static inline void gen_jmp(DisasContext *s, target_long diff) { - gen_jmp_tb(s, dest, 0); + gen_jmp_tb(s, diff, 0); } static inline void gen_mulxy(TCGv_i32 t0, TCGv_i32 t1, int x, int y) @@ -XXX,XX +XXX,XX @@ static bool trans_CLRM(DisasContext *s, arg_CLRM *a) static bool trans_B(DisasContext *s, arg_i *a) { - gen_jmp(s, read_pc(s) + a->imm); + gen_jmp(s, jmp_diff(s, a->imm)); return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_B_cond_thumb(DisasContext *s, arg_ci *a) return true; } arm_skip_unless(s, a->cond); - gen_jmp(s, read_pc(s) + a->imm); + gen_jmp(s, jmp_diff(s, a->imm)); return true; } static bool trans_BL(DisasContext *s, arg_i *a) { tcg_gen_movi_i32(cpu_R[14], s->base.pc_next | s->thumb); - gen_jmp(s, read_pc(s) + a->imm); + gen_jmp(s, jmp_diff(s, a->imm)); return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_BLX_i(DisasContext *s, arg_BLX_i *a) } tcg_gen_movi_i32(cpu_R[14], s->base.pc_next | s->thumb); store_cpu_field_constant(!s->thumb, thumb); - gen_jmp(s, (read_pc(s) & ~3) + a->imm); + /* This jump is computed from an aligned PC: subtract off the low bits. */ + gen_jmp(s, jmp_diff(s, a->imm - (s->pc_curr & 3))); return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_WLS(DisasContext *s, arg_WLS *a) * when we take this upcoming exit from this TB, so gen_jmp_tb() is OK. */ } - gen_jmp_tb(s, s->base.pc_next, 1); + gen_jmp_tb(s, curr_insn_len(s), 1); gen_set_label(nextlabel); - gen_jmp(s, read_pc(s) + a->imm); + gen_jmp(s, jmp_diff(s, a->imm)); return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_LE(DisasContext *s, arg_LE *a) if (a->f) { /* Loop-forever: just jump back to the loop start */ - gen_jmp(s, read_pc(s) - a->imm); + gen_jmp(s, jmp_diff(s, -a->imm)); return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_LE(DisasContext *s, arg_LE *a) tcg_temp_free_i32(decr); } /* Jump back to the loop start */ - gen_jmp(s, read_pc(s) - a->imm); + gen_jmp(s, jmp_diff(s, -a->imm)); gen_set_label(loopend); if (a->tp) { @@ -XXX,XX +XXX,XX @@ static bool trans_LE(DisasContext *s, arg_LE *a) store_cpu_field(tcg_constant_i32(4), v7m.ltpsize); } /* End TB, continuing to following insn */ - gen_jmp_tb(s, s->base.pc_next, 1); + gen_jmp_tb(s, curr_insn_len(s), 1); return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_CBZ(DisasContext *s, arg_CBZ *a) tcg_gen_brcondi_i32(a->nz ? TCG_COND_EQ : TCG_COND_NE, tmp, 0, s->condlabel); tcg_temp_free_i32(tmp); - gen_jmp(s, read_pc(s) + a->imm); + gen_jmp(s, jmp_diff(s, a->imm)); return true; } -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> In preparation for TARGET_TB_PCREL, reduce reliance on absolute values. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20221020030641.2066807-8-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-a64.c | 41 +++++++++++++++++++++++++++----------- 1 file changed, 29 insertions(+), 12 deletions(-) diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a64.c +++ b/target/arm/translate-a64.c @@ -XXX,XX +XXX,XX @@ static void reset_btype(DisasContext *s) } } +static void gen_pc_plus_diff(DisasContext *s, TCGv_i64 dest, target_long diff) +{ + tcg_gen_movi_i64(dest, s->pc_curr + diff); +} + void gen_a64_update_pc(DisasContext *s, target_long diff) { - tcg_gen_movi_i64(cpu_pc, s->pc_curr + diff); + gen_pc_plus_diff(s, cpu_pc, diff); } /* @@ -XXX,XX +XXX,XX @@ static void disas_uncond_b_imm(DisasContext *s, uint32_t insn) if (insn & (1U << 31)) { /* BL Branch with link */ - tcg_gen_movi_i64(cpu_reg(s, 30), s->base.pc_next); + gen_pc_plus_diff(s, cpu_reg(s, 30), curr_insn_len(s)); } /* B Branch / BL Branch with link */ @@ -XXX,XX +XXX,XX @@ static void disas_uncond_b_reg(DisasContext *s, uint32_t insn) default: goto do_unallocated; } - gen_a64_set_pc(s, dst); /* BLR also needs to load return address */ if (opc == 1) { - tcg_gen_movi_i64(cpu_reg(s, 30), s->base.pc_next); + TCGv_i64 lr = cpu_reg(s, 30); + if (dst == lr) { + TCGv_i64 tmp = new_tmp_a64(s); + tcg_gen_mov_i64(tmp, dst); + dst = tmp; + } + gen_pc_plus_diff(s, lr, curr_insn_len(s)); } + gen_a64_set_pc(s, dst); break; case 8: /* BRAA */ @@ -XXX,XX +XXX,XX @@ static void disas_uncond_b_reg(DisasContext *s, uint32_t insn) } else { dst = cpu_reg(s, rn); } - gen_a64_set_pc(s, dst); /* BLRAA also needs to load return address */ if (opc == 9) { - tcg_gen_movi_i64(cpu_reg(s, 30), s->base.pc_next); + TCGv_i64 lr = cpu_reg(s, 30); + if (dst == lr) { + TCGv_i64 tmp = new_tmp_a64(s); + tcg_gen_mov_i64(tmp, dst); + dst = tmp; + } + gen_pc_plus_diff(s, lr, curr_insn_len(s)); } + gen_a64_set_pc(s, dst); break; case 4: /* ERET */ @@ -XXX,XX +XXX,XX @@ static void disas_ld_lit(DisasContext *s, uint32_t insn) tcg_rt = cpu_reg(s, rt); - clean_addr = tcg_constant_i64(s->pc_curr + imm); + clean_addr = new_tmp_a64(s); + gen_pc_plus_diff(s, clean_addr, imm); if (is_vector) { do_fp_ld(s, rt, clean_addr, size); } else { @@ -XXX,XX +XXX,XX @@ static void disas_ldst(DisasContext *s, uint32_t insn) static void disas_pc_rel_adr(DisasContext *s, uint32_t insn) { unsigned int page, rd; - uint64_t base; - uint64_t offset; + int64_t offset; page = extract32(insn, 31, 1); /* SignExtend(immhi:immlo) -> offset */ offset = sextract64(insn, 5, 19); offset = offset << 2 | extract32(insn, 29, 2); rd = extract32(insn, 0, 5); - base = s->pc_curr; if (page) { /* ADRP (page based) */ - base &= ~0xfff; offset <<= 12; + /* The page offset is ok for TARGET_TB_PCREL. */ + offset -= s->pc_curr & 0xfff; } - tcg_gen_movi_i64(cpu_reg(s, rd), base + offset); + gen_pc_plus_diff(s, cpu_reg(s, rd), offset); } /* -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> In preparation for TARGET_TB_PCREL, reduce reliance on absolute values. Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20221020030641.2066807-9-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate.c | 38 +++++++++++++++++++++----------------- 1 file changed, 21 insertions(+), 17 deletions(-) diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static inline int get_a32_user_mem_index(DisasContext *s) } } -/* The architectural value of PC. */ -static uint32_t read_pc(DisasContext *s) -{ - return s->pc_curr + (s->thumb ? 4 : 8); -} - /* The pc_curr difference for an architectural jump. */ static target_long jmp_diff(DisasContext *s, target_long diff) { return diff + (s->thumb ? 4 : 8); } +static void gen_pc_plus_diff(DisasContext *s, TCGv_i32 var, target_long diff) +{ + tcg_gen_movi_i32(var, s->pc_curr + diff); +} + /* Set a variable to the value of a CPU register. */ void load_reg_var(DisasContext *s, TCGv_i32 var, int reg) { if (reg == 15) { - tcg_gen_movi_i32(var, read_pc(s)); + gen_pc_plus_diff(s, var, jmp_diff(s, 0)); } else { tcg_gen_mov_i32(var, cpu_R[reg]); } @@ -XXX,XX +XXX,XX @@ TCGv_i32 add_reg_for_lit(DisasContext *s, int reg, int ofs) TCGv_i32 tmp = tcg_temp_new_i32(); if (reg == 15) { - tcg_gen_movi_i32(tmp, (read_pc(s) & ~3) + ofs); + /* + * This address is computed from an aligned PC: + * subtract off the low bits. + */ + gen_pc_plus_diff(s, tmp, jmp_diff(s, ofs - (s->pc_curr & 3))); } else { tcg_gen_addi_i32(tmp, cpu_R[reg], ofs); } @@ -XXX,XX +XXX,XX @@ void unallocated_encoding(DisasContext *s) /* Force a TB lookup after an instruction that changes the CPU state. */ void gen_lookup_tb(DisasContext *s) { - tcg_gen_movi_i32(cpu_R[15], s->base.pc_next); + gen_pc_plus_diff(s, cpu_R[15], curr_insn_len(s)); s->base.is_jmp = DISAS_EXIT; } @@ -XXX,XX +XXX,XX @@ static bool trans_BLX_r(DisasContext *s, arg_BLX_r *a) return false; } tmp = load_reg(s, a->rm); - tcg_gen_movi_i32(cpu_R[14], s->base.pc_next | s->thumb); + gen_pc_plus_diff(s, cpu_R[14], curr_insn_len(s) | s->thumb); gen_bx(s, tmp); return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_B_cond_thumb(DisasContext *s, arg_ci *a) static bool trans_BL(DisasContext *s, arg_i *a) { - tcg_gen_movi_i32(cpu_R[14], s->base.pc_next | s->thumb); + gen_pc_plus_diff(s, cpu_R[14], curr_insn_len(s) | s->thumb); gen_jmp(s, jmp_diff(s, a->imm)); return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_BLX_i(DisasContext *s, arg_BLX_i *a) if (s->thumb && (a->imm & 2)) { return false; } - tcg_gen_movi_i32(cpu_R[14], s->base.pc_next | s->thumb); + gen_pc_plus_diff(s, cpu_R[14], curr_insn_len(s) | s->thumb); store_cpu_field_constant(!s->thumb, thumb); /* This jump is computed from an aligned PC: subtract off the low bits. */ gen_jmp(s, jmp_diff(s, a->imm - (s->pc_curr & 3))); @@ -XXX,XX +XXX,XX @@ static bool trans_BLX_i(DisasContext *s, arg_BLX_i *a) static bool trans_BL_BLX_prefix(DisasContext *s, arg_BL_BLX_prefix *a) { assert(!arm_dc_feature(s, ARM_FEATURE_THUMB2)); - tcg_gen_movi_i32(cpu_R[14], read_pc(s) + (a->imm << 12)); + gen_pc_plus_diff(s, cpu_R[14], jmp_diff(s, a->imm << 12)); return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_BL_suffix(DisasContext *s, arg_BL_suffix *a) assert(!arm_dc_feature(s, ARM_FEATURE_THUMB2)); tcg_gen_addi_i32(tmp, cpu_R[14], (a->imm << 1) | 1); - tcg_gen_movi_i32(cpu_R[14], s->base.pc_next | 1); + gen_pc_plus_diff(s, cpu_R[14], curr_insn_len(s) | 1); gen_bx(s, tmp); return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_BLX_suffix(DisasContext *s, arg_BLX_suffix *a) tmp = tcg_temp_new_i32(); tcg_gen_addi_i32(tmp, cpu_R[14], a->imm << 1); tcg_gen_andi_i32(tmp, tmp, 0xfffffffc); - tcg_gen_movi_i32(cpu_R[14], s->base.pc_next | 1); + gen_pc_plus_diff(s, cpu_R[14], curr_insn_len(s) | 1); gen_bx(s, tmp); return true; } @@ -XXX,XX +XXX,XX @@ static bool op_tbranch(DisasContext *s, arg_tbranch *a, bool half) tcg_gen_add_i32(addr, addr, tmp); gen_aa32_ld_i32(s, tmp, addr, get_mem_index(s), half ? MO_UW : MO_UB); - tcg_temp_free_i32(addr); tcg_gen_add_i32(tmp, tmp, tmp); - tcg_gen_addi_i32(tmp, tmp, read_pc(s)); + gen_pc_plus_diff(s, addr, jmp_diff(s, 0)); + tcg_gen_add_i32(tmp, tmp, addr); + tcg_temp_free_i32(addr); store_reg(s, 15, tmp); return true; } -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20221020030641.2066807-10-richard.henderson@linaro.org Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/cpu-param.h | 2 + target/arm/translate.h | 50 +++++++++++++++- target/arm/cpu.c | 23 ++++---- target/arm/translate-a64.c | 64 +++++++++++++------- target/arm/translate-m-nocp.c | 2 +- target/arm/translate.c | 108 +++++++++++++++++++++++----------- 6 files changed, 178 insertions(+), 71 deletions(-) diff --git a/target/arm/cpu-param.h b/target/arm/cpu-param.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu-param.h +++ b/target/arm/cpu-param.h @@ -XXX,XX +XXX,XX @@ # define TARGET_PAGE_BITS_VARY # define TARGET_PAGE_BITS_MIN 10 +# define TARGET_TB_PCREL 1 + /* * Cache the attrs and shareability fields from the page table entry. * diff --git a/target/arm/translate.h b/target/arm/translate.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.h +++ b/target/arm/translate.h @@ -XXX,XX +XXX,XX @@ /* internal defines */ + +/* + * Save pc_save across a branch, so that we may restore the value from + * before the branch at the point the label is emitted. + */ +typedef struct DisasLabel { + TCGLabel *label; + target_ulong pc_save; +} DisasLabel; + typedef struct DisasContext { DisasContextBase base; const ARMISARegisters *isar; /* The address of the current instruction being translated. */ target_ulong pc_curr; + /* + * For TARGET_TB_PCREL, the full value of cpu_pc is not known + * (although the page offset is known). For convenience, the + * translation loop uses the full virtual address that triggered + * the translation, from base.pc_start through pc_curr. + * For efficiency, we do not update cpu_pc for every instruction. + * Instead, pc_save has the value of pc_curr at the time of the + * last update to cpu_pc, which allows us to compute the addend + * needed to bring cpu_pc current: pc_curr - pc_save. + * If cpu_pc now contains the destination of an indirect branch, + * pc_save contains -1 to indicate that relative updates are no + * longer possible. + */ + target_ulong pc_save; target_ulong page_start; uint32_t insn; /* Nonzero if this instruction has been conditionally skipped. */ int condjmp; /* The label that will be jumped to when the instruction is skipped. */ - TCGLabel *condlabel; + DisasLabel condlabel; /* Thumb-2 conditional execution bits. */ int condexec_mask; int condexec_cond; @@ -XXX,XX +XXX,XX @@ typedef struct DisasContext { * after decode (ie after any UNDEF checks) */ bool eci_handled; - /* TCG op to rewind to if this turns out to be an invalid ECI state */ - TCGOp *insn_eci_rewind; int sctlr_b; MemOp be_data; #if !defined(CONFIG_USER_ONLY) @@ -XXX,XX +XXX,XX @@ static inline MemOp finalize_memop(DisasContext *s, MemOp opc) */ uint64_t asimd_imm_const(uint32_t imm, int cmode, int op); +/* + * gen_disas_label: + * Create a label and cache a copy of pc_save. + */ +static inline DisasLabel gen_disas_label(DisasContext *s) +{ + return (DisasLabel){ + .label = gen_new_label(), + .pc_save = s->pc_save, + }; +} + +/* + * set_disas_label: + * Emit a label and restore the cached copy of pc_save. + */ +static inline void set_disas_label(DisasContext *s, DisasLabel l) +{ + gen_set_label(l.label); + s->pc_save = l.pc_save; +} + /* * Helpers for implementing sets of trans_* functions. * Defer the implementation of NAME to FUNC, with optional extra arguments. diff --git a/target/arm/cpu.c b/target/arm/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.c +++ b/target/arm/cpu.c @@ -XXX,XX +XXX,XX @@ static vaddr arm_cpu_get_pc(CPUState *cs) void arm_cpu_synchronize_from_tb(CPUState *cs, const TranslationBlock *tb) { - ARMCPU *cpu = ARM_CPU(cs); - CPUARMState *env = &cpu->env; - - /* - * It's OK to look at env for the current mode here, because it's - * never possible for an AArch64 TB to chain to an AArch32 TB. - */ - if (is_a64(env)) { - env->pc = tb_pc(tb); - } else { - env->regs[15] = tb_pc(tb); + /* The program counter is always up to date with TARGET_TB_PCREL. */ + if (!TARGET_TB_PCREL) { + CPUARMState *env = cs->env_ptr; + /* + * It's OK to look at env for the current mode here, because it's + * never possible for an AArch64 TB to chain to an AArch32 TB. + */ + if (is_a64(env)) { + env->pc = tb_pc(tb); + } else { + env->regs[15] = tb_pc(tb); + } } } #endif /* CONFIG_TCG */ diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a64.c +++ b/target/arm/translate-a64.c @@ -XXX,XX +XXX,XX @@ static void reset_btype(DisasContext *s) static void gen_pc_plus_diff(DisasContext *s, TCGv_i64 dest, target_long diff) { - tcg_gen_movi_i64(dest, s->pc_curr + diff); + assert(s->pc_save != -1); + if (TARGET_TB_PCREL) { + tcg_gen_addi_i64(dest, cpu_pc, (s->pc_curr - s->pc_save) + diff); + } else { + tcg_gen_movi_i64(dest, s->pc_curr + diff); + } } void gen_a64_update_pc(DisasContext *s, target_long diff) { gen_pc_plus_diff(s, cpu_pc, diff); + s->pc_save = s->pc_curr + diff; } /* @@ -XXX,XX +XXX,XX @@ static void gen_a64_set_pc(DisasContext *s, TCGv_i64 src) * then loading an address into the PC will clear out any tag. */ gen_top_byte_ignore(s, cpu_pc, src, s->tbii); + s->pc_save = -1; } /* @@ -XXX,XX +XXX,XX @@ static inline bool use_goto_tb(DisasContext *s, uint64_t dest) static void gen_goto_tb(DisasContext *s, int n, int64_t diff) { - uint64_t dest = s->pc_curr + diff; - - if (use_goto_tb(s, dest)) { - tcg_gen_goto_tb(n); - gen_a64_update_pc(s, diff); + if (use_goto_tb(s, s->pc_curr + diff)) { + /* + * For pcrel, the pc must always be up-to-date on entry to + * the linked TB, so that it can use simple additions for all + * further adjustments. For !pcrel, the linked TB is compiled + * to know its full virtual address, so we can delay the + * update to pc to the unlinked path. A long chain of links + * can thus avoid many updates to the PC. + */ + if (TARGET_TB_PCREL) { + gen_a64_update_pc(s, diff); + tcg_gen_goto_tb(n); + } else { + tcg_gen_goto_tb(n); + gen_a64_update_pc(s, diff); + } tcg_gen_exit_tb(s->base.tb, n); s->base.is_jmp = DISAS_NORETURN; } else { @@ -XXX,XX +XXX,XX @@ static void disas_comp_b_imm(DisasContext *s, uint32_t insn) { unsigned int sf, op, rt; int64_t diff; - TCGLabel *label_match; + DisasLabel match; TCGv_i64 tcg_cmp; sf = extract32(insn, 31, 1); @@ -XXX,XX +XXX,XX @@ static void disas_comp_b_imm(DisasContext *s, uint32_t insn) diff = sextract32(insn, 5, 19) * 4; tcg_cmp = read_cpu_reg(s, rt, sf); - label_match = gen_new_label(); - reset_btype(s); - tcg_gen_brcondi_i64(op ? TCG_COND_NE : TCG_COND_EQ, - tcg_cmp, 0, label_match); + match = gen_disas_label(s); + tcg_gen_brcondi_i64(op ? TCG_COND_NE : TCG_COND_EQ, + tcg_cmp, 0, match.label); gen_goto_tb(s, 0, 4); - gen_set_label(label_match); + set_disas_label(s, match); gen_goto_tb(s, 1, diff); } @@ -XXX,XX +XXX,XX @@ static void disas_test_b_imm(DisasContext *s, uint32_t insn) { unsigned int bit_pos, op, rt; int64_t diff; - TCGLabel *label_match; + DisasLabel match; TCGv_i64 tcg_cmp; bit_pos = (extract32(insn, 31, 1) << 5) | extract32(insn, 19, 5); @@ -XXX,XX +XXX,XX @@ static void disas_test_b_imm(DisasContext *s, uint32_t insn) tcg_cmp = tcg_temp_new_i64(); tcg_gen_andi_i64(tcg_cmp, cpu_reg(s, rt), (1ULL << bit_pos)); - label_match = gen_new_label(); reset_btype(s); + + match = gen_disas_label(s); tcg_gen_brcondi_i64(op ? TCG_COND_NE : TCG_COND_EQ, - tcg_cmp, 0, label_match); + tcg_cmp, 0, match.label); tcg_temp_free_i64(tcg_cmp); gen_goto_tb(s, 0, 4); - gen_set_label(label_match); + set_disas_label(s, match); gen_goto_tb(s, 1, diff); } @@ -XXX,XX +XXX,XX @@ static void disas_cond_b_imm(DisasContext *s, uint32_t insn) reset_btype(s); if (cond < 0x0e) { /* genuinely conditional branches */ - TCGLabel *label_match = gen_new_label(); - arm_gen_test_cc(cond, label_match); + DisasLabel match = gen_disas_label(s); + arm_gen_test_cc(cond, match.label); gen_goto_tb(s, 0, 4); - gen_set_label(label_match); + set_disas_label(s, match); gen_goto_tb(s, 1, diff); } else { /* 0xe and 0xf are both "always" conditions */ @@ -XXX,XX +XXX,XX @@ static void aarch64_tr_init_disas_context(DisasContextBase *dcbase, dc->isar = &arm_cpu->isar; dc->condjmp = 0; - + dc->pc_save = dc->base.pc_first; dc->aarch64 = true; dc->thumb = false; dc->sctlr_b = 0; @@ -XXX,XX +XXX,XX @@ static void aarch64_tr_tb_start(DisasContextBase *db, CPUState *cpu) static void aarch64_tr_insn_start(DisasContextBase *dcbase, CPUState *cpu) { DisasContext *dc = container_of(dcbase, DisasContext, base); + target_ulong pc_arg = dc->base.pc_next; - tcg_gen_insn_start(dc->base.pc_next, 0, 0); + if (TARGET_TB_PCREL) { + pc_arg &= ~TARGET_PAGE_MASK; + } + tcg_gen_insn_start(pc_arg, 0, 0); dc->insn_start = tcg_last_op(); } diff --git a/target/arm/translate-m-nocp.c b/target/arm/translate-m-nocp.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-m-nocp.c +++ b/target/arm/translate-m-nocp.c @@ -XXX,XX +XXX,XX @@ static bool trans_VSCCLRM(DisasContext *s, arg_VSCCLRM *a) tcg_gen_andi_i32(sfpa, sfpa, R_V7M_CONTROL_SFPA_MASK); tcg_gen_or_i32(sfpa, sfpa, aspen); arm_gen_condlabel(s); - tcg_gen_brcondi_i32(TCG_COND_EQ, sfpa, 0, s->condlabel); + tcg_gen_brcondi_i32(TCG_COND_EQ, sfpa, 0, s->condlabel.label); if (s->fp_excp_el != 0) { gen_exception_insn_el(s, 0, EXCP_NOCP, diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ uint64_t asimd_imm_const(uint32_t imm, int cmode, int op) void arm_gen_condlabel(DisasContext *s) { if (!s->condjmp) { - s->condlabel = gen_new_label(); + s->condlabel = gen_disas_label(s); s->condjmp = 1; } } @@ -XXX,XX +XXX,XX @@ static target_long jmp_diff(DisasContext *s, target_long diff) static void gen_pc_plus_diff(DisasContext *s, TCGv_i32 var, target_long diff) { - tcg_gen_movi_i32(var, s->pc_curr + diff); + assert(s->pc_save != -1); + if (TARGET_TB_PCREL) { + tcg_gen_addi_i32(var, cpu_R[15], (s->pc_curr - s->pc_save) + diff); + } else { + tcg_gen_movi_i32(var, s->pc_curr + diff); + } } /* Set a variable to the value of a CPU register. */ @@ -XXX,XX +XXX,XX @@ void store_reg(DisasContext *s, int reg, TCGv_i32 var) */ tcg_gen_andi_i32(var, var, s->thumb ? ~1 : ~3); s->base.is_jmp = DISAS_JUMP; + s->pc_save = -1; } else if (reg == 13 && arm_dc_feature(s, ARM_FEATURE_M)) { /* For M-profile SP bits [1:0] are always zero */ tcg_gen_andi_i32(var, var, ~3); @@ -XXX,XX +XXX,XX @@ void gen_set_condexec(DisasContext *s) void gen_update_pc(DisasContext *s, target_long diff) { - tcg_gen_movi_i32(cpu_R[15], s->pc_curr + diff); + gen_pc_plus_diff(s, cpu_R[15], diff); + s->pc_save = s->pc_curr + diff; } /* Set PC and Thumb state from var. var is marked as dead. */ @@ -XXX,XX +XXX,XX @@ static inline void gen_bx(DisasContext *s, TCGv_i32 var) tcg_gen_andi_i32(cpu_R[15], var, ~1); tcg_gen_andi_i32(var, var, 1); store_cpu_field(var, thumb); + s->pc_save = -1; } /* @@ -XXX,XX +XXX,XX @@ static inline void gen_bx_excret(DisasContext *s, TCGv_i32 var) static inline void gen_bx_excret_final_code(DisasContext *s) { /* Generate the code to finish possible exception return and end the TB */ - TCGLabel *excret_label = gen_new_label(); + DisasLabel excret_label = gen_disas_label(s); uint32_t min_magic; if (arm_dc_feature(s, ARM_FEATURE_M_SECURITY)) { @@ -XXX,XX +XXX,XX @@ static inline void gen_bx_excret_final_code(DisasContext *s) } /* Is the new PC value in the magic range indicating exception return? */ - tcg_gen_brcondi_i32(TCG_COND_GEU, cpu_R[15], min_magic, excret_label); + tcg_gen_brcondi_i32(TCG_COND_GEU, cpu_R[15], min_magic, excret_label.label); /* No: end the TB as we would for a DISAS_JMP */ if (s->ss_active) { gen_singlestep_exception(s); } else { tcg_gen_exit_tb(NULL, 0); } - gen_set_label(excret_label); + set_disas_label(s, excret_label); /* Yes: this is an exception return. * At this point in runtime env->regs[15] and env->thumb will hold * the exception-return magic number, which do_v7m_exception_exit() @@ -XXX,XX +XXX,XX @@ static void gen_goto_ptr(void) */ static void gen_goto_tb(DisasContext *s, int n, target_long diff) { - target_ulong dest = s->pc_curr + diff; - - if (translator_use_goto_tb(&s->base, dest)) { - tcg_gen_goto_tb(n); - gen_update_pc(s, diff); + if (translator_use_goto_tb(&s->base, s->pc_curr + diff)) { + /* + * For pcrel, the pc must always be up-to-date on entry to + * the linked TB, so that it can use simple additions for all + * further adjustments. For !pcrel, the linked TB is compiled + * to know its full virtual address, so we can delay the + * update to pc to the unlinked path. A long chain of links + * can thus avoid many updates to the PC. + */ + if (TARGET_TB_PCREL) { + gen_update_pc(s, diff); + tcg_gen_goto_tb(n); + } else { + tcg_gen_goto_tb(n); + gen_update_pc(s, diff); + } tcg_gen_exit_tb(s->base.tb, n); } else { gen_update_pc(s, diff); @@ -XXX,XX +XXX,XX @@ static void gen_srs(DisasContext *s, static void arm_skip_unless(DisasContext *s, uint32_t cond) { arm_gen_condlabel(s); - arm_gen_test_cc(cond ^ 1, s->condlabel); + arm_gen_test_cc(cond ^ 1, s->condlabel.label); } @@ -XXX,XX +XXX,XX @@ static bool trans_WLS(DisasContext *s, arg_WLS *a) { /* M-profile low-overhead while-loop start */ TCGv_i32 tmp; - TCGLabel *nextlabel; + DisasLabel nextlabel; if (!dc_isar_feature(aa32_lob, s)) { return false; @@ -XXX,XX +XXX,XX @@ static bool trans_WLS(DisasContext *s, arg_WLS *a) } } - nextlabel = gen_new_label(); - tcg_gen_brcondi_i32(TCG_COND_EQ, cpu_R[a->rn], 0, nextlabel); + nextlabel = gen_disas_label(s); + tcg_gen_brcondi_i32(TCG_COND_EQ, cpu_R[a->rn], 0, nextlabel.label); tmp = load_reg(s, a->rn); store_reg(s, 14, tmp); if (a->size != 4) { @@ -XXX,XX +XXX,XX @@ static bool trans_WLS(DisasContext *s, arg_WLS *a) } gen_jmp_tb(s, curr_insn_len(s), 1); - gen_set_label(nextlabel); + set_disas_label(s, nextlabel); gen_jmp(s, jmp_diff(s, a->imm)); return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_LE(DisasContext *s, arg_LE *a) * any faster. */ TCGv_i32 tmp; - TCGLabel *loopend; + DisasLabel loopend; bool fpu_active; if (!dc_isar_feature(aa32_lob, s)) { @@ -XXX,XX +XXX,XX @@ static bool trans_LE(DisasContext *s, arg_LE *a) if (!a->tp && dc_isar_feature(aa32_mve, s) && fpu_active) { /* Need to do a runtime check for LTPSIZE != 4 */ - TCGLabel *skipexc = gen_new_label(); + DisasLabel skipexc = gen_disas_label(s); tmp = load_cpu_field(v7m.ltpsize); - tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, 4, skipexc); + tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, 4, skipexc.label); tcg_temp_free_i32(tmp); gen_exception_insn(s, 0, EXCP_INVSTATE, syn_uncategorized()); - gen_set_label(skipexc); + set_disas_label(s, skipexc); } if (a->f) { @@ -XXX,XX +XXX,XX @@ static bool trans_LE(DisasContext *s, arg_LE *a) * loop decrement value is 1. For LETP we need to calculate the decrement * value from LTPSIZE. */ - loopend = gen_new_label(); + loopend = gen_disas_label(s); if (!a->tp) { - tcg_gen_brcondi_i32(TCG_COND_LEU, cpu_R[14], 1, loopend); + tcg_gen_brcondi_i32(TCG_COND_LEU, cpu_R[14], 1, loopend.label); tcg_gen_addi_i32(cpu_R[14], cpu_R[14], -1); } else { /* @@ -XXX,XX +XXX,XX @@ static bool trans_LE(DisasContext *s, arg_LE *a) tcg_gen_shl_i32(decr, tcg_constant_i32(1), decr); tcg_temp_free_i32(ltpsize); - tcg_gen_brcond_i32(TCG_COND_LEU, cpu_R[14], decr, loopend); + tcg_gen_brcond_i32(TCG_COND_LEU, cpu_R[14], decr, loopend.label); tcg_gen_sub_i32(cpu_R[14], cpu_R[14], decr); tcg_temp_free_i32(decr); @@ -XXX,XX +XXX,XX @@ static bool trans_LE(DisasContext *s, arg_LE *a) /* Jump back to the loop start */ gen_jmp(s, jmp_diff(s, -a->imm)); - gen_set_label(loopend); + set_disas_label(s, loopend); if (a->tp) { /* Exits from tail-pred loops must reset LTPSIZE to 4 */ store_cpu_field(tcg_constant_i32(4), v7m.ltpsize); @@ -XXX,XX +XXX,XX @@ static bool trans_CBZ(DisasContext *s, arg_CBZ *a) arm_gen_condlabel(s); tcg_gen_brcondi_i32(a->nz ? TCG_COND_EQ : TCG_COND_NE, - tmp, 0, s->condlabel); + tmp, 0, s->condlabel.label); tcg_temp_free_i32(tmp); gen_jmp(s, jmp_diff(s, a->imm)); return true; @@ -XXX,XX +XXX,XX @@ static void arm_tr_init_disas_context(DisasContextBase *dcbase, CPUState *cs) dc->isar = &cpu->isar; dc->condjmp = 0; - + dc->pc_save = dc->base.pc_first; dc->aarch64 = false; dc->thumb = EX_TBFLAG_AM32(tb_flags, THUMB); dc->be_data = EX_TBFLAG_ANY(tb_flags, BE_DATA) ? MO_BE : MO_LE; @@ -XXX,XX +XXX,XX @@ static void arm_tr_init_disas_context(DisasContextBase *dcbase, CPUState *cs) */ dc->eci = dc->condexec_mask = dc->condexec_cond = 0; dc->eci_handled = false; - dc->insn_eci_rewind = NULL; if (condexec & 0xf) { dc->condexec_mask = (condexec & 0xf) << 1; dc->condexec_cond = condexec >> 4; @@ -XXX,XX +XXX,XX @@ static void arm_tr_insn_start(DisasContextBase *dcbase, CPUState *cpu) * fields here. */ uint32_t condexec_bits; + target_ulong pc_arg = dc->base.pc_next; + if (TARGET_TB_PCREL) { + pc_arg &= ~TARGET_PAGE_MASK; + } if (dc->eci) { condexec_bits = dc->eci << 4; } else { condexec_bits = (dc->condexec_cond << 4) | (dc->condexec_mask >> 1); } - tcg_gen_insn_start(dc->base.pc_next, condexec_bits, 0); + tcg_gen_insn_start(pc_arg, condexec_bits, 0); dc->insn_start = tcg_last_op(); } @@ -XXX,XX +XXX,XX @@ static bool arm_check_ss_active(DisasContext *dc) static void arm_post_translate_insn(DisasContext *dc) { - if (dc->condjmp && !dc->base.is_jmp) { - gen_set_label(dc->condlabel); + if (dc->condjmp && dc->base.is_jmp == DISAS_NEXT) { + if (dc->pc_save != dc->condlabel.pc_save) { + gen_update_pc(dc, dc->condlabel.pc_save - dc->pc_save); + } + gen_set_label(dc->condlabel.label); dc->condjmp = 0; } translator_loop_temp_check(&dc->base); @@ -XXX,XX +XXX,XX @@ static void thumb_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu) uint32_t pc = dc->base.pc_next; uint32_t insn; bool is_16bit; + /* TCG op to rewind to if this turns out to be an invalid ECI state */ + TCGOp *insn_eci_rewind = NULL; + target_ulong insn_eci_pc_save = -1; /* Misaligned thumb PC is architecturally impossible. */ assert((dc->base.pc_next & 1) == 0); @@ -XXX,XX +XXX,XX @@ static void thumb_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu) * insn" case. We will rewind to the marker (ie throwing away * all the generated code) and instead emit "take exception". */ - dc->insn_eci_rewind = tcg_last_op(); + insn_eci_rewind = tcg_last_op(); + insn_eci_pc_save = dc->pc_save; } if (dc->condexec_mask && !thumb_insn_is_unconditional(dc, insn)) { @@ -XXX,XX +XXX,XX @@ static void thumb_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu) * Insn wasn't valid for ECI/ICI at all: undo what we * just generated and instead emit an exception */ - tcg_remove_ops_after(dc->insn_eci_rewind); + tcg_remove_ops_after(insn_eci_rewind); + dc->pc_save = insn_eci_pc_save; dc->condjmp = 0; gen_exception_insn(dc, 0, EXCP_INVSTATE, syn_uncategorized()); } @@ -XXX,XX +XXX,XX @@ static void arm_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu) if (dc->condjmp) { /* "Condition failed" instruction codepath for the branch/trap insn */ - gen_set_label(dc->condlabel); + set_disas_label(dc, dc->condlabel); gen_set_condexec(dc); if (unlikely(dc->ss_active)) { gen_update_pc(dc, curr_insn_len(dc)); @@ -XXX,XX +XXX,XX @@ void restore_state_to_opc(CPUARMState *env, TranslationBlock *tb, target_ulong *data) { if (is_a64(env)) { - env->pc = data[0]; + if (TARGET_TB_PCREL) { + env->pc = (env->pc & TARGET_PAGE_MASK) | data[0]; + } else { + env->pc = data[0]; + } env->condexec_bits = 0; env->exception.syndrome = data[2] << ARM_INSN_START_WORD2_SHIFT; } else { - env->regs[15] = data[0]; + if (TARGET_TB_PCREL) { + env->regs[15] = (env->regs[15] & TARGET_PAGE_MASK) | data[0]; + } else { + env->regs[15] = data[0]; + } env->condexec_bits = data[1]; env->exception.syndrome = data[2] << ARM_INSN_START_WORD2_SHIFT; } -- 2.25.1 Currently the microdrive code uses device_legacy_reset() to reset itself, and has its reset method call reset on the IDE bus as the last thing it does. Switch to using device_cold_reset(). The only concrete microdrive device is the TYPE_DSCM1XXXX; it is not command-line pluggable, so it is used only by the old pxa2xx Arm boards 'akita', 'borzoi', 'spitz', 'terrier' and 'tosa'. You might think that this would result in the IDE bus being reset automatically, but it does not, because the IDEBus type does not set the BusClass::reset method. Instead the controller must explicitly call ide_bus_reset(). We therefore leave that call in md_reset(). Note also that because the PCMCIA card device is a direct subclass of TYPE_DEVICE and we don't model the PCMCIA controller-to-card interface as a qbus, PCMCIA cards are not on any qbus and so they don't get reset when the system is reset. The reset only happens via the dscm1xxxx_attach() and dscm1xxxx_detach() functions during machine creation. Because our aim here is merely to try to get rid of calls to the device_legacy_reset() function, we leave these other dubious reset-related issues alone. (They all stem from this code being absolutely ancient.) Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Message-id: 20221013174042.1602926-1-peter.maydell@linaro.org --- hw/ide/microdrive.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/hw/ide/microdrive.c b/hw/ide/microdrive.c index XXXXXXX..XXXXXXX 100644 --- a/hw/ide/microdrive.c +++ b/hw/ide/microdrive.c @@ -XXX,XX +XXX,XX @@ static void md_attr_write(PCMCIACardState *card, uint32_t at, uint8_t value) case 0x00: /* Configuration Option Register */ s->opt = value & 0xcf; if (value & OPT_SRESET) { - device_legacy_reset(DEVICE(s)); + device_cold_reset(DEVICE(s)); } md_interrupt_update(s); break; @@ -XXX,XX +XXX,XX @@ static void md_common_write(PCMCIACardState *card, uint32_t at, uint16_t value) case 0xe: /* Device Control */ s->ctrl = value; if (value & CTRL_SRST) { - device_legacy_reset(DEVICE(s)); + device_cold_reset(DEVICE(s)); } md_interrupt_update(s); break; @@ -XXX,XX +XXX,XX @@ static int dscm1xxxx_attach(PCMCIACardState *card) md->attr_base = pcc->cis[0x74] | (pcc->cis[0x76] << 8); md->io_base = 0x0; - device_legacy_reset(DEVICE(md)); + device_cold_reset(DEVICE(md)); md_interrupt_update(md); return 0; @@ -XXX,XX +XXX,XX @@ static int dscm1xxxx_detach(PCMCIACardState *card) { MicroDriveState *md = MICRODRIVE(card); - device_legacy_reset(DEVICE(md)); + device_cold_reset(DEVICE(md)); return 0; } -- 2.25.1