I don't have anything else queued up at the moment, so this is just Richard's SME patches. -- PMM The following changes since commit 63b38f6c85acd312c2cab68554abf33adf4ee2b3: Merge tag 'pull-target-arm-20220707' of https://git.linaro.org/people/pmaydell/qemu-arm into staging (2022-07-08 06:17:11 +0530) are available in the Git repository at: https://git.linaro.org/people/pmaydell/qemu-arm.git tags/pull-target-arm-20220711 for you to fetch changes up to f9982ceaf26df27d15547a3a7990a95019e9e3a8: linux-user/aarch64: Add SME related hwcap entries (2022-07-11 13:43:52 +0100) ---------------------------------------------------------------- target-arm: * Implement SME emulation, for both system and linux-user ---------------------------------------------------------------- Richard Henderson (45): target/arm: Handle SME in aarch64_cpu_dump_state target/arm: Add infrastructure for disas_sme target/arm: Trap non-streaming usage when Streaming SVE is active target/arm: Mark ADR as non-streaming target/arm: Mark RDFFR, WRFFR, SETFFR as non-streaming target/arm: Mark BDEP, BEXT, BGRP, COMPACT, FEXPA, FTSSEL as non-streaming target/arm: Mark PMULL, FMMLA as non-streaming target/arm: Mark FTSMUL, FTMAD, FADDA as non-streaming target/arm: Mark SMMLA, UMMLA, USMMLA as non-streaming target/arm: Mark string/histo/crypto as non-streaming target/arm: Mark gather/scatter load/store as non-streaming target/arm: Mark gather prefetch as non-streaming target/arm: Mark LDFF1 and LDNF1 as non-streaming target/arm: Mark LD1RO as non-streaming target/arm: Add SME enablement checks target/arm: Handle SME in sve_access_check target/arm: Implement SME RDSVL, ADDSVL, ADDSPL target/arm: Implement SME ZERO target/arm: Implement SME MOVA target/arm: Implement SME LD1, ST1 target/arm: Export unpredicated ld/st from translate-sve.c target/arm: Implement SME LDR, STR target/arm: Implement SME ADDHA, ADDVA target/arm: Implement FMOPA, FMOPS (non-widening) target/arm: Implement BFMOPA, BFMOPS target/arm: Implement FMOPA, FMOPS (widening) target/arm: Implement SME integer outer product target/arm: Implement PSEL target/arm: Implement REVD target/arm: Implement SCLAMP, UCLAMP target/arm: Reset streaming sve state on exception boundaries target/arm: Enable SME for -cpu max linux-user/aarch64: Clear tpidr2_el0 if CLONE_SETTLS linux-user/aarch64: Reset PSTATE.SM on syscalls linux-user/aarch64: Add SM bit to SVE signal context linux-user/aarch64: Tidy target_restore_sigframe error return linux-user/aarch64: Do not allow duplicate or short sve records linux-user/aarch64: Verify extra record lock succeeded linux-user/aarch64: Move sve record checks into restore linux-user/aarch64: Implement SME signal handling linux-user: Rename sve prctls linux-user/aarch64: Implement PR_SME_GET_VL, PR_SME_SET_VL target/arm: Only set ZEN in reset if SVE present target/arm: Enable SME for user-only linux-user/aarch64: Add SME related hwcap entries docs/system/arm/emulation.rst | 4 + linux-user/aarch64/target_cpu.h | 5 +- linux-user/aarch64/target_prctl.h | 62 +- target/arm/cpu.h | 7 + target/arm/helper-sme.h | 126 ++++ target/arm/helper-sve.h | 4 + target/arm/helper.h | 18 + target/arm/translate-a64.h | 45 ++ target/arm/translate.h | 16 + target/arm/sme-fa64.decode | 60 ++ target/arm/sme.decode | 88 +++ target/arm/sve.decode | 41 +- linux-user/aarch64/cpu_loop.c | 9 + linux-user/aarch64/signal.c | 243 ++++++-- linux-user/elfload.c | 20 + linux-user/syscall.c | 28 +- target/arm/cpu.c | 35 +- target/arm/cpu64.c | 11 + target/arm/helper.c | 56 +- target/arm/sme_helper.c | 1140 +++++++++++++++++++++++++++++++++++++ target/arm/sve_helper.c | 28 + target/arm/translate-a64.c | 103 +++- target/arm/translate-sme.c | 373 ++++++++++++ target/arm/translate-sve.c | 393 ++++++++++--- target/arm/translate-vfp.c | 12 + target/arm/translate.c | 2 + target/arm/vec_helper.c | 24 + target/arm/meson.build | 3 + 28 files changed, 2821 insertions(+), 135 deletions(-) create mode 100644 target/arm/sme-fa64.decode create mode 100644 target/arm/sme.decode create mode 100644 target/arm/translate-sme.c From: Richard Henderson <richard.henderson@linaro.org> Dump SVCR, plus use the correct access check for Streaming Mode. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-2-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/cpu.c | 17 ++++++++++++++++- 1 file changed, 16 insertions(+), 1 deletion(-) 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 aarch64_cpu_dump_state(CPUState *cs, FILE *f, int flags) int i; int el = arm_current_el(env); const char *ns_status; + bool sve; qemu_fprintf(f, " PC=%016" PRIx64 " ", env->pc); for (i = 0; i < 32; i++) { @@ -XXX,XX +XXX,XX @@ static void aarch64_cpu_dump_state(CPUState *cs, FILE *f, int flags) el, psr & PSTATE_SP ? 'h' : 't'); + if (cpu_isar_feature(aa64_sme, cpu)) { + qemu_fprintf(f, " SVCR=%08" PRIx64 " %c%c", + env->svcr, + (FIELD_EX64(env->svcr, SVCR, ZA) ? 'Z' : '-'), + (FIELD_EX64(env->svcr, SVCR, SM) ? 'S' : '-')); + } if (cpu_isar_feature(aa64_bti, cpu)) { qemu_fprintf(f, " BTYPE=%d", (psr & PSTATE_BTYPE) >> 10); } @@ -XXX,XX +XXX,XX @@ static void aarch64_cpu_dump_state(CPUState *cs, FILE *f, int flags) qemu_fprintf(f, " FPCR=%08x FPSR=%08x\n", vfp_get_fpcr(env), vfp_get_fpsr(env)); - if (cpu_isar_feature(aa64_sve, cpu) && sve_exception_el(env, el) == 0) { + if (cpu_isar_feature(aa64_sme, cpu) && FIELD_EX64(env->svcr, SVCR, SM)) { + sve = sme_exception_el(env, el) == 0; + } else if (cpu_isar_feature(aa64_sve, cpu)) { + sve = sve_exception_el(env, el) == 0; + } else { + sve = false; + } + + if (sve) { int j, zcr_len = sve_vqm1_for_el(env, el); for (i = 0; i <= FFR_PRED_NUM; i++) { -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> This includes the build rules for the decoder, and the new file for translation, but excludes any instructions. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-3-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-a64.h | 1 + target/arm/sme.decode | 20 ++++++++++++++++++++ target/arm/translate-a64.c | 7 ++++++- target/arm/translate-sme.c | 35 +++++++++++++++++++++++++++++++++++ target/arm/meson.build | 2 ++ 5 files changed, 64 insertions(+), 1 deletion(-) create mode 100644 target/arm/sme.decode create mode 100644 target/arm/translate-sme.c 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 pred_gvec_reg_size(DisasContext *s) } bool disas_sve(DisasContext *, uint32_t); +bool disas_sme(DisasContext *, uint32_t); void gen_gvec_rax1(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs, uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz); diff --git a/target/arm/sme.decode b/target/arm/sme.decode new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/target/arm/sme.decode @@ -XXX,XX +XXX,XX @@ +# AArch64 SME instruction descriptions +# +# Copyright (c) 2022 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 2.1 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 +# 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 aarch64_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu) } switch (extract32(insn, 25, 4)) { - case 0x0: case 0x1: case 0x3: /* UNALLOCATED */ + case 0x0: + if (!extract32(insn, 31, 1) || !disas_sme(s, insn)) { + unallocated_encoding(s); + } + break; + case 0x1: case 0x3: /* UNALLOCATED */ unallocated_encoding(s); break; case 0x2: diff --git a/target/arm/translate-sme.c b/target/arm/translate-sme.c new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/target/arm/translate-sme.c @@ -XXX,XX +XXX,XX @@ +/* + * AArch64 SME translation + * + * Copyright (c) 2022 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 2.1 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/>. + */ + +#include "qemu/osdep.h" +#include "cpu.h" +#include "tcg/tcg-op.h" +#include "tcg/tcg-op-gvec.h" +#include "tcg/tcg-gvec-desc.h" +#include "translate.h" +#include "exec/helper-gen.h" +#include "translate-a64.h" +#include "fpu/softfloat.h" + + +/* + * Include the generated decoder. + */ + +#include "decode-sme.c.inc" diff --git a/target/arm/meson.build b/target/arm/meson.build index XXXXXXX..XXXXXXX 100644 --- a/target/arm/meson.build +++ b/target/arm/meson.build @@ -XXX,XX +XXX,XX @@ gen = [ decodetree.process('sve.decode', extra_args: '--decode=disas_sve'), + decodetree.process('sme.decode', extra_args: '--decode=disas_sme'), decodetree.process('neon-shared.decode', extra_args: '--decode=disas_neon_shared'), decodetree.process('neon-dp.decode', extra_args: '--decode=disas_neon_dp'), decodetree.process('neon-ls.decode', extra_args: '--decode=disas_neon_ls'), @@ -XXX,XX +XXX,XX @@ arm_ss.add(when: 'TARGET_AARCH64', if_true: files( 'sme_helper.c', 'translate-a64.c', 'translate-sve.c', + 'translate-sme.c', )) arm_softmmu_ss = ss.source_set() -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> This new behaviour is in the ARM pseudocode function AArch64.CheckFPAdvSIMDEnabled, which applies to AArch32 via AArch32.CheckAdvSIMDOrFPEnabled when the EL to which the trap would be delivered is in AArch64 mode. Given that ARMv9 drops support for AArch32 outside EL0, the trap EL detection ought to be trivially true, but the pseudocode still contains a number of conditions, and QEMU has not yet committed to dropping A32 support for EL[12] when v9 features are present. Since the computation of SME_TRAP_NONSTREAMING is necessarily different for the two modes, we might as well preserve bits within TBFLAG_ANY and allocate separate bits within TBFLAG_A32 and TBFLAG_A64 instead. Note that DDI0616A.a has typos for bits [22:21] of LD1RO in the table of instructions illegal in streaming mode. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-4-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/cpu.h | 7 +++ target/arm/translate.h | 4 ++ target/arm/sme-fa64.decode | 90 ++++++++++++++++++++++++++++++++++++++ target/arm/helper.c | 41 +++++++++++++++++ target/arm/translate-a64.c | 40 ++++++++++++++++- target/arm/translate-vfp.c | 12 +++++ target/arm/translate.c | 2 + target/arm/meson.build | 1 + 8 files changed, 195 insertions(+), 2 deletions(-) create mode 100644 target/arm/sme-fa64.decode 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 @@ FIELD(TBFLAG_A32, HSTR_ACTIVE, 9, 1) * the same thing as the current security state of the processor! */ FIELD(TBFLAG_A32, NS, 10, 1) +/* + * Indicates that SME Streaming mode is active, and SMCR_ELx.FA64 is not. + * This requires an SME trap from AArch32 mode when using NEON. + */ +FIELD(TBFLAG_A32, SME_TRAP_NONSTREAMING, 11, 1) /* * Bit usage when in AArch32 state, for M-profile only. @@ -XXX,XX +XXX,XX @@ FIELD(TBFLAG_A64, SMEEXC_EL, 20, 2) FIELD(TBFLAG_A64, PSTATE_SM, 22, 1) FIELD(TBFLAG_A64, PSTATE_ZA, 23, 1) FIELD(TBFLAG_A64, SVL, 24, 4) +/* Indicates that SME Streaming mode is active, and SMCR_ELx.FA64 is not. */ +FIELD(TBFLAG_A64, SME_TRAP_NONSTREAMING, 28, 1) /* * Helpers for using the above. 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 struct DisasContext { bool pstate_sm; /* True if PSTATE.ZA is set. */ bool pstate_za; + /* True if non-streaming insns should raise an SME Streaming exception. */ + bool sme_trap_nonstreaming; + /* True if the current instruction is non-streaming. */ + bool is_nonstreaming; /* True if MVE insns are definitely not predicated by VPR or LTPSIZE */ bool mve_no_pred; /* diff --git a/target/arm/sme-fa64.decode b/target/arm/sme-fa64.decode new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/target/arm/sme-fa64.decode @@ -XXX,XX +XXX,XX @@ +# AArch64 SME allowed instruction decoding +# +# Copyright (c) 2022 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 2.1 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 +# + +# These patterns are taken from Appendix E1.1 of DDI0616 A.a, +# Arm Architecture Reference Manual Supplement, +# The Scalable Matrix Extension (SME), for Armv9-A + +{ + [ + OK 0-00 1110 0000 0001 0010 11-- ---- ---- # SMOV W|Xd,Vn.B[0] + OK 0-00 1110 0000 0010 0010 11-- ---- ---- # SMOV W|Xd,Vn.H[0] + OK 0100 1110 0000 0100 0010 11-- ---- ---- # SMOV Xd,Vn.S[0] + OK 0000 1110 0000 0001 0011 11-- ---- ---- # UMOV Wd,Vn.B[0] + OK 0000 1110 0000 0010 0011 11-- ---- ---- # UMOV Wd,Vn.H[0] + OK 0000 1110 0000 0100 0011 11-- ---- ---- # UMOV Wd,Vn.S[0] + OK 0100 1110 0000 1000 0011 11-- ---- ---- # UMOV Xd,Vn.D[0] + ] + FAIL 0--0 111- ---- ---- ---- ---- ---- ---- # Advanced SIMD vector operations +} + +{ + [ + OK 0101 1110 --1- ---- 11-1 11-- ---- ---- # FMULX/FRECPS/FRSQRTS (scalar) + OK 0101 1110 -10- ---- 00-1 11-- ---- ---- # FMULX/FRECPS/FRSQRTS (scalar, FP16) + OK 01-1 1110 1-10 0001 11-1 10-- ---- ---- # FRECPE/FRSQRTE/FRECPX (scalar) + OK 01-1 1110 1111 1001 11-1 10-- ---- ---- # FRECPE/FRSQRTE/FRECPX (scalar, FP16) + ] + FAIL 01-1 111- ---- ---- ---- ---- ---- ---- # Advanced SIMD single-element operations +} + +FAIL 0-00 110- ---- ---- ---- ---- ---- ---- # Advanced SIMD structure load/store +FAIL 1100 1110 ---- ---- ---- ---- ---- ---- # Advanced SIMD cryptography extensions +FAIL 0001 1110 0111 1110 0000 00-- ---- ---- # FJCVTZS + +# These are the "avoidance of doubt" final table of Illegal Advanced SIMD instructions +# We don't actually need to include these, as the default is OK. +# -001 111- ---- ---- ---- ---- ---- ---- # Scalar floating-point operations +# --10 110- ---- ---- ---- ---- ---- ---- # Load/store pair of FP registers +# --01 1100 ---- ---- ---- ---- ---- ---- # Load FP register (PC-relative literal) +# --11 1100 --0- ---- ---- ---- ---- ---- # Load/store FP register (unscaled imm) +# --11 1100 --1- ---- ---- ---- ---- --10 # Load/store FP register (register offset) +# --11 1101 ---- ---- ---- ---- ---- ---- # Load/store FP register (scaled imm) + +FAIL 0000 0100 --1- ---- 1010 ---- ---- ---- # ADR +FAIL 0000 0100 --1- ---- 1011 -0-- ---- ---- # FTSSEL, FEXPA +FAIL 0000 0101 --10 0001 100- ---- ---- ---- # COMPACT +FAIL 0010 0101 --01 100- 1111 000- ---0 ---- # RDFFR, RDFFRS +FAIL 0010 0101 --10 1--- 1001 ---- ---- ---- # WRFFR, SETFFR +FAIL 0100 0101 --0- ---- 1011 ---- ---- ---- # BDEP, BEXT, BGRP +FAIL 0100 0101 000- ---- 0110 1--- ---- ---- # PMULLB, PMULLT (128b result) +FAIL 0110 0100 --1- ---- 1110 01-- ---- ---- # FMMLA, BFMMLA +FAIL 0110 0101 --0- ---- 0000 11-- ---- ---- # FTSMUL +FAIL 0110 0101 --01 0--- 100- ---- ---- ---- # FTMAD +FAIL 0110 0101 --01 1--- 001- ---- ---- ---- # FADDA +FAIL 0100 0101 --0- ---- 1001 10-- ---- ---- # SMMLA, UMMLA, USMMLA +FAIL 0100 0101 --1- ---- 1--- ---- ---- ---- # SVE2 string/histo/crypto instructions +FAIL 1000 010- -00- ---- 10-- ---- ---- ---- # SVE2 32-bit gather NT load (vector+scalar) +FAIL 1000 010- -00- ---- 111- ---- ---- ---- # SVE 32-bit gather prefetch (vector+imm) +FAIL 1000 0100 0-1- ---- 0--- ---- ---- ---- # SVE 32-bit gather prefetch (scalar+vector) +FAIL 1000 010- -01- ---- 1--- ---- ---- ---- # SVE 32-bit gather load (vector+imm) +FAIL 1000 0100 0-0- ---- 0--- ---- ---- ---- # SVE 32-bit gather load byte (scalar+vector) +FAIL 1000 0100 1--- ---- 0--- ---- ---- ---- # SVE 32-bit gather load half (scalar+vector) +FAIL 1000 0101 0--- ---- 0--- ---- ---- ---- # SVE 32-bit gather load word (scalar+vector) +FAIL 1010 010- ---- ---- 011- ---- ---- ---- # SVE contiguous FF load (scalar+scalar) +FAIL 1010 010- ---1 ---- 101- ---- ---- ---- # SVE contiguous NF load (scalar+imm) +FAIL 1010 010- -01- ---- 000- ---- ---- ---- # SVE load & replicate 32 bytes (scalar+scalar) +FAIL 1010 010- -010 ---- 001- ---- ---- ---- # SVE load & replicate 32 bytes (scalar+imm) +FAIL 1100 010- ---- ---- ---- ---- ---- ---- # SVE 64-bit gather load/prefetch +FAIL 1110 010- -00- ---- 001- ---- ---- ---- # SVE2 64-bit scatter NT store (vector+scalar) +FAIL 1110 010- -10- ---- 001- ---- ---- ---- # SVE2 32-bit scatter NT store (vector+scalar) +FAIL 1110 010- ---- ---- 1-0- ---- ---- ---- # SVE scatter store (scalar+32-bit vector) +FAIL 1110 010- ---- ---- 101- ---- ---- ---- # SVE scatter store (misc) 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 @@ int sme_exception_el(CPUARMState *env, int el) return 0; } +/* This corresponds to the ARM pseudocode function IsFullA64Enabled(). */ +static bool sme_fa64(CPUARMState *env, int el) +{ + if (!cpu_isar_feature(aa64_sme_fa64, env_archcpu(env))) { + return false; + } + + if (el <= 1 && !el_is_in_host(env, el)) { + if (!FIELD_EX64(env->vfp.smcr_el[1], SMCR, FA64)) { + return false; + } + } + if (el <= 2 && arm_is_el2_enabled(env)) { + if (!FIELD_EX64(env->vfp.smcr_el[2], SMCR, FA64)) { + return false; + } + } + if (arm_feature(env, ARM_FEATURE_EL3)) { + if (!FIELD_EX64(env->vfp.smcr_el[3], SMCR, FA64)) { + return false; + } + } + + return true; +} + /* * Given that SVE is enabled, return the vector length for EL. */ @@ -XXX,XX +XXX,XX @@ static CPUARMTBFlags rebuild_hflags_a32(CPUARMState *env, int fp_el, DP_TBFLAG_ANY(flags, PSTATE__IL, 1); } + /* + * The SME exception we are testing for is raised via + * AArch64.CheckFPAdvSIMDEnabled(), as called from + * AArch32.CheckAdvSIMDOrFPEnabled(). + */ + if (el == 0 + && FIELD_EX64(env->svcr, SVCR, SM) + && (!arm_is_el2_enabled(env) + || (arm_el_is_aa64(env, 2) && !(env->cp15.hcr_el2 & HCR_TGE))) + && arm_el_is_aa64(env, 1) + && !sme_fa64(env, el)) { + DP_TBFLAG_A32(flags, SME_TRAP_NONSTREAMING, 1); + } + return rebuild_hflags_common_32(env, fp_el, mmu_idx, flags); } @@ -XXX,XX +XXX,XX @@ static CPUARMTBFlags rebuild_hflags_a64(CPUARMState *env, int el, int fp_el, } if (FIELD_EX64(env->svcr, SVCR, SM)) { DP_TBFLAG_A64(flags, PSTATE_SM, 1); + DP_TBFLAG_A64(flags, SME_TRAP_NONSTREAMING, !sme_fa64(env, el)); } DP_TBFLAG_A64(flags, PSTATE_ZA, FIELD_EX64(env->svcr, SVCR, ZA)); } 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 do_vec_ld(DisasContext *s, int destidx, int element, * unallocated-encoding checks (otherwise the syndrome information * for the resulting exception will be incorrect). */ -static bool fp_access_check(DisasContext *s) +static bool fp_access_check_only(DisasContext *s) { if (s->fp_excp_el) { assert(!s->fp_access_checked); @@ -XXX,XX +XXX,XX @@ static bool fp_access_check(DisasContext *s) return true; } +static bool fp_access_check(DisasContext *s) +{ + if (!fp_access_check_only(s)) { + return false; + } + if (s->sme_trap_nonstreaming && s->is_nonstreaming) { + gen_exception_insn(s, s->pc_curr, EXCP_UDEF, + syn_smetrap(SME_ET_Streaming, false)); + return false; + } + return true; +} + /* Check that SVE access is enabled. If it is, return true. * If not, emit code to generate an appropriate exception and return false. */ @@ -XXX,XX +XXX,XX @@ static void handle_sys(DisasContext *s, uint32_t insn, bool isread, default: g_assert_not_reached(); } - if ((ri->type & ARM_CP_FPU) && !fp_access_check(s)) { + if ((ri->type & ARM_CP_FPU) && !fp_access_check_only(s)) { return; } else if ((ri->type & ARM_CP_SVE) && !sve_access_check(s)) { return; @@ -XXX,XX +XXX,XX @@ static void disas_data_proc_simd_fp(DisasContext *s, uint32_t insn) } } +/* + * Include the generated SME FA64 decoder. + */ + +#include "decode-sme-fa64.c.inc" + +static bool trans_OK(DisasContext *s, arg_OK *a) +{ + return true; +} + +static bool trans_FAIL(DisasContext *s, arg_OK *a) +{ + s->is_nonstreaming = true; + return true; +} + /** * is_guarded_page: * @env: The cpu environment @@ -XXX,XX +XXX,XX @@ static void aarch64_tr_init_disas_context(DisasContextBase *dcbase, dc->mte_active[1] = EX_TBFLAG_A64(tb_flags, MTE0_ACTIVE); dc->pstate_sm = EX_TBFLAG_A64(tb_flags, PSTATE_SM); dc->pstate_za = EX_TBFLAG_A64(tb_flags, PSTATE_ZA); + dc->sme_trap_nonstreaming = EX_TBFLAG_A64(tb_flags, SME_TRAP_NONSTREAMING); dc->vec_len = 0; dc->vec_stride = 0; dc->cp_regs = arm_cpu->cp_regs; @@ -XXX,XX +XXX,XX @@ static void aarch64_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu) } } + s->is_nonstreaming = false; + if (s->sme_trap_nonstreaming) { + disas_sme_fa64(s, insn); + } + switch (extract32(insn, 25, 4)) { case 0x0: if (!extract32(insn, 31, 1) || !disas_sme(s, insn)) { 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) return false; } + /* + * Note that rebuild_hflags_a32 has already accounted for being in EL0 + * and the higher EL in A64 mode, etc. Unlike A64 mode, there do not + * 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, + syn_smetrap(SME_ET_Streaming, + s->base.pc_next - s->pc_curr == 2)); + return false; + } + if (!s->vfp_enabled && !ignore_vfp_enabled) { assert(!arm_dc_feature(s, ARM_FEATURE_M)); 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 void arm_tr_init_disas_context(DisasContextBase *dcbase, CPUState *cs) dc->vec_len = EX_TBFLAG_A32(tb_flags, VECLEN); dc->vec_stride = EX_TBFLAG_A32(tb_flags, VECSTRIDE); } + dc->sme_trap_nonstreaming = + EX_TBFLAG_A32(tb_flags, SME_TRAP_NONSTREAMING); } dc->cp_regs = cpu->cp_regs; dc->features = env->features; diff --git a/target/arm/meson.build b/target/arm/meson.build index XXXXXXX..XXXXXXX 100644 --- a/target/arm/meson.build +++ b/target/arm/meson.build @@ -XXX,XX +XXX,XX @@ gen = [ decodetree.process('sve.decode', extra_args: '--decode=disas_sve'), decodetree.process('sme.decode', extra_args: '--decode=disas_sme'), + decodetree.process('sme-fa64.decode', extra_args: '--static-decode=disas_sme_fa64'), decodetree.process('neon-shared.decode', extra_args: '--decode=disas_neon_shared'), decodetree.process('neon-dp.decode', extra_args: '--decode=disas_neon_dp'), decodetree.process('neon-ls.decode', extra_args: '--decode=disas_neon_ls'), -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Mark ADR as a non-streaming instruction, which should trap if full a64 support is not enabled in streaming mode. Removing entries from sme-fa64.decode is an easy way to see what remains to be done. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-5-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate.h | 7 +++++++ target/arm/sme-fa64.decode | 1 - target/arm/translate-sve.c | 8 ++++---- 3 files changed, 11 insertions(+), 5 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 @@ uint64_t asimd_imm_const(uint32_t imm, int cmode, int op); static bool trans_##NAME(DisasContext *s, arg_##NAME *a) \ { return dc_isar_feature(FEAT, s) && FUNC(s, __VA_ARGS__); } +#define TRANS_FEAT_NONSTREAMING(NAME, FEAT, FUNC, ...) \ + static bool trans_##NAME(DisasContext *s, arg_##NAME *a) \ + { \ + s->is_nonstreaming = true; \ + return dc_isar_feature(FEAT, s) && FUNC(s, __VA_ARGS__); \ + } + #endif /* TARGET_ARM_TRANSLATE_H */ diff --git a/target/arm/sme-fa64.decode b/target/arm/sme-fa64.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme-fa64.decode +++ b/target/arm/sme-fa64.decode @@ -XXX,XX +XXX,XX @@ FAIL 0001 1110 0111 1110 0000 00-- ---- ---- # FJCVTZS # --11 1100 --1- ---- ---- ---- ---- --10 # Load/store FP register (register offset) # --11 1101 ---- ---- ---- ---- ---- ---- # Load/store FP register (scaled imm) -FAIL 0000 0100 --1- ---- 1010 ---- ---- ---- # ADR FAIL 0000 0100 --1- ---- 1011 -0-- ---- ---- # FTSSEL, FEXPA FAIL 0000 0101 --10 0001 100- ---- ---- ---- # COMPACT FAIL 0010 0101 --01 100- 1111 000- ---0 ---- # RDFFR, RDFFRS diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool do_adr(DisasContext *s, arg_rrri *a, gen_helper_gvec_3 *fn) return gen_gvec_ool_zzz(s, fn, a->rd, a->rn, a->rm, a->imm); } -TRANS_FEAT(ADR_p32, aa64_sve, do_adr, a, gen_helper_sve_adr_p32) -TRANS_FEAT(ADR_p64, aa64_sve, do_adr, a, gen_helper_sve_adr_p64) -TRANS_FEAT(ADR_s32, aa64_sve, do_adr, a, gen_helper_sve_adr_s32) -TRANS_FEAT(ADR_u32, aa64_sve, do_adr, a, gen_helper_sve_adr_u32) +TRANS_FEAT_NONSTREAMING(ADR_p32, aa64_sve, do_adr, a, gen_helper_sve_adr_p32) +TRANS_FEAT_NONSTREAMING(ADR_p64, aa64_sve, do_adr, a, gen_helper_sve_adr_p64) +TRANS_FEAT_NONSTREAMING(ADR_s32, aa64_sve, do_adr, a, gen_helper_sve_adr_s32) +TRANS_FEAT_NONSTREAMING(ADR_u32, aa64_sve, do_adr, a, gen_helper_sve_adr_u32) /* *** SVE Integer Misc - Unpredicated Group -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Mark these as a non-streaming instructions, which should trap if full a64 support is not enabled in streaming mode. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-6-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/sme-fa64.decode | 2 -- target/arm/translate-sve.c | 9 ++++++--- 2 files changed, 6 insertions(+), 5 deletions(-) diff --git a/target/arm/sme-fa64.decode b/target/arm/sme-fa64.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme-fa64.decode +++ b/target/arm/sme-fa64.decode @@ -XXX,XX +XXX,XX @@ FAIL 0001 1110 0111 1110 0000 00-- ---- ---- # FJCVTZS FAIL 0000 0100 --1- ---- 1011 -0-- ---- ---- # FTSSEL, FEXPA FAIL 0000 0101 --10 0001 100- ---- ---- ---- # COMPACT -FAIL 0010 0101 --01 100- 1111 000- ---0 ---- # RDFFR, RDFFRS -FAIL 0010 0101 --10 1--- 1001 ---- ---- ---- # WRFFR, SETFFR FAIL 0100 0101 --0- ---- 1011 ---- ---- ---- # BDEP, BEXT, BGRP FAIL 0100 0101 000- ---- 0110 1--- ---- ---- # PMULLB, PMULLT (128b result) FAIL 0110 0100 --1- ---- 1110 01-- ---- ---- # FMMLA, BFMMLA diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool do_predset(DisasContext *s, int esz, int rd, int pat, bool setflag) TRANS_FEAT(PTRUE, aa64_sve, do_predset, a->esz, a->rd, a->pat, a->s) /* Note pat == 31 is #all, to set all elements. */ -TRANS_FEAT(SETFFR, aa64_sve, do_predset, 0, FFR_PRED_NUM, 31, false) +TRANS_FEAT_NONSTREAMING(SETFFR, aa64_sve, + do_predset, 0, FFR_PRED_NUM, 31, false) /* Note pat == 32 is #unimp, to set no elements. */ TRANS_FEAT(PFALSE, aa64_sve, do_predset, 0, a->rd, 32, false) @@ -XXX,XX +XXX,XX @@ static bool trans_RDFFR_p(DisasContext *s, arg_RDFFR_p *a) .rd = a->rd, .pg = a->pg, .s = a->s, .rn = FFR_PRED_NUM, .rm = FFR_PRED_NUM, }; + + s->is_nonstreaming = true; return trans_AND_pppp(s, &alt_a); } -TRANS_FEAT(RDFFR, aa64_sve, do_mov_p, a->rd, FFR_PRED_NUM) -TRANS_FEAT(WRFFR, aa64_sve, do_mov_p, FFR_PRED_NUM, a->rn) +TRANS_FEAT_NONSTREAMING(RDFFR, aa64_sve, do_mov_p, a->rd, FFR_PRED_NUM) +TRANS_FEAT_NONSTREAMING(WRFFR, aa64_sve, do_mov_p, FFR_PRED_NUM, a->rn) static bool do_pfirst_pnext(DisasContext *s, arg_rr_esz *a, void (*gen_fn)(TCGv_i32, TCGv_ptr, -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Mark these as a non-streaming instructions, which should trap if full a64 support is not enabled in streaming mode. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-7-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/sme-fa64.decode | 3 --- target/arm/translate-sve.c | 22 ++++++++++++---------- 2 files changed, 12 insertions(+), 13 deletions(-) diff --git a/target/arm/sme-fa64.decode b/target/arm/sme-fa64.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme-fa64.decode +++ b/target/arm/sme-fa64.decode @@ -XXX,XX +XXX,XX @@ FAIL 0001 1110 0111 1110 0000 00-- ---- ---- # FJCVTZS # --11 1100 --1- ---- ---- ---- ---- --10 # Load/store FP register (register offset) # --11 1101 ---- ---- ---- ---- ---- ---- # Load/store FP register (scaled imm) -FAIL 0000 0100 --1- ---- 1011 -0-- ---- ---- # FTSSEL, FEXPA -FAIL 0000 0101 --10 0001 100- ---- ---- ---- # COMPACT -FAIL 0100 0101 --0- ---- 1011 ---- ---- ---- # BDEP, BEXT, BGRP FAIL 0100 0101 000- ---- 0110 1--- ---- ---- # PMULLB, PMULLT (128b result) FAIL 0110 0100 --1- ---- 1110 01-- ---- ---- # FMMLA, BFMMLA FAIL 0110 0101 --0- ---- 0000 11-- ---- ---- # FTSMUL diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static gen_helper_gvec_2 * const fexpa_fns[4] = { NULL, gen_helper_sve_fexpa_h, gen_helper_sve_fexpa_s, gen_helper_sve_fexpa_d, }; -TRANS_FEAT(FEXPA, aa64_sve, gen_gvec_ool_zz, - fexpa_fns[a->esz], a->rd, a->rn, 0) +TRANS_FEAT_NONSTREAMING(FEXPA, aa64_sve, gen_gvec_ool_zz, + fexpa_fns[a->esz], a->rd, a->rn, 0) static gen_helper_gvec_3 * const ftssel_fns[4] = { NULL, gen_helper_sve_ftssel_h, gen_helper_sve_ftssel_s, gen_helper_sve_ftssel_d, }; -TRANS_FEAT(FTSSEL, aa64_sve, gen_gvec_ool_arg_zzz, ftssel_fns[a->esz], a, 0) +TRANS_FEAT_NONSTREAMING(FTSSEL, aa64_sve, gen_gvec_ool_arg_zzz, + ftssel_fns[a->esz], a, 0) /* *** SVE Predicate Logical Operations Group @@ -XXX,XX +XXX,XX @@ TRANS_FEAT(TRN2_q, aa64_sve_f64mm, gen_gvec_ool_arg_zzz, static gen_helper_gvec_3 * const compact_fns[4] = { NULL, NULL, gen_helper_sve_compact_s, gen_helper_sve_compact_d }; -TRANS_FEAT(COMPACT, aa64_sve, gen_gvec_ool_arg_zpz, compact_fns[a->esz], a, 0) +TRANS_FEAT_NONSTREAMING(COMPACT, aa64_sve, gen_gvec_ool_arg_zpz, + compact_fns[a->esz], a, 0) /* Call the helper that computes the ARM LastActiveElement pseudocode * function, scaled by the element size. This includes the not found @@ -XXX,XX +XXX,XX @@ static gen_helper_gvec_3 * const bext_fns[4] = { gen_helper_sve2_bext_b, gen_helper_sve2_bext_h, gen_helper_sve2_bext_s, gen_helper_sve2_bext_d, }; -TRANS_FEAT(BEXT, aa64_sve2_bitperm, gen_gvec_ool_arg_zzz, - bext_fns[a->esz], a, 0) +TRANS_FEAT_NONSTREAMING(BEXT, aa64_sve2_bitperm, gen_gvec_ool_arg_zzz, + bext_fns[a->esz], a, 0) static gen_helper_gvec_3 * const bdep_fns[4] = { gen_helper_sve2_bdep_b, gen_helper_sve2_bdep_h, gen_helper_sve2_bdep_s, gen_helper_sve2_bdep_d, }; -TRANS_FEAT(BDEP, aa64_sve2_bitperm, gen_gvec_ool_arg_zzz, - bdep_fns[a->esz], a, 0) +TRANS_FEAT_NONSTREAMING(BDEP, aa64_sve2_bitperm, gen_gvec_ool_arg_zzz, + bdep_fns[a->esz], a, 0) static gen_helper_gvec_3 * const bgrp_fns[4] = { gen_helper_sve2_bgrp_b, gen_helper_sve2_bgrp_h, gen_helper_sve2_bgrp_s, gen_helper_sve2_bgrp_d, }; -TRANS_FEAT(BGRP, aa64_sve2_bitperm, gen_gvec_ool_arg_zzz, - bgrp_fns[a->esz], a, 0) +TRANS_FEAT_NONSTREAMING(BGRP, aa64_sve2_bitperm, gen_gvec_ool_arg_zzz, + bgrp_fns[a->esz], a, 0) static gen_helper_gvec_3 * const cadd_fns[4] = { gen_helper_sve2_cadd_b, gen_helper_sve2_cadd_h, -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Mark these as a non-streaming instructions, which should trap if full a64 support is not enabled in streaming mode. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-8-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/sme-fa64.decode | 2 -- target/arm/translate-sve.c | 24 +++++++++++++++--------- 2 files changed, 15 insertions(+), 11 deletions(-) diff --git a/target/arm/sme-fa64.decode b/target/arm/sme-fa64.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme-fa64.decode +++ b/target/arm/sme-fa64.decode @@ -XXX,XX +XXX,XX @@ FAIL 0001 1110 0111 1110 0000 00-- ---- ---- # FJCVTZS # --11 1100 --1- ---- ---- ---- ---- --10 # Load/store FP register (register offset) # --11 1101 ---- ---- ---- ---- ---- ---- # Load/store FP register (scaled imm) -FAIL 0100 0101 000- ---- 0110 1--- ---- ---- # PMULLB, PMULLT (128b result) -FAIL 0110 0100 --1- ---- 1110 01-- ---- ---- # FMMLA, BFMMLA FAIL 0110 0101 --0- ---- 0000 11-- ---- ---- # FTSMUL FAIL 0110 0101 --01 0--- 100- ---- ---- ---- # FTMAD FAIL 0110 0101 --01 1--- 001- ---- ---- ---- # FADDA diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool do_trans_pmull(DisasContext *s, arg_rrr_esz *a, bool sel) gen_helper_gvec_pmull_q, gen_helper_sve2_pmull_h, NULL, gen_helper_sve2_pmull_d, }; - if (a->esz == 0 - ? !dc_isar_feature(aa64_sve2_pmull128, s) - : !dc_isar_feature(aa64_sve, s)) { + + if (a->esz == 0) { + if (!dc_isar_feature(aa64_sve2_pmull128, s)) { + return false; + } + s->is_nonstreaming = true; + } else if (!dc_isar_feature(aa64_sve, s)) { return false; } return gen_gvec_ool_arg_zzz(s, fns[a->esz], a, sel); @@ -XXX,XX +XXX,XX @@ DO_ZPZZ_FP(FMINP, aa64_sve2, sve2_fminp_zpzz) * SVE Integer Multiply-Add (unpredicated) */ -TRANS_FEAT(FMMLA_s, aa64_sve_f32mm, gen_gvec_fpst_zzzz, gen_helper_fmmla_s, - a->rd, a->rn, a->rm, a->ra, 0, FPST_FPCR) -TRANS_FEAT(FMMLA_d, aa64_sve_f64mm, gen_gvec_fpst_zzzz, gen_helper_fmmla_d, - a->rd, a->rn, a->rm, a->ra, 0, FPST_FPCR) +TRANS_FEAT_NONSTREAMING(FMMLA_s, aa64_sve_f32mm, gen_gvec_fpst_zzzz, + gen_helper_fmmla_s, a->rd, a->rn, a->rm, a->ra, + 0, FPST_FPCR) +TRANS_FEAT_NONSTREAMING(FMMLA_d, aa64_sve_f64mm, gen_gvec_fpst_zzzz, + gen_helper_fmmla_d, a->rd, a->rn, a->rm, a->ra, + 0, FPST_FPCR) static gen_helper_gvec_4 * const sqdmlal_zzzw_fns[] = { NULL, gen_helper_sve2_sqdmlal_zzzw_h, @@ -XXX,XX +XXX,XX @@ TRANS_FEAT(BFDOT_zzzz, aa64_sve_bf16, gen_gvec_ool_arg_zzzz, TRANS_FEAT(BFDOT_zzxz, aa64_sve_bf16, gen_gvec_ool_arg_zzxz, gen_helper_gvec_bfdot_idx, a) -TRANS_FEAT(BFMMLA, aa64_sve_bf16, gen_gvec_ool_arg_zzzz, - gen_helper_gvec_bfmmla, a, 0) +TRANS_FEAT_NONSTREAMING(BFMMLA, aa64_sve_bf16, gen_gvec_ool_arg_zzzz, + gen_helper_gvec_bfmmla, a, 0) static bool do_BFMLAL_zzzw(DisasContext *s, arg_rrrr_esz *a, bool sel) { -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Mark these as a non-streaming instructions, which should trap if full a64 support is not enabled in streaming mode. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-9-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/sme-fa64.decode | 3 --- target/arm/translate-sve.c | 15 +++++++++++---- 2 files changed, 11 insertions(+), 7 deletions(-) diff --git a/target/arm/sme-fa64.decode b/target/arm/sme-fa64.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme-fa64.decode +++ b/target/arm/sme-fa64.decode @@ -XXX,XX +XXX,XX @@ FAIL 0001 1110 0111 1110 0000 00-- ---- ---- # FJCVTZS # --11 1100 --1- ---- ---- ---- ---- --10 # Load/store FP register (register offset) # --11 1101 ---- ---- ---- ---- ---- ---- # Load/store FP register (scaled imm) -FAIL 0110 0101 --0- ---- 0000 11-- ---- ---- # FTSMUL -FAIL 0110 0101 --01 0--- 100- ---- ---- ---- # FTMAD -FAIL 0110 0101 --01 1--- 001- ---- ---- ---- # FADDA FAIL 0100 0101 --0- ---- 1001 10-- ---- ---- # SMMLA, UMMLA, USMMLA FAIL 0100 0101 --1- ---- 1--- ---- ---- ---- # SVE2 string/histo/crypto instructions FAIL 1000 010- -00- ---- 10-- ---- ---- ---- # SVE2 32-bit gather NT load (vector+scalar) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static gen_helper_gvec_3_ptr * const ftmad_fns[4] = { NULL, gen_helper_sve_ftmad_h, gen_helper_sve_ftmad_s, gen_helper_sve_ftmad_d, }; -TRANS_FEAT(FTMAD, aa64_sve, gen_gvec_fpst_zzz, - ftmad_fns[a->esz], a->rd, a->rn, a->rm, a->imm, - a->esz == MO_16 ? FPST_FPCR_F16 : FPST_FPCR) +TRANS_FEAT_NONSTREAMING(FTMAD, aa64_sve, gen_gvec_fpst_zzz, + ftmad_fns[a->esz], a->rd, a->rn, a->rm, a->imm, + a->esz == MO_16 ? FPST_FPCR_F16 : FPST_FPCR) /* *** SVE Floating Point Accumulating Reduction Group @@ -XXX,XX +XXX,XX @@ static bool trans_FADDA(DisasContext *s, arg_rprr_esz *a) if (a->esz == 0 || !dc_isar_feature(aa64_sve, s)) { return false; } + s->is_nonstreaming = true; if (!sve_access_check(s)) { return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_FADDA(DisasContext *s, arg_rprr_esz *a) DO_FP3(FADD_zzz, fadd) DO_FP3(FSUB_zzz, fsub) DO_FP3(FMUL_zzz, fmul) -DO_FP3(FTSMUL, ftsmul) DO_FP3(FRECPS, recps) DO_FP3(FRSQRTS, rsqrts) #undef DO_FP3 +static gen_helper_gvec_3_ptr * const ftsmul_fns[4] = { + NULL, gen_helper_gvec_ftsmul_h, + gen_helper_gvec_ftsmul_s, gen_helper_gvec_ftsmul_d +}; +TRANS_FEAT_NONSTREAMING(FTSMUL, aa64_sve, gen_gvec_fpst_arg_zzz, + ftsmul_fns[a->esz], a, 0) + /* *** SVE Floating Point Arithmetic - Predicated Group */ -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Mark these as a non-streaming instructions, which should trap if full a64 support is not enabled in streaming mode. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-10-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/sme-fa64.decode | 1 - target/arm/translate-sve.c | 12 ++++++------ 2 files changed, 6 insertions(+), 7 deletions(-) diff --git a/target/arm/sme-fa64.decode b/target/arm/sme-fa64.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme-fa64.decode +++ b/target/arm/sme-fa64.decode @@ -XXX,XX +XXX,XX @@ FAIL 0001 1110 0111 1110 0000 00-- ---- ---- # FJCVTZS # --11 1100 --1- ---- ---- ---- ---- --10 # Load/store FP register (register offset) # --11 1101 ---- ---- ---- ---- ---- ---- # Load/store FP register (scaled imm) -FAIL 0100 0101 --0- ---- 1001 10-- ---- ---- # SMMLA, UMMLA, USMMLA FAIL 0100 0101 --1- ---- 1--- ---- ---- ---- # SVE2 string/histo/crypto instructions FAIL 1000 010- -00- ---- 10-- ---- ---- ---- # SVE2 32-bit gather NT load (vector+scalar) FAIL 1000 010- -00- ---- 111- ---- ---- ---- # SVE 32-bit gather prefetch (vector+imm) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ TRANS_FEAT(FMLALT_zzxw, aa64_sve2, do_FMLAL_zzxw, a, false, true) TRANS_FEAT(FMLSLB_zzxw, aa64_sve2, do_FMLAL_zzxw, a, true, false) TRANS_FEAT(FMLSLT_zzxw, aa64_sve2, do_FMLAL_zzxw, a, true, true) -TRANS_FEAT(SMMLA, aa64_sve_i8mm, gen_gvec_ool_arg_zzzz, - gen_helper_gvec_smmla_b, a, 0) -TRANS_FEAT(USMMLA, aa64_sve_i8mm, gen_gvec_ool_arg_zzzz, - gen_helper_gvec_usmmla_b, a, 0) -TRANS_FEAT(UMMLA, aa64_sve_i8mm, gen_gvec_ool_arg_zzzz, - gen_helper_gvec_ummla_b, a, 0) +TRANS_FEAT_NONSTREAMING(SMMLA, aa64_sve_i8mm, gen_gvec_ool_arg_zzzz, + gen_helper_gvec_smmla_b, a, 0) +TRANS_FEAT_NONSTREAMING(USMMLA, aa64_sve_i8mm, gen_gvec_ool_arg_zzzz, + gen_helper_gvec_usmmla_b, a, 0) +TRANS_FEAT_NONSTREAMING(UMMLA, aa64_sve_i8mm, gen_gvec_ool_arg_zzzz, + gen_helper_gvec_ummla_b, a, 0) TRANS_FEAT(BFDOT_zzzz, aa64_sve_bf16, gen_gvec_ool_arg_zzzz, gen_helper_gvec_bfdot, a, 0) -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Mark these as non-streaming instructions, which should trap if full a64 support is not enabled in streaming mode. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-11-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/sme-fa64.decode | 1 - target/arm/translate-sve.c | 35 ++++++++++++++++++----------------- 2 files changed, 18 insertions(+), 18 deletions(-) diff --git a/target/arm/sme-fa64.decode b/target/arm/sme-fa64.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme-fa64.decode +++ b/target/arm/sme-fa64.decode @@ -XXX,XX +XXX,XX @@ FAIL 0001 1110 0111 1110 0000 00-- ---- ---- # FJCVTZS # --11 1100 --1- ---- ---- ---- ---- --10 # Load/store FP register (register offset) # --11 1101 ---- ---- ---- ---- ---- ---- # Load/store FP register (scaled imm) -FAIL 0100 0101 --1- ---- 1--- ---- ---- ---- # SVE2 string/histo/crypto instructions FAIL 1000 010- -00- ---- 10-- ---- ---- ---- # SVE2 32-bit gather NT load (vector+scalar) FAIL 1000 010- -00- ---- 111- ---- ---- ---- # SVE 32-bit gather prefetch (vector+imm) FAIL 1000 0100 0-1- ---- 0--- ---- ---- ---- # SVE 32-bit gather prefetch (scalar+vector) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ DO_SVE2_ZZZ_NARROW(RSUBHNT, rsubhnt) static gen_helper_gvec_flags_4 * const match_fns[4] = { gen_helper_sve2_match_ppzz_b, gen_helper_sve2_match_ppzz_h, NULL, NULL }; -TRANS_FEAT(MATCH, aa64_sve2, do_ppzz_flags, a, match_fns[a->esz]) +TRANS_FEAT_NONSTREAMING(MATCH, aa64_sve2, do_ppzz_flags, a, match_fns[a->esz]) static gen_helper_gvec_flags_4 * const nmatch_fns[4] = { gen_helper_sve2_nmatch_ppzz_b, gen_helper_sve2_nmatch_ppzz_h, NULL, NULL }; -TRANS_FEAT(NMATCH, aa64_sve2, do_ppzz_flags, a, nmatch_fns[a->esz]) +TRANS_FEAT_NONSTREAMING(NMATCH, aa64_sve2, do_ppzz_flags, a, nmatch_fns[a->esz]) static gen_helper_gvec_4 * const histcnt_fns[4] = { NULL, NULL, gen_helper_sve2_histcnt_s, gen_helper_sve2_histcnt_d }; -TRANS_FEAT(HISTCNT, aa64_sve2, gen_gvec_ool_arg_zpzz, - histcnt_fns[a->esz], a, 0) +TRANS_FEAT_NONSTREAMING(HISTCNT, aa64_sve2, gen_gvec_ool_arg_zpzz, + histcnt_fns[a->esz], a, 0) -TRANS_FEAT(HISTSEG, aa64_sve2, gen_gvec_ool_arg_zzz, - a->esz == 0 ? gen_helper_sve2_histseg : NULL, a, 0) +TRANS_FEAT_NONSTREAMING(HISTSEG, aa64_sve2, gen_gvec_ool_arg_zzz, + a->esz == 0 ? gen_helper_sve2_histseg : NULL, a, 0) DO_ZPZZ_FP(FADDP, aa64_sve2, sve2_faddp_zpzz) DO_ZPZZ_FP(FMAXNMP, aa64_sve2, sve2_fmaxnmp_zpzz) @@ -XXX,XX +XXX,XX @@ TRANS_FEAT(SQRDCMLAH_zzzz, aa64_sve2, gen_gvec_ool_zzzz, TRANS_FEAT(USDOT_zzzz, aa64_sve_i8mm, gen_gvec_ool_arg_zzzz, a->esz == 2 ? gen_helper_gvec_usdot_b : NULL, a, 0) -TRANS_FEAT(AESMC, aa64_sve2_aes, gen_gvec_ool_zz, - gen_helper_crypto_aesmc, a->rd, a->rd, a->decrypt) +TRANS_FEAT_NONSTREAMING(AESMC, aa64_sve2_aes, gen_gvec_ool_zz, + gen_helper_crypto_aesmc, a->rd, a->rd, a->decrypt) -TRANS_FEAT(AESE, aa64_sve2_aes, gen_gvec_ool_arg_zzz, - gen_helper_crypto_aese, a, false) -TRANS_FEAT(AESD, aa64_sve2_aes, gen_gvec_ool_arg_zzz, - gen_helper_crypto_aese, a, true) +TRANS_FEAT_NONSTREAMING(AESE, aa64_sve2_aes, gen_gvec_ool_arg_zzz, + gen_helper_crypto_aese, a, false) +TRANS_FEAT_NONSTREAMING(AESD, aa64_sve2_aes, gen_gvec_ool_arg_zzz, + gen_helper_crypto_aese, a, true) -TRANS_FEAT(SM4E, aa64_sve2_sm4, gen_gvec_ool_arg_zzz, - gen_helper_crypto_sm4e, a, 0) -TRANS_FEAT(SM4EKEY, aa64_sve2_sm4, gen_gvec_ool_arg_zzz, - gen_helper_crypto_sm4ekey, a, 0) +TRANS_FEAT_NONSTREAMING(SM4E, aa64_sve2_sm4, gen_gvec_ool_arg_zzz, + gen_helper_crypto_sm4e, a, 0) +TRANS_FEAT_NONSTREAMING(SM4EKEY, aa64_sve2_sm4, gen_gvec_ool_arg_zzz, + gen_helper_crypto_sm4ekey, a, 0) -TRANS_FEAT(RAX1, aa64_sve2_sha3, gen_gvec_fn_arg_zzz, gen_gvec_rax1, a) +TRANS_FEAT_NONSTREAMING(RAX1, aa64_sve2_sha3, gen_gvec_fn_arg_zzz, + gen_gvec_rax1, a) TRANS_FEAT(FCVTNT_sh, aa64_sve2, gen_gvec_fpst_arg_zpz, gen_helper_sve2_fcvtnt_sh, a, 0, FPST_FPCR) -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Mark these as a non-streaming instructions, which should trap if full a64 support is not enabled in streaming mode. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-12-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/sme-fa64.decode | 9 --------- target/arm/translate-sve.c | 6 ++++++ 2 files changed, 6 insertions(+), 9 deletions(-) diff --git a/target/arm/sme-fa64.decode b/target/arm/sme-fa64.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme-fa64.decode +++ b/target/arm/sme-fa64.decode @@ -XXX,XX +XXX,XX @@ FAIL 0001 1110 0111 1110 0000 00-- ---- ---- # FJCVTZS # --11 1100 --1- ---- ---- ---- ---- --10 # Load/store FP register (register offset) # --11 1101 ---- ---- ---- ---- ---- ---- # Load/store FP register (scaled imm) -FAIL 1000 010- -00- ---- 10-- ---- ---- ---- # SVE2 32-bit gather NT load (vector+scalar) FAIL 1000 010- -00- ---- 111- ---- ---- ---- # SVE 32-bit gather prefetch (vector+imm) FAIL 1000 0100 0-1- ---- 0--- ---- ---- ---- # SVE 32-bit gather prefetch (scalar+vector) -FAIL 1000 010- -01- ---- 1--- ---- ---- ---- # SVE 32-bit gather load (vector+imm) -FAIL 1000 0100 0-0- ---- 0--- ---- ---- ---- # SVE 32-bit gather load byte (scalar+vector) -FAIL 1000 0100 1--- ---- 0--- ---- ---- ---- # SVE 32-bit gather load half (scalar+vector) -FAIL 1000 0101 0--- ---- 0--- ---- ---- ---- # SVE 32-bit gather load word (scalar+vector) FAIL 1010 010- ---- ---- 011- ---- ---- ---- # SVE contiguous FF load (scalar+scalar) FAIL 1010 010- ---1 ---- 101- ---- ---- ---- # SVE contiguous NF load (scalar+imm) FAIL 1010 010- -01- ---- 000- ---- ---- ---- # SVE load & replicate 32 bytes (scalar+scalar) FAIL 1010 010- -010 ---- 001- ---- ---- ---- # SVE load & replicate 32 bytes (scalar+imm) FAIL 1100 010- ---- ---- ---- ---- ---- ---- # SVE 64-bit gather load/prefetch -FAIL 1110 010- -00- ---- 001- ---- ---- ---- # SVE2 64-bit scatter NT store (vector+scalar) -FAIL 1110 010- -10- ---- 001- ---- ---- ---- # SVE2 32-bit scatter NT store (vector+scalar) -FAIL 1110 010- ---- ---- 1-0- ---- ---- ---- # SVE scatter store (scalar+32-bit vector) -FAIL 1110 010- ---- ---- 101- ---- ---- ---- # SVE scatter store (misc) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool trans_LD1_zprz(DisasContext *s, arg_LD1_zprz *a) if (!dc_isar_feature(aa64_sve, s)) { return false; } + s->is_nonstreaming = true; if (!sve_access_check(s)) { return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_LD1_zpiz(DisasContext *s, arg_LD1_zpiz *a) if (!dc_isar_feature(aa64_sve, s)) { return false; } + s->is_nonstreaming = true; if (!sve_access_check(s)) { return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_LDNT1_zprz(DisasContext *s, arg_LD1_zprz *a) if (!dc_isar_feature(aa64_sve2, s)) { return false; } + s->is_nonstreaming = true; if (!sve_access_check(s)) { return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_ST1_zprz(DisasContext *s, arg_ST1_zprz *a) if (!dc_isar_feature(aa64_sve, s)) { return false; } + s->is_nonstreaming = true; if (!sve_access_check(s)) { return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_ST1_zpiz(DisasContext *s, arg_ST1_zpiz *a) if (!dc_isar_feature(aa64_sve, s)) { return false; } + s->is_nonstreaming = true; if (!sve_access_check(s)) { return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_STNT1_zprz(DisasContext *s, arg_ST1_zprz *a) if (!dc_isar_feature(aa64_sve2, s)) { return false; } + s->is_nonstreaming = true; if (!sve_access_check(s)) { return true; } -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Mark these as a non-streaming instructions, which should trap if full a64 support is not enabled in streaming mode. In this case, introduce PRF_ns (prefetch non-streaming) to handle the checks. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-13-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/sme-fa64.decode | 3 --- target/arm/sve.decode | 10 +++++----- target/arm/translate-sve.c | 11 +++++++++++ 3 files changed, 16 insertions(+), 8 deletions(-) diff --git a/target/arm/sme-fa64.decode b/target/arm/sme-fa64.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme-fa64.decode +++ b/target/arm/sme-fa64.decode @@ -XXX,XX +XXX,XX @@ FAIL 0001 1110 0111 1110 0000 00-- ---- ---- # FJCVTZS # --11 1100 --1- ---- ---- ---- ---- --10 # Load/store FP register (register offset) # --11 1101 ---- ---- ---- ---- ---- ---- # Load/store FP register (scaled imm) -FAIL 1000 010- -00- ---- 111- ---- ---- ---- # SVE 32-bit gather prefetch (vector+imm) -FAIL 1000 0100 0-1- ---- 0--- ---- ---- ---- # SVE 32-bit gather prefetch (scalar+vector) FAIL 1010 010- ---- ---- 011- ---- ---- ---- # SVE contiguous FF load (scalar+scalar) FAIL 1010 010- ---1 ---- 101- ---- ---- ---- # SVE contiguous NF load (scalar+imm) FAIL 1010 010- -01- ---- 000- ---- ---- ---- # SVE load & replicate 32 bytes (scalar+scalar) FAIL 1010 010- -010 ---- 001- ---- ---- ---- # SVE load & replicate 32 bytes (scalar+imm) -FAIL 1100 010- ---- ---- ---- ---- ---- ---- # SVE 64-bit gather load/prefetch diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ LD1RO_zpri 1010010 .. 01 0.... 001 ... ..... ..... \ @rpri_load_msz nreg=0 # SVE 32-bit gather prefetch (scalar plus 32-bit scaled offsets) -PRF 1000010 00 -1 ----- 0-- --- ----- 0 ---- +PRF_ns 1000010 00 -1 ----- 0-- --- ----- 0 ---- # SVE 32-bit gather prefetch (vector plus immediate) -PRF 1000010 -- 00 ----- 111 --- ----- 0 ---- +PRF_ns 1000010 -- 00 ----- 111 --- ----- 0 ---- # SVE contiguous prefetch (scalar plus immediate) PRF 1000010 11 1- ----- 0-- --- ----- 0 ---- @@ -XXX,XX +XXX,XX @@ LD1_zpiz 1100010 .. 01 ..... 1.. ... ..... ..... \ @rpri_g_load esz=3 # SVE 64-bit gather prefetch (scalar plus 64-bit scaled offsets) -PRF 1100010 00 11 ----- 1-- --- ----- 0 ---- +PRF_ns 1100010 00 11 ----- 1-- --- ----- 0 ---- # SVE 64-bit gather prefetch (scalar plus unpacked 32-bit scaled offsets) -PRF 1100010 00 -1 ----- 0-- --- ----- 0 ---- +PRF_ns 1100010 00 -1 ----- 0-- --- ----- 0 ---- # SVE 64-bit gather prefetch (vector plus immediate) -PRF 1100010 -- 00 ----- 111 --- ----- 0 ---- +PRF_ns 1100010 -- 00 ----- 111 --- ----- 0 ---- ### SVE Memory Store Group diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool trans_PRF_rr(DisasContext *s, arg_PRF_rr *a) return true; } +static bool trans_PRF_ns(DisasContext *s, arg_PRF_ns *a) +{ + if (!dc_isar_feature(aa64_sve, s)) { + return false; + } + /* Prefetch is a nop within QEMU. */ + s->is_nonstreaming = true; + (void)sve_access_check(s); + return true; +} + /* * Move Prefix * -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Mark these as a non-streaming instructions, which should trap if full a64 support is not enabled in streaming mode. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-14-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/sme-fa64.decode | 2 -- target/arm/translate-sve.c | 2 ++ 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/target/arm/sme-fa64.decode b/target/arm/sme-fa64.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme-fa64.decode +++ b/target/arm/sme-fa64.decode @@ -XXX,XX +XXX,XX @@ FAIL 0001 1110 0111 1110 0000 00-- ---- ---- # FJCVTZS # --11 1100 --1- ---- ---- ---- ---- --10 # Load/store FP register (register offset) # --11 1101 ---- ---- ---- ---- ---- ---- # Load/store FP register (scaled imm) -FAIL 1010 010- ---- ---- 011- ---- ---- ---- # SVE contiguous FF load (scalar+scalar) -FAIL 1010 010- ---1 ---- 101- ---- ---- ---- # SVE contiguous NF load (scalar+imm) FAIL 1010 010- -01- ---- 000- ---- ---- ---- # SVE load & replicate 32 bytes (scalar+scalar) FAIL 1010 010- -010 ---- 001- ---- ---- ---- # SVE load & replicate 32 bytes (scalar+imm) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool trans_LDFF1_zprr(DisasContext *s, arg_rprr_load *a) if (!dc_isar_feature(aa64_sve, s)) { return false; } + s->is_nonstreaming = true; if (sve_access_check(s)) { TCGv_i64 addr = new_tmp_a64(s); tcg_gen_shli_i64(addr, cpu_reg(s, a->rm), dtype_msz(a->dtype)); @@ -XXX,XX +XXX,XX @@ static bool trans_LDNF1_zpri(DisasContext *s, arg_rpri_load *a) if (!dc_isar_feature(aa64_sve, s)) { return false; } + s->is_nonstreaming = true; if (sve_access_check(s)) { int vsz = vec_full_reg_size(s); int elements = vsz >> dtype_esz[a->dtype]; -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Mark these as a non-streaming instructions, which should trap if full a64 support is not enabled in streaming mode. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-15-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/sme-fa64.decode | 3 --- target/arm/translate-sve.c | 2 ++ 2 files changed, 2 insertions(+), 3 deletions(-) diff --git a/target/arm/sme-fa64.decode b/target/arm/sme-fa64.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme-fa64.decode +++ b/target/arm/sme-fa64.decode @@ -XXX,XX +XXX,XX @@ FAIL 0001 1110 0111 1110 0000 00-- ---- ---- # FJCVTZS # --11 1100 --0- ---- ---- ---- ---- ---- # Load/store FP register (unscaled imm) # --11 1100 --1- ---- ---- ---- ---- --10 # Load/store FP register (register offset) # --11 1101 ---- ---- ---- ---- ---- ---- # Load/store FP register (scaled imm) - -FAIL 1010 010- -01- ---- 000- ---- ---- ---- # SVE load & replicate 32 bytes (scalar+scalar) -FAIL 1010 010- -010 ---- 001- ---- ---- ---- # SVE load & replicate 32 bytes (scalar+imm) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool trans_LD1RO_zprr(DisasContext *s, arg_rprr_load *a) if (a->rm == 31) { return false; } + s->is_nonstreaming = true; if (sve_access_check(s)) { TCGv_i64 addr = new_tmp_a64(s); tcg_gen_shli_i64(addr, cpu_reg(s, a->rm), dtype_msz(a->dtype)); @@ -XXX,XX +XXX,XX @@ static bool trans_LD1RO_zpri(DisasContext *s, arg_rpri_load *a) if (!dc_isar_feature(aa64_sve_f64mm, s)) { return false; } + s->is_nonstreaming = true; if (sve_access_check(s)) { TCGv_i64 addr = new_tmp_a64(s); tcg_gen_addi_i64(addr, cpu_reg_sp(s, a->rn), a->imm * 32); -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> These functions will be used to verify that the cpu is in the correct state for a given instruction. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-16-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-a64.h | 21 +++++++++++++++++++++ target/arm/translate-a64.c | 34 ++++++++++++++++++++++++++++++++++ 2 files changed, 55 insertions(+) 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 @@ void write_fp_dreg(DisasContext *s, int reg, TCGv_i64 v); bool logic_imm_decode_wmask(uint64_t *result, unsigned int immn, unsigned int imms, unsigned int immr); bool sve_access_check(DisasContext *s); +bool sme_enabled_check(DisasContext *s); +bool sme_enabled_check_with_svcr(DisasContext *s, unsigned); + +/* This function corresponds to CheckStreamingSVEEnabled. */ +static inline bool sme_sm_enabled_check(DisasContext *s) +{ + return sme_enabled_check_with_svcr(s, R_SVCR_SM_MASK); +} + +/* This function corresponds to CheckSMEAndZAEnabled. */ +static inline bool sme_za_enabled_check(DisasContext *s) +{ + return sme_enabled_check_with_svcr(s, R_SVCR_ZA_MASK); +} + +/* Note that this function corresponds to CheckStreamingSVEAndZAEnabled. */ +static inline bool sme_smza_enabled_check(DisasContext *s) +{ + return sme_enabled_check_with_svcr(s, R_SVCR_SM_MASK | R_SVCR_ZA_MASK); +} + TCGv_i64 clean_data_tbi(DisasContext *s, TCGv_i64 addr); TCGv_i64 gen_mte_check1(DisasContext *s, TCGv_i64 addr, bool is_write, bool tag_checked, int log2_size); 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 sme_access_check(DisasContext *s) return true; } +/* This function corresponds to CheckSMEEnabled. */ +bool sme_enabled_check(DisasContext *s) +{ + /* + * Note that unlike sve_excp_el, we have not constrained sme_excp_el + * to be zero when fp_excp_el has priority. This is because we need + * sme_excp_el by itself for cpregs access checks. + */ + if (!s->fp_excp_el || s->sme_excp_el < s->fp_excp_el) { + s->fp_access_checked = true; + return sme_access_check(s); + } + return fp_access_check_only(s); +} + +/* Common subroutine for CheckSMEAnd*Enabled. */ +bool sme_enabled_check_with_svcr(DisasContext *s, unsigned req) +{ + if (!sme_enabled_check(s)) { + return false; + } + if (FIELD_EX64(req, SVCR, SM) && !s->pstate_sm) { + gen_exception_insn(s, s->pc_curr, 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, + syn_smetrap(SME_ET_InactiveZA, false)); + return false; + } + return true; +} + /* * This utility function is for doing register extension with an * optional shift. You will likely want to pass a temporary for the -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> The pseudocode for CheckSVEEnabled gains a check for Streaming SVE mode, and for SME present but SVE absent. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-17-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-a64.c | 22 ++++++++++++++++------ 1 file changed, 16 insertions(+), 6 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 bool fp_access_check(DisasContext *s) return true; } -/* Check that SVE access is enabled. If it is, return true. +/* + * Check that SVE access is enabled. If it is, return true. * If not, emit code to generate an appropriate exception and return false. + * This function corresponds to CheckSVEEnabled(). */ bool sve_access_check(DisasContext *s) { - if (s->sve_excp_el) { - assert(!s->sve_access_checked); - s->sve_access_checked = true; - + if (s->pstate_sm || !dc_isar_feature(aa64_sve, s)) { + assert(dc_isar_feature(aa64_sme, s)); + if (!sme_sm_enabled_check(s)) { + goto fail_exit; + } + } else if (s->sve_excp_el) { gen_exception_insn_el(s, s->pc_curr, EXCP_UDEF, syn_sve_access_trap(), s->sve_excp_el); - return false; + goto fail_exit; } s->sve_access_checked = true; return fp_access_check(s); + + fail_exit: + /* Assert that we only raise one exception per instruction. */ + assert(!s->sve_access_checked); + s->sve_access_checked = true; + return false; } /* -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> These SME instructions are nominally within the SVE decode space, so we add them to sve.decode and translate-sve.c. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-18-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-a64.h | 12 ++++++++++++ target/arm/sve.decode | 5 ++++- target/arm/translate-sve.c | 38 ++++++++++++++++++++++++++++++++++++++ 3 files changed, 54 insertions(+), 1 deletion(-) 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) return s->vl; } +/* Return the byte size of the vector register, SVL / 8. */ +static inline int streaming_vec_reg_size(DisasContext *s) +{ + return s->svl; +} + /* * Return the offset info CPUARMState of the predicate vector register Pn. * Note for this purpose, FFR is P16. @@ -XXX,XX +XXX,XX @@ static inline int pred_full_reg_size(DisasContext *s) return s->vl >> 3; } +/* Return the byte size of the predicate register, SVL / 64. */ +static inline int streaming_pred_reg_size(DisasContext *s) +{ + return s->svl >> 3; +} + /* * Round up the size of a register to a size allowed by * the tcg vector infrastructure. Any operation which uses this diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ INDEX_ri 00000100 esz:2 1 imm:s5 010001 rn:5 rd:5 # SVE index generation (register start, register increment) INDEX_rr 00000100 .. 1 ..... 010011 ..... ..... @rd_rn_rm -### SVE Stack Allocation Group +### SVE / Streaming SVE Stack Allocation Group # SVE stack frame adjustment ADDVL 00000100 001 ..... 01010 ...... ..... @rd_rn_i6 +ADDSVL 00000100 001 ..... 01011 ...... ..... @rd_rn_i6 ADDPL 00000100 011 ..... 01010 ...... ..... @rd_rn_i6 +ADDSPL 00000100 011 ..... 01011 ...... ..... @rd_rn_i6 # SVE stack frame size RDVL 00000100 101 11111 01010 imm:s6 rd:5 +RDSVL 00000100 101 11111 01011 imm:s6 rd:5 ### SVE Bitwise Shift - Unpredicated Group diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool trans_ADDVL(DisasContext *s, arg_ADDVL *a) return true; } +static bool trans_ADDSVL(DisasContext *s, arg_ADDSVL *a) +{ + if (!dc_isar_feature(aa64_sme, s)) { + return false; + } + if (sme_enabled_check(s)) { + TCGv_i64 rd = cpu_reg_sp(s, a->rd); + TCGv_i64 rn = cpu_reg_sp(s, a->rn); + tcg_gen_addi_i64(rd, rn, a->imm * streaming_vec_reg_size(s)); + } + return true; +} + static bool trans_ADDPL(DisasContext *s, arg_ADDPL *a) { if (!dc_isar_feature(aa64_sve, s)) { @@ -XXX,XX +XXX,XX @@ static bool trans_ADDPL(DisasContext *s, arg_ADDPL *a) return true; } +static bool trans_ADDSPL(DisasContext *s, arg_ADDSPL *a) +{ + if (!dc_isar_feature(aa64_sme, s)) { + return false; + } + if (sme_enabled_check(s)) { + TCGv_i64 rd = cpu_reg_sp(s, a->rd); + TCGv_i64 rn = cpu_reg_sp(s, a->rn); + tcg_gen_addi_i64(rd, rn, a->imm * streaming_pred_reg_size(s)); + } + return true; +} + static bool trans_RDVL(DisasContext *s, arg_RDVL *a) { if (!dc_isar_feature(aa64_sve, s)) { @@ -XXX,XX +XXX,XX @@ static bool trans_RDVL(DisasContext *s, arg_RDVL *a) return true; } +static bool trans_RDSVL(DisasContext *s, arg_RDSVL *a) +{ + if (!dc_isar_feature(aa64_sme, s)) { + return false; + } + if (sme_enabled_check(s)) { + TCGv_i64 reg = cpu_reg(s, a->rd); + tcg_gen_movi_i64(reg, a->imm * streaming_vec_reg_size(s)); + } + return true; +} + /* *** SVE Compute Vector Address Group */ -- 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: 20220708151540.18136-19-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sme.h | 2 ++ target/arm/sme.decode | 4 ++++ target/arm/sme_helper.c | 25 +++++++++++++++++++++++++ target/arm/translate-sme.c | 13 +++++++++++++ 4 files changed, 44 insertions(+) diff --git a/target/arm/helper-sme.h b/target/arm/helper-sme.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sme.h +++ b/target/arm/helper-sme.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_2(set_pstate_sm, TCG_CALL_NO_RWG, void, env, i32) DEF_HELPER_FLAGS_2(set_pstate_za, TCG_CALL_NO_RWG, void, env, i32) + +DEF_HELPER_FLAGS_3(sme_zero, TCG_CALL_NO_RWG, void, env, i32, i32) diff --git a/target/arm/sme.decode b/target/arm/sme.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme.decode +++ b/target/arm/sme.decode @@ -XXX,XX +XXX,XX @@ # # This file is processed by scripts/decodetree.py # + +### SME Misc + +ZERO 11000000 00 001 00000000000 imm:8 diff --git a/target/arm/sme_helper.c b/target/arm/sme_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme_helper.c +++ b/target/arm/sme_helper.c @@ -XXX,XX +XXX,XX @@ void helper_set_pstate_za(CPUARMState *env, uint32_t i) memset(env->zarray, 0, sizeof(env->zarray)); } } + +void helper_sme_zero(CPUARMState *env, uint32_t imm, uint32_t svl) +{ + uint32_t i; + + /* + * Special case clearing the entire ZA space. + * This falls into the CONSTRAINED UNPREDICTABLE zeroing of any + * parts of the ZA storage outside of SVL. + */ + if (imm == 0xff) { + memset(env->zarray, 0, sizeof(env->zarray)); + return; + } + + /* + * Recall that ZAnH.D[m] is spread across ZA[n+8*m], + * so each row is discontiguous within ZA[]. + */ + for (i = 0; i < svl; i++) { + if (imm & (1 << (i % 8))) { + memset(&env->zarray[i], 0, svl); + } + } +} diff --git a/target/arm/translate-sme.c b/target/arm/translate-sme.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sme.c +++ b/target/arm/translate-sme.c @@ -XXX,XX +XXX,XX @@ */ #include "decode-sme.c.inc" + + +static bool trans_ZERO(DisasContext *s, arg_ZERO *a) +{ + if (!dc_isar_feature(aa64_sme, s)) { + return false; + } + if (sme_za_enabled_check(s)) { + gen_helper_sme_zero(cpu_env, tcg_constant_i32(a->imm), + tcg_constant_i32(streaming_vec_reg_size(s))); + } + return true; +} -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> We can reuse the SVE functions for implementing moves to/from horizontal tile slices, but we need new ones for moves to/from vertical tile slices. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-20-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sme.h | 12 +++ target/arm/helper-sve.h | 2 + target/arm/translate-a64.h | 8 ++ target/arm/translate.h | 5 ++ target/arm/sme.decode | 15 ++++ target/arm/sme_helper.c | 151 ++++++++++++++++++++++++++++++++++++- target/arm/sve_helper.c | 12 +++ target/arm/translate-sme.c | 127 +++++++++++++++++++++++++++++++ 8 files changed, 331 insertions(+), 1 deletion(-) diff --git a/target/arm/helper-sme.h b/target/arm/helper-sme.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sme.h +++ b/target/arm/helper-sme.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_2(set_pstate_sm, TCG_CALL_NO_RWG, void, env, i32) DEF_HELPER_FLAGS_2(set_pstate_za, TCG_CALL_NO_RWG, void, env, i32) DEF_HELPER_FLAGS_3(sme_zero, TCG_CALL_NO_RWG, void, env, i32, i32) + +/* Move to/from vertical array slices, i.e. columns, so 'c'. */ +DEF_HELPER_FLAGS_4(sme_mova_cz_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sme_mova_zc_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sme_mova_cz_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sme_mova_zc_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sme_mova_cz_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sme_mova_zc_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sme_mova_cz_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sme_mova_zc_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sme_mova_cz_q, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sme_mova_zc_q, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_5(sve_sel_zpzz_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_sel_zpzz_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_sel_zpzz_q, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve2_addp_zpzz_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) 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 pred_gvec_reg_size(DisasContext *s) return size_for_gvec(pred_full_reg_size(s)); } +/* Return a newly allocated pointer to the predicate register. */ +static inline TCGv_ptr pred_full_reg_ptr(DisasContext *s, int regno) +{ + TCGv_ptr ret = tcg_temp_new_ptr(); + tcg_gen_addi_ptr(ret, cpu_env, pred_full_reg_offset(s, regno)); + return ret; +} + bool disas_sve(DisasContext *, uint32_t); bool disas_sme(DisasContext *, uint32_t); 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 plus_2(DisasContext *s, int x) return x + 2; } +static inline int plus_12(DisasContext *s, int x) +{ + return x + 12; +} + static inline int times_2(DisasContext *s, int x) { return x * 2; diff --git a/target/arm/sme.decode b/target/arm/sme.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme.decode +++ b/target/arm/sme.decode @@ -XXX,XX +XXX,XX @@ ### SME Misc ZERO 11000000 00 001 00000000000 imm:8 + +### SME Move into/from Array + +%mova_rs 13:2 !function=plus_12 +&mova esz rs pg zr za_imm v:bool to_vec:bool + +MOVA 11000000 esz:2 00000 0 v:1 .. pg:3 zr:5 0 za_imm:4 \ + &mova to_vec=0 rs=%mova_rs +MOVA 11000000 11 00000 1 v:1 .. pg:3 zr:5 0 za_imm:4 \ + &mova to_vec=0 rs=%mova_rs esz=4 + +MOVA 11000000 esz:2 00001 0 v:1 .. pg:3 0 za_imm:4 zr:5 \ + &mova to_vec=1 rs=%mova_rs +MOVA 11000000 11 00001 1 v:1 .. pg:3 0 za_imm:4 zr:5 \ + &mova to_vec=1 rs=%mova_rs esz=4 diff --git a/target/arm/sme_helper.c b/target/arm/sme_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme_helper.c +++ b/target/arm/sme_helper.c @@ -XXX,XX +XXX,XX @@ #include "qemu/osdep.h" #include "cpu.h" -#include "internals.h" +#include "tcg/tcg-gvec-desc.h" #include "exec/helper-proto.h" +#include "qemu/int128.h" +#include "vec_internal.h" /* ResetSVEState */ void arm_reset_sve_state(CPUARMState *env) @@ -XXX,XX +XXX,XX @@ void helper_sme_zero(CPUARMState *env, uint32_t imm, uint32_t svl) } } } + + +/* + * When considering the ZA storage as an array of elements of + * type T, the index within that array of the Nth element of + * a vertical slice of a tile can be calculated like this, + * regardless of the size of type T. This is because the tiles + * are interleaved, so if type T is size N bytes then row 1 of + * the tile is N rows away from row 0. The division by N to + * convert a byte offset into an array index and the multiplication + * by N to convert from vslice-index-within-the-tile to + * the index within the ZA storage cancel out. + */ +#define tile_vslice_index(i) ((i) * sizeof(ARMVectorReg)) + +/* + * When doing byte arithmetic on the ZA storage, the element + * byteoff bytes away in a tile vertical slice is always this + * many bytes away in the ZA storage, regardless of the + * size of the tile element, assuming that byteoff is a multiple + * of the element size. Again this is because of the interleaving + * of the tiles. For instance if we have 1 byte per element then + * each row of the ZA storage has one byte of the vslice data, + * and (counting from 0) byte 8 goes in row 8 of the storage + * at offset (8 * row-size-in-bytes). + * If we have 8 bytes per element then each row of the ZA storage + * has 8 bytes of the data, but there are 8 interleaved tiles and + * so byte 8 of the data goes into row 1 of the tile, + * which is again row 8 of the storage, so the offset is still + * (8 * row-size-in-bytes). Similarly for other element sizes. + */ +#define tile_vslice_offset(byteoff) ((byteoff) * sizeof(ARMVectorReg)) + + +/* + * Move Zreg vector to ZArray column. + */ +#define DO_MOVA_C(NAME, TYPE, H) \ +void HELPER(NAME)(void *za, void *vn, void *vg, uint32_t desc) \ +{ \ + int i, oprsz = simd_oprsz(desc); \ + for (i = 0; i < oprsz; ) { \ + uint16_t pg = *(uint16_t *)(vg + H1_2(i >> 3)); \ + do { \ + if (pg & 1) { \ + *(TYPE *)(za + tile_vslice_offset(i)) = *(TYPE *)(vn + H(i)); \ + } \ + i += sizeof(TYPE); \ + pg >>= sizeof(TYPE); \ + } while (i & 15); \ + } \ +} + +DO_MOVA_C(sme_mova_cz_b, uint8_t, H1) +DO_MOVA_C(sme_mova_cz_h, uint16_t, H1_2) +DO_MOVA_C(sme_mova_cz_s, uint32_t, H1_4) + +void HELPER(sme_mova_cz_d)(void *za, void *vn, void *vg, uint32_t desc) +{ + int i, oprsz = simd_oprsz(desc) / 8; + uint8_t *pg = vg; + uint64_t *n = vn; + uint64_t *a = za; + + for (i = 0; i < oprsz; i++) { + if (pg[H1(i)] & 1) { + a[tile_vslice_index(i)] = n[i]; + } + } +} + +void HELPER(sme_mova_cz_q)(void *za, void *vn, void *vg, uint32_t desc) +{ + int i, oprsz = simd_oprsz(desc) / 16; + uint16_t *pg = vg; + Int128 *n = vn; + Int128 *a = za; + + /* + * Int128 is used here simply to copy 16 bytes, and to simplify + * the address arithmetic. + */ + for (i = 0; i < oprsz; i++) { + if (pg[H2(i)] & 1) { + a[tile_vslice_index(i)] = n[i]; + } + } +} + +#undef DO_MOVA_C + +/* + * Move ZArray column to Zreg vector. + */ +#define DO_MOVA_Z(NAME, TYPE, H) \ +void HELPER(NAME)(void *vd, void *za, void *vg, uint32_t desc) \ +{ \ + int i, oprsz = simd_oprsz(desc); \ + for (i = 0; i < oprsz; ) { \ + uint16_t pg = *(uint16_t *)(vg + H1_2(i >> 3)); \ + do { \ + if (pg & 1) { \ + *(TYPE *)(vd + H(i)) = *(TYPE *)(za + tile_vslice_offset(i)); \ + } \ + i += sizeof(TYPE); \ + pg >>= sizeof(TYPE); \ + } while (i & 15); \ + } \ +} + +DO_MOVA_Z(sme_mova_zc_b, uint8_t, H1) +DO_MOVA_Z(sme_mova_zc_h, uint16_t, H1_2) +DO_MOVA_Z(sme_mova_zc_s, uint32_t, H1_4) + +void HELPER(sme_mova_zc_d)(void *vd, void *za, void *vg, uint32_t desc) +{ + int i, oprsz = simd_oprsz(desc) / 8; + uint8_t *pg = vg; + uint64_t *d = vd; + uint64_t *a = za; + + for (i = 0; i < oprsz; i++) { + if (pg[H1(i)] & 1) { + d[i] = a[tile_vslice_index(i)]; + } + } +} + +void HELPER(sme_mova_zc_q)(void *vd, void *za, void *vg, uint32_t desc) +{ + int i, oprsz = simd_oprsz(desc) / 16; + uint16_t *pg = vg; + Int128 *d = vd; + Int128 *a = za; + + /* + * Int128 is used here simply to copy 16 bytes, and to simplify + * the address arithmetic. + */ + for (i = 0; i < oprsz; i++, za += sizeof(ARMVectorReg)) { + if (pg[H2(i)] & 1) { + d[i] = a[tile_vslice_index(i)]; + } + } +} + +#undef DO_MOVA_Z 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 @@ void HELPER(sve_sel_zpzz_d)(void *vd, void *vn, void *vm, } } +void HELPER(sve_sel_zpzz_q)(void *vd, void *vn, void *vm, + void *vg, uint32_t desc) +{ + intptr_t i, opr_sz = simd_oprsz(desc) / 16; + Int128 *d = vd, *n = vn, *m = vm; + uint16_t *pg = vg; + + for (i = 0; i < opr_sz; i += 1) { + d[i] = (pg[H2(i)] & 1 ? n : m)[i]; + } +} + /* Two operand comparison controlled by a predicate. * ??? It is very tempting to want to be able to expand this inline * with x86 instructions, e.g. diff --git a/target/arm/translate-sme.c b/target/arm/translate-sme.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sme.c +++ b/target/arm/translate-sme.c @@ -XXX,XX +XXX,XX @@ #include "decode-sme.c.inc" +/* + * Resolve tile.size[index] to a host pointer, where tile and index + * are always decoded together, dependent on the element size. + */ +static TCGv_ptr get_tile_rowcol(DisasContext *s, int esz, int rs, + int tile_index, bool vertical) +{ + int tile = tile_index >> (4 - esz); + int index = esz == MO_128 ? 0 : extract32(tile_index, 0, 4 - esz); + int pos, len, offset; + TCGv_i32 tmp; + TCGv_ptr addr; + + /* Compute the final index, which is Rs+imm. */ + tmp = tcg_temp_new_i32(); + tcg_gen_trunc_tl_i32(tmp, cpu_reg(s, rs)); + tcg_gen_addi_i32(tmp, tmp, index); + + /* Prepare a power-of-two modulo via extraction of @len bits. */ + len = ctz32(streaming_vec_reg_size(s)) - esz; + + if (vertical) { + /* + * Compute the byte offset of the index within the tile: + * (index % (svl / size)) * size + * = (index % (svl >> esz)) << esz + * Perform the power-of-two modulo via extraction of the low @len bits. + * Perform the multiply by shifting left by @pos bits. + * Perform these operations simultaneously via deposit into zero. + */ + pos = esz; + tcg_gen_deposit_z_i32(tmp, tmp, pos, len); + + /* + * For big-endian, adjust the indexed column byte offset within + * the uint64_t host words that make up env->zarray[]. + */ + if (HOST_BIG_ENDIAN && esz < MO_64) { + tcg_gen_xori_i32(tmp, tmp, 8 - (1 << esz)); + } + } else { + /* + * Compute the byte offset of the index within the tile: + * (index % (svl / size)) * (size * sizeof(row)) + * = (index % (svl >> esz)) << (esz + log2(sizeof(row))) + */ + pos = esz + ctz32(sizeof(ARMVectorReg)); + tcg_gen_deposit_z_i32(tmp, tmp, pos, len); + + /* Row slices are always aligned and need no endian adjustment. */ + } + + /* The tile byte offset within env->zarray is the row. */ + offset = tile * sizeof(ARMVectorReg); + + /* Include the byte offset of zarray to make this relative to env. */ + offset += offsetof(CPUARMState, zarray); + tcg_gen_addi_i32(tmp, tmp, offset); + + /* Add the byte offset to env to produce the final pointer. */ + addr = tcg_temp_new_ptr(); + tcg_gen_ext_i32_ptr(addr, tmp); + tcg_temp_free_i32(tmp); + tcg_gen_add_ptr(addr, addr, cpu_env); + + return addr; +} + static bool trans_ZERO(DisasContext *s, arg_ZERO *a) { if (!dc_isar_feature(aa64_sme, s)) { @@ -XXX,XX +XXX,XX @@ static bool trans_ZERO(DisasContext *s, arg_ZERO *a) } return true; } + +static bool trans_MOVA(DisasContext *s, arg_MOVA *a) +{ + static gen_helper_gvec_4 * const h_fns[5] = { + gen_helper_sve_sel_zpzz_b, gen_helper_sve_sel_zpzz_h, + gen_helper_sve_sel_zpzz_s, gen_helper_sve_sel_zpzz_d, + gen_helper_sve_sel_zpzz_q + }; + static gen_helper_gvec_3 * const cz_fns[5] = { + gen_helper_sme_mova_cz_b, gen_helper_sme_mova_cz_h, + gen_helper_sme_mova_cz_s, gen_helper_sme_mova_cz_d, + gen_helper_sme_mova_cz_q, + }; + static gen_helper_gvec_3 * const zc_fns[5] = { + gen_helper_sme_mova_zc_b, gen_helper_sme_mova_zc_h, + gen_helper_sme_mova_zc_s, gen_helper_sme_mova_zc_d, + gen_helper_sme_mova_zc_q, + }; + + TCGv_ptr t_za, t_zr, t_pg; + TCGv_i32 t_desc; + int svl; + + if (!dc_isar_feature(aa64_sme, s)) { + return false; + } + if (!sme_smza_enabled_check(s)) { + return true; + } + + t_za = get_tile_rowcol(s, a->esz, a->rs, a->za_imm, a->v); + t_zr = vec_full_reg_ptr(s, a->zr); + t_pg = pred_full_reg_ptr(s, a->pg); + + svl = streaming_vec_reg_size(s); + t_desc = tcg_constant_i32(simd_desc(svl, svl, 0)); + + if (a->v) { + /* Vertical slice -- use sme mova helpers. */ + if (a->to_vec) { + zc_fns[a->esz](t_zr, t_za, t_pg, t_desc); + } else { + cz_fns[a->esz](t_za, t_zr, t_pg, t_desc); + } + } else { + /* Horizontal slice -- reuse sve sel helpers. */ + if (a->to_vec) { + h_fns[a->esz](t_zr, t_za, t_zr, t_pg, t_desc); + } else { + h_fns[a->esz](t_za, t_zr, t_za, t_pg, t_desc); + } + } + + tcg_temp_free_ptr(t_za); + tcg_temp_free_ptr(t_zr); + tcg_temp_free_ptr(t_pg); + + return true; +} -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> We cannot reuse the SVE functions for LD[1-4] and ST[1-4], because those functions accept only a Zreg register number. For SME, we want to pass a pointer into ZA storage. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-21-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sme.h | 82 +++++ target/arm/sme.decode | 9 + target/arm/sme_helper.c | 595 +++++++++++++++++++++++++++++++++++++ target/arm/translate-sme.c | 70 +++++ 4 files changed, 756 insertions(+) diff --git a/target/arm/helper-sme.h b/target/arm/helper-sme.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sme.h +++ b/target/arm/helper-sme.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(sme_mova_cz_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(sme_mova_zc_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(sme_mova_cz_q, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(sme_mova_zc_q, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_5(sme_ld1b_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1b_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1b_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1b_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_5(sme_ld1h_be_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1h_le_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1h_be_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1h_le_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1h_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1h_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1h_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1h_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_5(sme_ld1s_be_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1s_le_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1s_be_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1s_le_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1s_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1s_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1s_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1s_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_5(sme_ld1d_be_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1d_le_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1d_be_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1d_le_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1d_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1d_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1d_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1d_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_5(sme_ld1q_be_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1q_le_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1q_be_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1q_le_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1q_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1q_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1q_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1q_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_5(sme_st1b_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1b_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1b_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1b_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_5(sme_st1h_be_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1h_le_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1h_be_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1h_le_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1h_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1h_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1h_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1h_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_5(sme_st1s_be_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1s_le_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1s_be_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1s_le_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1s_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1s_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1s_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1s_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_5(sme_st1d_be_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1d_le_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1d_be_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1d_le_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1d_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1d_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1d_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1d_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_5(sme_st1q_be_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1q_le_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1q_be_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1q_le_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1q_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1q_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1q_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1q_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) diff --git a/target/arm/sme.decode b/target/arm/sme.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme.decode +++ b/target/arm/sme.decode @@ -XXX,XX +XXX,XX @@ MOVA 11000000 esz:2 00001 0 v:1 .. pg:3 0 za_imm:4 zr:5 \ &mova to_vec=1 rs=%mova_rs MOVA 11000000 11 00001 1 v:1 .. pg:3 0 za_imm:4 zr:5 \ &mova to_vec=1 rs=%mova_rs esz=4 + +### SME Memory + +&ldst esz rs pg rn rm za_imm v:bool st:bool + +LDST1 1110000 0 esz:2 st:1 rm:5 v:1 .. pg:3 rn:5 0 za_imm:4 \ + &ldst rs=%mova_rs +LDST1 1110000 111 st:1 rm:5 v:1 .. pg:3 rn:5 0 za_imm:4 \ + &ldst esz=4 rs=%mova_rs diff --git a/target/arm/sme_helper.c b/target/arm/sme_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme_helper.c +++ b/target/arm/sme_helper.c @@ -XXX,XX +XXX,XX @@ #include "qemu/osdep.h" #include "cpu.h" +#include "internals.h" #include "tcg/tcg-gvec-desc.h" #include "exec/helper-proto.h" +#include "exec/cpu_ldst.h" +#include "exec/exec-all.h" #include "qemu/int128.h" #include "vec_internal.h" +#include "sve_ldst_internal.h" /* ResetSVEState */ void arm_reset_sve_state(CPUARMState *env) @@ -XXX,XX +XXX,XX @@ void HELPER(sme_mova_zc_q)(void *vd, void *za, void *vg, uint32_t desc) } #undef DO_MOVA_Z + +/* + * Clear elements in a tile slice comprising len bytes. + */ + +typedef void ClearFn(void *ptr, size_t off, size_t len); + +static void clear_horizontal(void *ptr, size_t off, size_t len) +{ + memset(ptr + off, 0, len); +} + +static void clear_vertical_b(void *vptr, size_t off, size_t len) +{ + for (size_t i = 0; i < len; ++i) { + *(uint8_t *)(vptr + tile_vslice_offset(i + off)) = 0; + } +} + +static void clear_vertical_h(void *vptr, size_t off, size_t len) +{ + for (size_t i = 0; i < len; i += 2) { + *(uint16_t *)(vptr + tile_vslice_offset(i + off)) = 0; + } +} + +static void clear_vertical_s(void *vptr, size_t off, size_t len) +{ + for (size_t i = 0; i < len; i += 4) { + *(uint32_t *)(vptr + tile_vslice_offset(i + off)) = 0; + } +} + +static void clear_vertical_d(void *vptr, size_t off, size_t len) +{ + for (size_t i = 0; i < len; i += 8) { + *(uint64_t *)(vptr + tile_vslice_offset(i + off)) = 0; + } +} + +static void clear_vertical_q(void *vptr, size_t off, size_t len) +{ + for (size_t i = 0; i < len; i += 16) { + memset(vptr + tile_vslice_offset(i + off), 0, 16); + } +} + +/* + * Copy elements from an array into a tile slice comprising len bytes. + */ + +typedef void CopyFn(void *dst, const void *src, size_t len); + +static void copy_horizontal(void *dst, const void *src, size_t len) +{ + memcpy(dst, src, len); +} + +static void copy_vertical_b(void *vdst, const void *vsrc, size_t len) +{ + const uint8_t *src = vsrc; + uint8_t *dst = vdst; + size_t i; + + for (i = 0; i < len; ++i) { + dst[tile_vslice_index(i)] = src[i]; + } +} + +static void copy_vertical_h(void *vdst, const void *vsrc, size_t len) +{ + const uint16_t *src = vsrc; + uint16_t *dst = vdst; + size_t i; + + for (i = 0; i < len / 2; ++i) { + dst[tile_vslice_index(i)] = src[i]; + } +} + +static void copy_vertical_s(void *vdst, const void *vsrc, size_t len) +{ + const uint32_t *src = vsrc; + uint32_t *dst = vdst; + size_t i; + + for (i = 0; i < len / 4; ++i) { + dst[tile_vslice_index(i)] = src[i]; + } +} + +static void copy_vertical_d(void *vdst, const void *vsrc, size_t len) +{ + const uint64_t *src = vsrc; + uint64_t *dst = vdst; + size_t i; + + for (i = 0; i < len / 8; ++i) { + dst[tile_vslice_index(i)] = src[i]; + } +} + +static void copy_vertical_q(void *vdst, const void *vsrc, size_t len) +{ + for (size_t i = 0; i < len; i += 16) { + memcpy(vdst + tile_vslice_offset(i), vsrc + i, 16); + } +} + +/* + * Host and TLB primitives for vertical tile slice addressing. + */ + +#define DO_LD(NAME, TYPE, HOST, TLB) \ +static inline void sme_##NAME##_v_host(void *za, intptr_t off, void *host) \ +{ \ + TYPE val = HOST(host); \ + *(TYPE *)(za + tile_vslice_offset(off)) = val; \ +} \ +static inline void sme_##NAME##_v_tlb(CPUARMState *env, void *za, \ + intptr_t off, target_ulong addr, uintptr_t ra) \ +{ \ + TYPE val = TLB(env, useronly_clean_ptr(addr), ra); \ + *(TYPE *)(za + tile_vslice_offset(off)) = val; \ +} + +#define DO_ST(NAME, TYPE, HOST, TLB) \ +static inline void sme_##NAME##_v_host(void *za, intptr_t off, void *host) \ +{ \ + TYPE val = *(TYPE *)(za + tile_vslice_offset(off)); \ + HOST(host, val); \ +} \ +static inline void sme_##NAME##_v_tlb(CPUARMState *env, void *za, \ + intptr_t off, target_ulong addr, uintptr_t ra) \ +{ \ + TYPE val = *(TYPE *)(za + tile_vslice_offset(off)); \ + TLB(env, useronly_clean_ptr(addr), val, ra); \ +} + +/* + * The ARMVectorReg elements are stored in host-endian 64-bit units. + * For 128-bit quantities, the sequence defined by the Elem[] pseudocode + * corresponds to storing the two 64-bit pieces in little-endian order. + */ +#define DO_LDQ(HNAME, VNAME, BE, HOST, TLB) \ +static inline void HNAME##_host(void *za, intptr_t off, void *host) \ +{ \ + uint64_t val0 = HOST(host), val1 = HOST(host + 8); \ + uint64_t *ptr = za + off; \ + ptr[0] = BE ? val1 : val0, ptr[1] = BE ? val0 : val1; \ +} \ +static inline void VNAME##_v_host(void *za, intptr_t off, void *host) \ +{ \ + HNAME##_host(za, tile_vslice_offset(off), host); \ +} \ +static inline void HNAME##_tlb(CPUARMState *env, void *za, intptr_t off, \ + target_ulong addr, uintptr_t ra) \ +{ \ + uint64_t val0 = TLB(env, useronly_clean_ptr(addr), ra); \ + uint64_t val1 = TLB(env, useronly_clean_ptr(addr + 8), ra); \ + uint64_t *ptr = za + off; \ + ptr[0] = BE ? val1 : val0, ptr[1] = BE ? val0 : val1; \ +} \ +static inline void VNAME##_v_tlb(CPUARMState *env, void *za, intptr_t off, \ + target_ulong addr, uintptr_t ra) \ +{ \ + HNAME##_tlb(env, za, tile_vslice_offset(off), addr, ra); \ +} + +#define DO_STQ(HNAME, VNAME, BE, HOST, TLB) \ +static inline void HNAME##_host(void *za, intptr_t off, void *host) \ +{ \ + uint64_t *ptr = za + off; \ + HOST(host, ptr[BE]); \ + HOST(host + 1, ptr[!BE]); \ +} \ +static inline void VNAME##_v_host(void *za, intptr_t off, void *host) \ +{ \ + HNAME##_host(za, tile_vslice_offset(off), host); \ +} \ +static inline void HNAME##_tlb(CPUARMState *env, void *za, intptr_t off, \ + target_ulong addr, uintptr_t ra) \ +{ \ + uint64_t *ptr = za + off; \ + TLB(env, useronly_clean_ptr(addr), ptr[BE], ra); \ + TLB(env, useronly_clean_ptr(addr + 8), ptr[!BE], ra); \ +} \ +static inline void VNAME##_v_tlb(CPUARMState *env, void *za, intptr_t off, \ + target_ulong addr, uintptr_t ra) \ +{ \ + HNAME##_tlb(env, za, tile_vslice_offset(off), addr, ra); \ +} + +DO_LD(ld1b, uint8_t, ldub_p, cpu_ldub_data_ra) +DO_LD(ld1h_be, uint16_t, lduw_be_p, cpu_lduw_be_data_ra) +DO_LD(ld1h_le, uint16_t, lduw_le_p, cpu_lduw_le_data_ra) +DO_LD(ld1s_be, uint32_t, ldl_be_p, cpu_ldl_be_data_ra) +DO_LD(ld1s_le, uint32_t, ldl_le_p, cpu_ldl_le_data_ra) +DO_LD(ld1d_be, uint64_t, ldq_be_p, cpu_ldq_be_data_ra) +DO_LD(ld1d_le, uint64_t, ldq_le_p, cpu_ldq_le_data_ra) + +DO_LDQ(sve_ld1qq_be, sme_ld1q_be, 1, ldq_be_p, cpu_ldq_be_data_ra) +DO_LDQ(sve_ld1qq_le, sme_ld1q_le, 0, ldq_le_p, cpu_ldq_le_data_ra) + +DO_ST(st1b, uint8_t, stb_p, cpu_stb_data_ra) +DO_ST(st1h_be, uint16_t, stw_be_p, cpu_stw_be_data_ra) +DO_ST(st1h_le, uint16_t, stw_le_p, cpu_stw_le_data_ra) +DO_ST(st1s_be, uint32_t, stl_be_p, cpu_stl_be_data_ra) +DO_ST(st1s_le, uint32_t, stl_le_p, cpu_stl_le_data_ra) +DO_ST(st1d_be, uint64_t, stq_be_p, cpu_stq_be_data_ra) +DO_ST(st1d_le, uint64_t, stq_le_p, cpu_stq_le_data_ra) + +DO_STQ(sve_st1qq_be, sme_st1q_be, 1, stq_be_p, cpu_stq_be_data_ra) +DO_STQ(sve_st1qq_le, sme_st1q_le, 0, stq_le_p, cpu_stq_le_data_ra) + +#undef DO_LD +#undef DO_ST +#undef DO_LDQ +#undef DO_STQ + +/* + * Common helper for all contiguous predicated loads. + */ + +static inline QEMU_ALWAYS_INLINE +void sme_ld1(CPUARMState *env, void *za, uint64_t *vg, + const target_ulong addr, uint32_t desc, const uintptr_t ra, + const int esz, uint32_t mtedesc, bool vertical, + sve_ldst1_host_fn *host_fn, + sve_ldst1_tlb_fn *tlb_fn, + ClearFn *clr_fn, + CopyFn *cpy_fn) +{ + const intptr_t reg_max = simd_oprsz(desc); + const intptr_t esize = 1 << esz; + intptr_t reg_off, reg_last; + SVEContLdSt info; + void *host; + int flags; + + /* Find the active elements. */ + if (!sve_cont_ldst_elements(&info, addr, vg, reg_max, esz, esize)) { + /* The entire predicate was false; no load occurs. */ + clr_fn(za, 0, reg_max); + return; + } + + /* Probe the page(s). Exit with exception for any invalid page. */ + sve_cont_ldst_pages(&info, FAULT_ALL, env, addr, MMU_DATA_LOAD, ra); + + /* Handle watchpoints for all active elements. */ + sve_cont_ldst_watchpoints(&info, env, vg, addr, esize, esize, + BP_MEM_READ, ra); + + /* + * Handle mte checks for all active elements. + * Since TBI must be set for MTE, !mtedesc => !mte_active. + */ + if (mtedesc) { + sve_cont_ldst_mte_check(&info, env, vg, addr, esize, esize, + mtedesc, ra); + } + + flags = info.page[0].flags | info.page[1].flags; + if (unlikely(flags != 0)) { +#ifdef CONFIG_USER_ONLY + g_assert_not_reached(); +#else + /* + * At least one page includes MMIO. + * Any bus operation can fail with cpu_transaction_failed, + * which for ARM will raise SyncExternal. Perform the load + * into scratch memory to preserve register state until the end. + */ + ARMVectorReg scratch = { }; + + reg_off = info.reg_off_first[0]; + reg_last = info.reg_off_last[1]; + if (reg_last < 0) { + reg_last = info.reg_off_split; + if (reg_last < 0) { + reg_last = info.reg_off_last[0]; + } + } + + do { + uint64_t pg = vg[reg_off >> 6]; + do { + if ((pg >> (reg_off & 63)) & 1) { + tlb_fn(env, &scratch, reg_off, addr + reg_off, ra); + } + reg_off += esize; + } while (reg_off & 63); + } while (reg_off <= reg_last); + + cpy_fn(za, &scratch, reg_max); + return; +#endif + } + + /* The entire operation is in RAM, on valid pages. */ + + reg_off = info.reg_off_first[0]; + reg_last = info.reg_off_last[0]; + host = info.page[0].host; + + if (!vertical) { + memset(za, 0, reg_max); + } else if (reg_off) { + clr_fn(za, 0, reg_off); + } + + while (reg_off <= reg_last) { + uint64_t pg = vg[reg_off >> 6]; + do { + if ((pg >> (reg_off & 63)) & 1) { + host_fn(za, reg_off, host + reg_off); + } else if (vertical) { + clr_fn(za, reg_off, esize); + } + reg_off += esize; + } while (reg_off <= reg_last && (reg_off & 63)); + } + + /* + * Use the slow path to manage the cross-page misalignment. + * But we know this is RAM and cannot trap. + */ + reg_off = info.reg_off_split; + if (unlikely(reg_off >= 0)) { + tlb_fn(env, za, reg_off, addr + reg_off, ra); + } + + reg_off = info.reg_off_first[1]; + if (unlikely(reg_off >= 0)) { + reg_last = info.reg_off_last[1]; + host = info.page[1].host; + + do { + uint64_t pg = vg[reg_off >> 6]; + do { + if ((pg >> (reg_off & 63)) & 1) { + host_fn(za, reg_off, host + reg_off); + } else if (vertical) { + clr_fn(za, reg_off, esize); + } + reg_off += esize; + } while (reg_off & 63); + } while (reg_off <= reg_last); + } +} + +static inline QEMU_ALWAYS_INLINE +void sme_ld1_mte(CPUARMState *env, void *za, uint64_t *vg, + target_ulong addr, uint32_t desc, uintptr_t ra, + const int esz, bool vertical, + sve_ldst1_host_fn *host_fn, + sve_ldst1_tlb_fn *tlb_fn, + ClearFn *clr_fn, + CopyFn *cpy_fn) +{ + uint32_t mtedesc = desc >> (SIMD_DATA_SHIFT + SVE_MTEDESC_SHIFT); + int bit55 = extract64(addr, 55, 1); + + /* Remove mtedesc from the normal sve descriptor. */ + desc = extract32(desc, 0, SIMD_DATA_SHIFT + SVE_MTEDESC_SHIFT); + + /* Perform gross MTE suppression early. */ + if (!tbi_check(desc, bit55) || + tcma_check(desc, bit55, allocation_tag_from_addr(addr))) { + mtedesc = 0; + } + + sme_ld1(env, za, vg, addr, desc, ra, esz, mtedesc, vertical, + host_fn, tlb_fn, clr_fn, cpy_fn); +} + +#define DO_LD(L, END, ESZ) \ +void HELPER(sme_ld1##L##END##_h)(CPUARMState *env, void *za, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + sme_ld1(env, za, vg, addr, desc, GETPC(), ESZ, 0, false, \ + sve_ld1##L##L##END##_host, sve_ld1##L##L##END##_tlb, \ + clear_horizontal, copy_horizontal); \ +} \ +void HELPER(sme_ld1##L##END##_v)(CPUARMState *env, void *za, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + sme_ld1(env, za, vg, addr, desc, GETPC(), ESZ, 0, true, \ + sme_ld1##L##END##_v_host, sme_ld1##L##END##_v_tlb, \ + clear_vertical_##L, copy_vertical_##L); \ +} \ +void HELPER(sme_ld1##L##END##_h_mte)(CPUARMState *env, void *za, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + sme_ld1_mte(env, za, vg, addr, desc, GETPC(), ESZ, false, \ + sve_ld1##L##L##END##_host, sve_ld1##L##L##END##_tlb, \ + clear_horizontal, copy_horizontal); \ +} \ +void HELPER(sme_ld1##L##END##_v_mte)(CPUARMState *env, void *za, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + sme_ld1_mte(env, za, vg, addr, desc, GETPC(), ESZ, true, \ + sme_ld1##L##END##_v_host, sme_ld1##L##END##_v_tlb, \ + clear_vertical_##L, copy_vertical_##L); \ +} + +DO_LD(b, , MO_8) +DO_LD(h, _be, MO_16) +DO_LD(h, _le, MO_16) +DO_LD(s, _be, MO_32) +DO_LD(s, _le, MO_32) +DO_LD(d, _be, MO_64) +DO_LD(d, _le, MO_64) +DO_LD(q, _be, MO_128) +DO_LD(q, _le, MO_128) + +#undef DO_LD + +/* + * Common helper for all contiguous predicated stores. + */ + +static inline QEMU_ALWAYS_INLINE +void sme_st1(CPUARMState *env, void *za, uint64_t *vg, + const target_ulong addr, uint32_t desc, const uintptr_t ra, + const int esz, uint32_t mtedesc, bool vertical, + sve_ldst1_host_fn *host_fn, + sve_ldst1_tlb_fn *tlb_fn) +{ + const intptr_t reg_max = simd_oprsz(desc); + const intptr_t esize = 1 << esz; + intptr_t reg_off, reg_last; + SVEContLdSt info; + void *host; + int flags; + + /* Find the active elements. */ + if (!sve_cont_ldst_elements(&info, addr, vg, reg_max, esz, esize)) { + /* The entire predicate was false; no store occurs. */ + return; + } + + /* Probe the page(s). Exit with exception for any invalid page. */ + sve_cont_ldst_pages(&info, FAULT_ALL, env, addr, MMU_DATA_STORE, ra); + + /* Handle watchpoints for all active elements. */ + sve_cont_ldst_watchpoints(&info, env, vg, addr, esize, esize, + BP_MEM_WRITE, ra); + + /* + * Handle mte checks for all active elements. + * Since TBI must be set for MTE, !mtedesc => !mte_active. + */ + if (mtedesc) { + sve_cont_ldst_mte_check(&info, env, vg, addr, esize, esize, + mtedesc, ra); + } + + flags = info.page[0].flags | info.page[1].flags; + if (unlikely(flags != 0)) { +#ifdef CONFIG_USER_ONLY + g_assert_not_reached(); +#else + /* + * At least one page includes MMIO. + * Any bus operation can fail with cpu_transaction_failed, + * which for ARM will raise SyncExternal. We cannot avoid + * this fault and will leave with the store incomplete. + */ + reg_off = info.reg_off_first[0]; + reg_last = info.reg_off_last[1]; + if (reg_last < 0) { + reg_last = info.reg_off_split; + if (reg_last < 0) { + reg_last = info.reg_off_last[0]; + } + } + + do { + uint64_t pg = vg[reg_off >> 6]; + do { + if ((pg >> (reg_off & 63)) & 1) { + tlb_fn(env, za, reg_off, addr + reg_off, ra); + } + reg_off += esize; + } while (reg_off & 63); + } while (reg_off <= reg_last); + return; +#endif + } + + reg_off = info.reg_off_first[0]; + reg_last = info.reg_off_last[0]; + host = info.page[0].host; + + while (reg_off <= reg_last) { + uint64_t pg = vg[reg_off >> 6]; + do { + if ((pg >> (reg_off & 63)) & 1) { + host_fn(za, reg_off, host + reg_off); + } + reg_off += 1 << esz; + } while (reg_off <= reg_last && (reg_off & 63)); + } + + /* + * Use the slow path to manage the cross-page misalignment. + * But we know this is RAM and cannot trap. + */ + reg_off = info.reg_off_split; + if (unlikely(reg_off >= 0)) { + tlb_fn(env, za, reg_off, addr + reg_off, ra); + } + + reg_off = info.reg_off_first[1]; + if (unlikely(reg_off >= 0)) { + reg_last = info.reg_off_last[1]; + host = info.page[1].host; + + do { + uint64_t pg = vg[reg_off >> 6]; + do { + if ((pg >> (reg_off & 63)) & 1) { + host_fn(za, reg_off, host + reg_off); + } + reg_off += 1 << esz; + } while (reg_off & 63); + } while (reg_off <= reg_last); + } +} + +static inline QEMU_ALWAYS_INLINE +void sme_st1_mte(CPUARMState *env, void *za, uint64_t *vg, target_ulong addr, + uint32_t desc, uintptr_t ra, int esz, bool vertical, + sve_ldst1_host_fn *host_fn, + sve_ldst1_tlb_fn *tlb_fn) +{ + uint32_t mtedesc = desc >> (SIMD_DATA_SHIFT + SVE_MTEDESC_SHIFT); + int bit55 = extract64(addr, 55, 1); + + /* Remove mtedesc from the normal sve descriptor. */ + desc = extract32(desc, 0, SIMD_DATA_SHIFT + SVE_MTEDESC_SHIFT); + + /* Perform gross MTE suppression early. */ + if (!tbi_check(desc, bit55) || + tcma_check(desc, bit55, allocation_tag_from_addr(addr))) { + mtedesc = 0; + } + + sme_st1(env, za, vg, addr, desc, ra, esz, mtedesc, + vertical, host_fn, tlb_fn); +} + +#define DO_ST(L, END, ESZ) \ +void HELPER(sme_st1##L##END##_h)(CPUARMState *env, void *za, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + sme_st1(env, za, vg, addr, desc, GETPC(), ESZ, 0, false, \ + sve_st1##L##L##END##_host, sve_st1##L##L##END##_tlb); \ +} \ +void HELPER(sme_st1##L##END##_v)(CPUARMState *env, void *za, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + sme_st1(env, za, vg, addr, desc, GETPC(), ESZ, 0, true, \ + sme_st1##L##END##_v_host, sme_st1##L##END##_v_tlb); \ +} \ +void HELPER(sme_st1##L##END##_h_mte)(CPUARMState *env, void *za, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + sme_st1_mte(env, za, vg, addr, desc, GETPC(), ESZ, false, \ + sve_st1##L##L##END##_host, sve_st1##L##L##END##_tlb); \ +} \ +void HELPER(sme_st1##L##END##_v_mte)(CPUARMState *env, void *za, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + sme_st1_mte(env, za, vg, addr, desc, GETPC(), ESZ, true, \ + sme_st1##L##END##_v_host, sme_st1##L##END##_v_tlb); \ +} + +DO_ST(b, , MO_8) +DO_ST(h, _be, MO_16) +DO_ST(h, _le, MO_16) +DO_ST(s, _be, MO_32) +DO_ST(s, _le, MO_32) +DO_ST(d, _be, MO_64) +DO_ST(d, _le, MO_64) +DO_ST(q, _be, MO_128) +DO_ST(q, _le, MO_128) + +#undef DO_ST diff --git a/target/arm/translate-sme.c b/target/arm/translate-sme.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sme.c +++ b/target/arm/translate-sme.c @@ -XXX,XX +XXX,XX @@ static bool trans_MOVA(DisasContext *s, arg_MOVA *a) return true; } + +static bool trans_LDST1(DisasContext *s, arg_LDST1 *a) +{ + typedef void GenLdSt1(TCGv_env, TCGv_ptr, TCGv_ptr, TCGv, TCGv_i32); + + /* + * Indexed by [esz][be][v][mte][st], which is (except for load/store) + * also the order in which the elements appear in the function names, + * and so how we must concatenate the pieces. + */ + +#define FN_LS(F) { gen_helper_sme_ld1##F, gen_helper_sme_st1##F } +#define FN_MTE(F) { FN_LS(F), FN_LS(F##_mte) } +#define FN_HV(F) { FN_MTE(F##_h), FN_MTE(F##_v) } +#define FN_END(L, B) { FN_HV(L), FN_HV(B) } + + static GenLdSt1 * const fns[5][2][2][2][2] = { + FN_END(b, b), + FN_END(h_le, h_be), + FN_END(s_le, s_be), + FN_END(d_le, d_be), + FN_END(q_le, q_be), + }; + +#undef FN_LS +#undef FN_MTE +#undef FN_HV +#undef FN_END + + TCGv_ptr t_za, t_pg; + TCGv_i64 addr; + int svl, desc = 0; + bool be = s->be_data == MO_BE; + bool mte = s->mte_active[0]; + + if (!dc_isar_feature(aa64_sme, s)) { + return false; + } + if (!sme_smza_enabled_check(s)) { + return true; + } + + t_za = get_tile_rowcol(s, a->esz, a->rs, a->za_imm, a->v); + t_pg = pred_full_reg_ptr(s, a->pg); + addr = tcg_temp_new_i64(); + + tcg_gen_shli_i64(addr, cpu_reg(s, a->rm), a->esz); + tcg_gen_add_i64(addr, addr, cpu_reg_sp(s, a->rn)); + + if (mte) { + desc = FIELD_DP32(desc, MTEDESC, MIDX, get_mem_index(s)); + desc = FIELD_DP32(desc, MTEDESC, TBI, s->tbid); + desc = FIELD_DP32(desc, MTEDESC, TCMA, s->tcma); + desc = FIELD_DP32(desc, MTEDESC, WRITE, a->st); + desc = FIELD_DP32(desc, MTEDESC, SIZEM1, (1 << a->esz) - 1); + desc <<= SVE_MTEDESC_SHIFT; + } else { + addr = clean_data_tbi(s, addr); + } + svl = streaming_vec_reg_size(s); + desc = simd_desc(svl, svl, desc); + + fns[a->esz][be][a->v][mte][a->st](cpu_env, t_za, t_pg, addr, + tcg_constant_i32(desc)); + + tcg_temp_free_ptr(t_za); + tcg_temp_free_ptr(t_pg); + tcg_temp_free_i64(addr); + return true; +} -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Add a TCGv_ptr base argument, which will be cpu_env for SVE. We will reuse this for SME save and restore array insns. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-22-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-a64.h | 3 +++ target/arm/translate-sve.c | 48 ++++++++++++++++++++++++++++---------- 2 files changed, 39 insertions(+), 12 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 @@ void gen_gvec_xar(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs, uint32_t rm_ofs, int64_t shift, uint32_t opr_sz, uint32_t max_sz); +void gen_sve_ldr(DisasContext *s, TCGv_ptr, int vofs, int len, int rn, int imm); +void gen_sve_str(DisasContext *s, TCGv_ptr, int vofs, int len, int rn, int imm); + #endif /* TARGET_ARM_TRANSLATE_A64_H */ diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ TRANS_FEAT(UCVTF_dd, aa64_sve, gen_gvec_fpst_arg_zpz, * The load should begin at the address Rn + IMM. */ -static void do_ldr(DisasContext *s, uint32_t vofs, int len, int rn, int imm) +void gen_sve_ldr(DisasContext *s, TCGv_ptr base, int vofs, + int len, int rn, int imm) { int len_align = QEMU_ALIGN_DOWN(len, 8); int len_remain = len % 8; @@ -XXX,XX +XXX,XX @@ static void do_ldr(DisasContext *s, uint32_t vofs, int len, int rn, int imm) t0 = tcg_temp_new_i64(); for (i = 0; i < len_align; i += 8) { tcg_gen_qemu_ld_i64(t0, clean_addr, midx, MO_LEUQ); - tcg_gen_st_i64(t0, cpu_env, vofs + i); + tcg_gen_st_i64(t0, base, vofs + i); tcg_gen_addi_i64(clean_addr, clean_addr, 8); } tcg_temp_free_i64(t0); @@ -XXX,XX +XXX,XX @@ static void do_ldr(DisasContext *s, uint32_t vofs, int len, int rn, int imm) clean_addr = new_tmp_a64_local(s); tcg_gen_mov_i64(clean_addr, t0); + if (base != cpu_env) { + TCGv_ptr b = tcg_temp_local_new_ptr(); + tcg_gen_mov_ptr(b, base); + base = b; + } + gen_set_label(loop); t0 = tcg_temp_new_i64(); @@ -XXX,XX +XXX,XX @@ static void do_ldr(DisasContext *s, uint32_t vofs, int len, int rn, int imm) tcg_gen_addi_i64(clean_addr, clean_addr, 8); tp = tcg_temp_new_ptr(); - tcg_gen_add_ptr(tp, cpu_env, i); + tcg_gen_add_ptr(tp, base, i); tcg_gen_addi_ptr(i, i, 8); tcg_gen_st_i64(t0, tp, vofs); tcg_temp_free_ptr(tp); @@ -XXX,XX +XXX,XX @@ static void do_ldr(DisasContext *s, uint32_t vofs, int len, int rn, int imm) tcg_gen_brcondi_ptr(TCG_COND_LTU, i, len_align, loop); tcg_temp_free_ptr(i); + + if (base != cpu_env) { + tcg_temp_free_ptr(base); + assert(len_remain == 0); + } } /* @@ -XXX,XX +XXX,XX @@ static void do_ldr(DisasContext *s, uint32_t vofs, int len, int rn, int imm) default: g_assert_not_reached(); } - tcg_gen_st_i64(t0, cpu_env, vofs + len_align); + tcg_gen_st_i64(t0, base, vofs + len_align); tcg_temp_free_i64(t0); } } /* Similarly for stores. */ -static void do_str(DisasContext *s, uint32_t vofs, int len, int rn, int imm) +void gen_sve_str(DisasContext *s, TCGv_ptr base, int vofs, + int len, int rn, int imm) { int len_align = QEMU_ALIGN_DOWN(len, 8); int len_remain = len % 8; @@ -XXX,XX +XXX,XX @@ static void do_str(DisasContext *s, uint32_t vofs, int len, int rn, int imm) t0 = tcg_temp_new_i64(); for (i = 0; i < len_align; i += 8) { - tcg_gen_ld_i64(t0, cpu_env, vofs + i); + tcg_gen_ld_i64(t0, base, vofs + i); tcg_gen_qemu_st_i64(t0, clean_addr, midx, MO_LEUQ); tcg_gen_addi_i64(clean_addr, clean_addr, 8); } @@ -XXX,XX +XXX,XX @@ static void do_str(DisasContext *s, uint32_t vofs, int len, int rn, int imm) clean_addr = new_tmp_a64_local(s); tcg_gen_mov_i64(clean_addr, t0); + if (base != cpu_env) { + TCGv_ptr b = tcg_temp_local_new_ptr(); + tcg_gen_mov_ptr(b, base); + base = b; + } + gen_set_label(loop); t0 = tcg_temp_new_i64(); tp = tcg_temp_new_ptr(); - tcg_gen_add_ptr(tp, cpu_env, i); + tcg_gen_add_ptr(tp, base, i); tcg_gen_ld_i64(t0, tp, vofs); tcg_gen_addi_ptr(i, i, 8); tcg_temp_free_ptr(tp); @@ -XXX,XX +XXX,XX @@ static void do_str(DisasContext *s, uint32_t vofs, int len, int rn, int imm) tcg_gen_brcondi_ptr(TCG_COND_LTU, i, len_align, loop); tcg_temp_free_ptr(i); + + if (base != cpu_env) { + tcg_temp_free_ptr(base); + assert(len_remain == 0); + } } /* Predicate register stores can be any multiple of 2. */ if (len_remain) { t0 = tcg_temp_new_i64(); - tcg_gen_ld_i64(t0, cpu_env, vofs + len_align); + tcg_gen_ld_i64(t0, base, vofs + len_align); switch (len_remain) { case 2: @@ -XXX,XX +XXX,XX @@ static bool trans_LDR_zri(DisasContext *s, arg_rri *a) if (sve_access_check(s)) { int size = vec_full_reg_size(s); int off = vec_full_reg_offset(s, a->rd); - do_ldr(s, off, size, a->rn, a->imm * size); + gen_sve_ldr(s, cpu_env, off, size, a->rn, a->imm * size); } return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_LDR_pri(DisasContext *s, arg_rri *a) if (sve_access_check(s)) { int size = pred_full_reg_size(s); int off = pred_full_reg_offset(s, a->rd); - do_ldr(s, off, size, a->rn, a->imm * size); + gen_sve_ldr(s, cpu_env, off, size, a->rn, a->imm * size); } return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_STR_zri(DisasContext *s, arg_rri *a) if (sve_access_check(s)) { int size = vec_full_reg_size(s); int off = vec_full_reg_offset(s, a->rd); - do_str(s, off, size, a->rn, a->imm * size); + gen_sve_str(s, cpu_env, off, size, a->rn, a->imm * size); } return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_STR_pri(DisasContext *s, arg_rri *a) if (sve_access_check(s)) { int size = pred_full_reg_size(s); int off = pred_full_reg_offset(s, a->rd); - do_str(s, off, size, a->rn, a->imm * size); + gen_sve_str(s, cpu_env, off, size, a->rn, a->imm * size); } return true; } -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> We can reuse the SVE functions for LDR and STR, passing in the base of the ZA vector and a zero offset. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-23-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/sme.decode | 7 +++++++ target/arm/translate-sme.c | 24 ++++++++++++++++++++++++ 2 files changed, 31 insertions(+) diff --git a/target/arm/sme.decode b/target/arm/sme.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme.decode +++ b/target/arm/sme.decode @@ -XXX,XX +XXX,XX @@ LDST1 1110000 0 esz:2 st:1 rm:5 v:1 .. pg:3 rn:5 0 za_imm:4 \ &ldst rs=%mova_rs LDST1 1110000 111 st:1 rm:5 v:1 .. pg:3 rn:5 0 za_imm:4 \ &ldst esz=4 rs=%mova_rs + +&ldstr rv rn imm +@ldstr ....... ... . ...... .. ... rn:5 . imm:4 \ + &ldstr rv=%mova_rs + +LDR 1110000 100 0 000000 .. 000 ..... 0 .... @ldstr +STR 1110000 100 1 000000 .. 000 ..... 0 .... @ldstr diff --git a/target/arm/translate-sme.c b/target/arm/translate-sme.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sme.c +++ b/target/arm/translate-sme.c @@ -XXX,XX +XXX,XX @@ static bool trans_LDST1(DisasContext *s, arg_LDST1 *a) tcg_temp_free_i64(addr); return true; } + +typedef void GenLdStR(DisasContext *, TCGv_ptr, int, int, int, int); + +static bool do_ldst_r(DisasContext *s, arg_ldstr *a, GenLdStR *fn) +{ + int svl = streaming_vec_reg_size(s); + int imm = a->imm; + TCGv_ptr base; + + if (!sme_za_enabled_check(s)) { + return true; + } + + /* ZA[n] equates to ZA0H.B[n]. */ + base = get_tile_rowcol(s, MO_8, a->rv, imm, false); + + fn(s, base, 0, svl, a->rn, imm * svl); + + tcg_temp_free_ptr(base); + return true; +} + +TRANS_FEAT(LDR, aa64_sme, do_ldst_r, a, gen_sve_ldr) +TRANS_FEAT(STR, aa64_sme, do_ldst_r, a, gen_sve_str) -- 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: 20220708151540.18136-24-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sme.h | 5 +++ target/arm/sme.decode | 11 +++++ target/arm/sme_helper.c | 90 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-sme.c | 31 +++++++++++++ 4 files changed, 137 insertions(+) diff --git a/target/arm/helper-sme.h b/target/arm/helper-sme.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sme.h +++ b/target/arm/helper-sme.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_5(sme_st1q_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i DEF_HELPER_FLAGS_5(sme_st1q_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) DEF_HELPER_FLAGS_5(sme_st1q_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) DEF_HELPER_FLAGS_5(sme_st1q_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_5(sme_addha_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sme_addva_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sme_addha_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sme_addva_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) diff --git a/target/arm/sme.decode b/target/arm/sme.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme.decode +++ b/target/arm/sme.decode @@ -XXX,XX +XXX,XX @@ LDST1 1110000 111 st:1 rm:5 v:1 .. pg:3 rn:5 0 za_imm:4 \ LDR 1110000 100 0 000000 .. 000 ..... 0 .... @ldstr STR 1110000 100 1 000000 .. 000 ..... 0 .... @ldstr + +### SME Add Vector to Array + +&adda zad zn pm pn +@adda_32 ........ .. ..... . pm:3 pn:3 zn:5 ... zad:2 &adda +@adda_64 ........ .. ..... . pm:3 pn:3 zn:5 .. zad:3 &adda + +ADDHA_s 11000000 10 01000 0 ... ... ..... 000 .. @adda_32 +ADDVA_s 11000000 10 01000 1 ... ... ..... 000 .. @adda_32 +ADDHA_d 11000000 11 01000 0 ... ... ..... 00 ... @adda_64 +ADDVA_d 11000000 11 01000 1 ... ... ..... 00 ... @adda_64 diff --git a/target/arm/sme_helper.c b/target/arm/sme_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme_helper.c +++ b/target/arm/sme_helper.c @@ -XXX,XX +XXX,XX @@ DO_ST(q, _be, MO_128) DO_ST(q, _le, MO_128) #undef DO_ST + +void HELPER(sme_addha_s)(void *vzda, void *vzn, void *vpn, + void *vpm, uint32_t desc) +{ + intptr_t row, col, oprsz = simd_oprsz(desc) / 4; + uint64_t *pn = vpn, *pm = vpm; + uint32_t *zda = vzda, *zn = vzn; + + for (row = 0; row < oprsz; ) { + uint64_t pa = pn[row >> 4]; + do { + if (pa & 1) { + for (col = 0; col < oprsz; ) { + uint64_t pb = pm[col >> 4]; + do { + if (pb & 1) { + zda[tile_vslice_index(row) + H4(col)] += zn[H4(col)]; + } + pb >>= 4; + } while (++col & 15); + } + } + pa >>= 4; + } while (++row & 15); + } +} + +void HELPER(sme_addha_d)(void *vzda, void *vzn, void *vpn, + void *vpm, uint32_t desc) +{ + intptr_t row, col, oprsz = simd_oprsz(desc) / 8; + uint8_t *pn = vpn, *pm = vpm; + uint64_t *zda = vzda, *zn = vzn; + + for (row = 0; row < oprsz; ++row) { + if (pn[H1(row)] & 1) { + for (col = 0; col < oprsz; ++col) { + if (pm[H1(col)] & 1) { + zda[tile_vslice_index(row) + col] += zn[col]; + } + } + } + } +} + +void HELPER(sme_addva_s)(void *vzda, void *vzn, void *vpn, + void *vpm, uint32_t desc) +{ + intptr_t row, col, oprsz = simd_oprsz(desc) / 4; + uint64_t *pn = vpn, *pm = vpm; + uint32_t *zda = vzda, *zn = vzn; + + for (row = 0; row < oprsz; ) { + uint64_t pa = pn[row >> 4]; + do { + if (pa & 1) { + uint32_t zn_row = zn[H4(row)]; + for (col = 0; col < oprsz; ) { + uint64_t pb = pm[col >> 4]; + do { + if (pb & 1) { + zda[tile_vslice_index(row) + H4(col)] += zn_row; + } + pb >>= 4; + } while (++col & 15); + } + } + pa >>= 4; + } while (++row & 15); + } +} + +void HELPER(sme_addva_d)(void *vzda, void *vzn, void *vpn, + void *vpm, uint32_t desc) +{ + intptr_t row, col, oprsz = simd_oprsz(desc) / 8; + uint8_t *pn = vpn, *pm = vpm; + uint64_t *zda = vzda, *zn = vzn; + + for (row = 0; row < oprsz; ++row) { + if (pn[H1(row)] & 1) { + uint64_t zn_row = zn[row]; + for (col = 0; col < oprsz; ++col) { + if (pm[H1(col)] & 1) { + zda[tile_vslice_index(row) + col] += zn_row; + } + } + } + } +} diff --git a/target/arm/translate-sme.c b/target/arm/translate-sme.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sme.c +++ b/target/arm/translate-sme.c @@ -XXX,XX +XXX,XX @@ static bool do_ldst_r(DisasContext *s, arg_ldstr *a, GenLdStR *fn) TRANS_FEAT(LDR, aa64_sme, do_ldst_r, a, gen_sve_ldr) TRANS_FEAT(STR, aa64_sme, do_ldst_r, a, gen_sve_str) + +static bool do_adda(DisasContext *s, arg_adda *a, MemOp esz, + gen_helper_gvec_4 *fn) +{ + int svl = streaming_vec_reg_size(s); + uint32_t desc = simd_desc(svl, svl, 0); + TCGv_ptr za, zn, pn, pm; + + if (!sme_smza_enabled_check(s)) { + return true; + } + + /* Sum XZR+zad to find ZAd. */ + za = get_tile_rowcol(s, esz, 31, a->zad, false); + zn = vec_full_reg_ptr(s, a->zn); + pn = pred_full_reg_ptr(s, a->pn); + pm = pred_full_reg_ptr(s, a->pm); + + fn(za, zn, pn, pm, tcg_constant_i32(desc)); + + tcg_temp_free_ptr(za); + tcg_temp_free_ptr(zn); + tcg_temp_free_ptr(pn); + tcg_temp_free_ptr(pm); + return true; +} + +TRANS_FEAT(ADDHA_s, aa64_sme, do_adda, a, MO_32, gen_helper_sme_addha_s) +TRANS_FEAT(ADDVA_s, aa64_sme, do_adda, a, MO_32, gen_helper_sme_addva_s) +TRANS_FEAT(ADDHA_d, aa64_sme_i16i64, do_adda, a, MO_64, gen_helper_sme_addha_d) +TRANS_FEAT(ADDVA_d, aa64_sme_i16i64, do_adda, a, MO_64, gen_helper_sme_addva_d) -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-25-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sme.h | 5 +++ target/arm/sme.decode | 9 +++++ target/arm/sme_helper.c | 69 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-sme.c | 32 ++++++++++++++++++ 4 files changed, 115 insertions(+) diff --git a/target/arm/helper-sme.h b/target/arm/helper-sme.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sme.h +++ b/target/arm/helper-sme.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_5(sme_addha_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sme_addva_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sme_addha_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sme_addva_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_7(sme_fmopa_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_7(sme_fmopa_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, ptr, i32) diff --git a/target/arm/sme.decode b/target/arm/sme.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme.decode +++ b/target/arm/sme.decode @@ -XXX,XX +XXX,XX @@ ADDHA_s 11000000 10 01000 0 ... ... ..... 000 .. @adda_32 ADDVA_s 11000000 10 01000 1 ... ... ..... 000 .. @adda_32 ADDHA_d 11000000 11 01000 0 ... ... ..... 00 ... @adda_64 ADDVA_d 11000000 11 01000 1 ... ... ..... 00 ... @adda_64 + +### SME Outer Product + +&op zad zn zm pm pn sub:bool +@op_32 ........ ... zm:5 pm:3 pn:3 zn:5 sub:1 .. zad:2 &op +@op_64 ........ ... zm:5 pm:3 pn:3 zn:5 sub:1 . zad:3 &op + +FMOPA_s 10000000 100 ..... ... ... ..... . 00 .. @op_32 +FMOPA_d 10000000 110 ..... ... ... ..... . 0 ... @op_64 diff --git a/target/arm/sme_helper.c b/target/arm/sme_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme_helper.c +++ b/target/arm/sme_helper.c @@ -XXX,XX +XXX,XX @@ #include "exec/cpu_ldst.h" #include "exec/exec-all.h" #include "qemu/int128.h" +#include "fpu/softfloat.h" #include "vec_internal.h" #include "sve_ldst_internal.h" @@ -XXX,XX +XXX,XX @@ void HELPER(sme_addva_d)(void *vzda, void *vzn, void *vpn, } } } + +void HELPER(sme_fmopa_s)(void *vza, void *vzn, void *vzm, void *vpn, + void *vpm, void *vst, uint32_t desc) +{ + intptr_t row, col, oprsz = simd_maxsz(desc); + uint32_t neg = simd_data(desc) << 31; + uint16_t *pn = vpn, *pm = vpm; + float_status fpst; + + /* + * Make a copy of float_status because this operation does not + * update the cumulative fp exception status. It also produces + * default nans. + */ + fpst = *(float_status *)vst; + set_default_nan_mode(true, &fpst); + + for (row = 0; row < oprsz; ) { + uint16_t pa = pn[H2(row >> 4)]; + do { + if (pa & 1) { + void *vza_row = vza + tile_vslice_offset(row); + uint32_t n = *(uint32_t *)(vzn + H1_4(row)) ^ neg; + + for (col = 0; col < oprsz; ) { + uint16_t pb = pm[H2(col >> 4)]; + do { + if (pb & 1) { + uint32_t *a = vza_row + H1_4(col); + uint32_t *m = vzm + H1_4(col); + *a = float32_muladd(n, *m, *a, 0, vst); + } + col += 4; + pb >>= 4; + } while (col & 15); + } + } + row += 4; + pa >>= 4; + } while (row & 15); + } +} + +void HELPER(sme_fmopa_d)(void *vza, void *vzn, void *vzm, void *vpn, + void *vpm, void *vst, uint32_t desc) +{ + intptr_t row, col, oprsz = simd_oprsz(desc) / 8; + uint64_t neg = (uint64_t)simd_data(desc) << 63; + uint64_t *za = vza, *zn = vzn, *zm = vzm; + uint8_t *pn = vpn, *pm = vpm; + float_status fpst = *(float_status *)vst; + + set_default_nan_mode(true, &fpst); + + for (row = 0; row < oprsz; ++row) { + if (pn[H1(row)] & 1) { + uint64_t *za_row = &za[tile_vslice_index(row)]; + uint64_t n = zn[row] ^ neg; + + for (col = 0; col < oprsz; ++col) { + if (pm[H1(col)] & 1) { + uint64_t *a = &za_row[col]; + *a = float64_muladd(n, zm[col], *a, 0, &fpst); + } + } + } + } +} diff --git a/target/arm/translate-sme.c b/target/arm/translate-sme.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sme.c +++ b/target/arm/translate-sme.c @@ -XXX,XX +XXX,XX @@ TRANS_FEAT(ADDHA_s, aa64_sme, do_adda, a, MO_32, gen_helper_sme_addha_s) TRANS_FEAT(ADDVA_s, aa64_sme, do_adda, a, MO_32, gen_helper_sme_addva_s) TRANS_FEAT(ADDHA_d, aa64_sme_i16i64, do_adda, a, MO_64, gen_helper_sme_addha_d) TRANS_FEAT(ADDVA_d, aa64_sme_i16i64, do_adda, a, MO_64, gen_helper_sme_addva_d) + +static bool do_outprod_fpst(DisasContext *s, arg_op *a, MemOp esz, + gen_helper_gvec_5_ptr *fn) +{ + int svl = streaming_vec_reg_size(s); + uint32_t desc = simd_desc(svl, svl, a->sub); + TCGv_ptr za, zn, zm, pn, pm, fpst; + + if (!sme_smza_enabled_check(s)) { + return true; + } + + /* Sum XZR+zad to find ZAd. */ + za = get_tile_rowcol(s, esz, 31, a->zad, false); + zn = vec_full_reg_ptr(s, a->zn); + zm = vec_full_reg_ptr(s, a->zm); + pn = pred_full_reg_ptr(s, a->pn); + pm = pred_full_reg_ptr(s, a->pm); + fpst = fpstatus_ptr(FPST_FPCR); + + fn(za, zn, zm, pn, pm, fpst, tcg_constant_i32(desc)); + + tcg_temp_free_ptr(za); + tcg_temp_free_ptr(zn); + tcg_temp_free_ptr(pn); + tcg_temp_free_ptr(pm); + tcg_temp_free_ptr(fpst); + return true; +} + +TRANS_FEAT(FMOPA_s, aa64_sme, do_outprod_fpst, a, MO_32, gen_helper_sme_fmopa_s) +TRANS_FEAT(FMOPA_d, aa64_sme_f64f64, do_outprod_fpst, a, MO_64, gen_helper_sme_fmopa_d) -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-26-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sme.h | 2 ++ target/arm/sme.decode | 2 ++ target/arm/sme_helper.c | 56 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-sme.c | 30 ++++++++++++++++++++ 4 files changed, 90 insertions(+) diff --git a/target/arm/helper-sme.h b/target/arm/helper-sme.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sme.h +++ b/target/arm/helper-sme.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_7(sme_fmopa_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_7(sme_fmopa_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sme_bfmopa, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) diff --git a/target/arm/sme.decode b/target/arm/sme.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme.decode +++ b/target/arm/sme.decode @@ -XXX,XX +XXX,XX @@ ADDVA_d 11000000 11 01000 1 ... ... ..... 00 ... @adda_64 FMOPA_s 10000000 100 ..... ... ... ..... . 00 .. @op_32 FMOPA_d 10000000 110 ..... ... ... ..... . 0 ... @op_64 + +BFMOPA 10000001 100 ..... ... ... ..... . 00 .. @op_32 diff --git a/target/arm/sme_helper.c b/target/arm/sme_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme_helper.c +++ b/target/arm/sme_helper.c @@ -XXX,XX +XXX,XX @@ void HELPER(sme_fmopa_d)(void *vza, void *vzn, void *vzm, void *vpn, } } } + +/* + * Alter PAIR as needed for controlling predicates being false, + * and for NEG on an enabled row element. + */ +static inline uint32_t f16mop_adj_pair(uint32_t pair, uint32_t pg, uint32_t neg) +{ + /* + * The pseudocode uses a conditional negate after the conditional zero. + * It is simpler here to unconditionally negate before conditional zero. + */ + pair ^= neg; + if (!(pg & 1)) { + pair &= 0xffff0000u; + } + if (!(pg & 4)) { + pair &= 0x0000ffffu; + } + return pair; +} + +void HELPER(sme_bfmopa)(void *vza, void *vzn, void *vzm, void *vpn, + void *vpm, uint32_t desc) +{ + intptr_t row, col, oprsz = simd_maxsz(desc); + uint32_t neg = simd_data(desc) * 0x80008000u; + uint16_t *pn = vpn, *pm = vpm; + + for (row = 0; row < oprsz; ) { + uint16_t prow = pn[H2(row >> 4)]; + do { + void *vza_row = vza + tile_vslice_offset(row); + uint32_t n = *(uint32_t *)(vzn + H1_4(row)); + + n = f16mop_adj_pair(n, prow, neg); + + for (col = 0; col < oprsz; ) { + uint16_t pcol = pm[H2(col >> 4)]; + do { + if (prow & pcol & 0b0101) { + uint32_t *a = vza_row + H1_4(col); + uint32_t m = *(uint32_t *)(vzm + H1_4(col)); + + m = f16mop_adj_pair(m, pcol, 0); + *a = bfdotadd(*a, n, m); + + col += 4; + pcol >>= 4; + } + } while (col & 15); + } + row += 4; + prow >>= 4; + } while (row & 15); + } +} diff --git a/target/arm/translate-sme.c b/target/arm/translate-sme.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sme.c +++ b/target/arm/translate-sme.c @@ -XXX,XX +XXX,XX @@ TRANS_FEAT(ADDVA_s, aa64_sme, do_adda, a, MO_32, gen_helper_sme_addva_s) TRANS_FEAT(ADDHA_d, aa64_sme_i16i64, do_adda, a, MO_64, gen_helper_sme_addha_d) TRANS_FEAT(ADDVA_d, aa64_sme_i16i64, do_adda, a, MO_64, gen_helper_sme_addva_d) +static bool do_outprod(DisasContext *s, arg_op *a, MemOp esz, + gen_helper_gvec_5 *fn) +{ + int svl = streaming_vec_reg_size(s); + uint32_t desc = simd_desc(svl, svl, a->sub); + TCGv_ptr za, zn, zm, pn, pm; + + if (!sme_smza_enabled_check(s)) { + return true; + } + + /* Sum XZR+zad to find ZAd. */ + za = get_tile_rowcol(s, esz, 31, a->zad, false); + zn = vec_full_reg_ptr(s, a->zn); + zm = vec_full_reg_ptr(s, a->zm); + pn = pred_full_reg_ptr(s, a->pn); + pm = pred_full_reg_ptr(s, a->pm); + + fn(za, zn, zm, pn, pm, tcg_constant_i32(desc)); + + tcg_temp_free_ptr(za); + tcg_temp_free_ptr(zn); + tcg_temp_free_ptr(pn); + tcg_temp_free_ptr(pm); + return true; +} + static bool do_outprod_fpst(DisasContext *s, arg_op *a, MemOp esz, gen_helper_gvec_5_ptr *fn) { @@ -XXX,XX +XXX,XX @@ static bool do_outprod_fpst(DisasContext *s, arg_op *a, MemOp esz, TRANS_FEAT(FMOPA_s, aa64_sme, do_outprod_fpst, a, MO_32, gen_helper_sme_fmopa_s) TRANS_FEAT(FMOPA_d, aa64_sme_f64f64, do_outprod_fpst, a, MO_64, gen_helper_sme_fmopa_d) + +/* TODO: FEAT_EBF16 */ +TRANS_FEAT(BFMOPA, aa64_sme, do_outprod, a, MO_32, gen_helper_sme_bfmopa) -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-27-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sme.h | 2 ++ target/arm/sme.decode | 1 + target/arm/sme_helper.c | 74 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-sme.c | 1 + 4 files changed, 78 insertions(+) diff --git a/target/arm/helper-sme.h b/target/arm/helper-sme.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sme.h +++ b/target/arm/helper-sme.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_5(sme_addva_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sme_addha_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sme_addva_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_7(sme_fmopa_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_7(sme_fmopa_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_7(sme_fmopa_d, TCG_CALL_NO_RWG, diff --git a/target/arm/sme.decode b/target/arm/sme.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme.decode +++ b/target/arm/sme.decode @@ -XXX,XX +XXX,XX @@ FMOPA_s 10000000 100 ..... ... ... ..... . 00 .. @op_32 FMOPA_d 10000000 110 ..... ... ... ..... . 0 ... @op_64 BFMOPA 10000001 100 ..... ... ... ..... . 00 .. @op_32 +FMOPA_h 10000001 101 ..... ... ... ..... . 00 .. @op_32 diff --git a/target/arm/sme_helper.c b/target/arm/sme_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme_helper.c +++ b/target/arm/sme_helper.c @@ -XXX,XX +XXX,XX @@ static inline uint32_t f16mop_adj_pair(uint32_t pair, uint32_t pg, uint32_t neg) return pair; } +static float32 f16_dotadd(float32 sum, uint32_t e1, uint32_t e2, + float_status *s_std, float_status *s_odd) +{ + float64 e1r = float16_to_float64(e1 & 0xffff, true, s_std); + float64 e1c = float16_to_float64(e1 >> 16, true, s_std); + float64 e2r = float16_to_float64(e2 & 0xffff, true, s_std); + float64 e2c = float16_to_float64(e2 >> 16, true, s_std); + float64 t64; + float32 t32; + + /* + * The ARM pseudocode function FPDot performs both multiplies + * and the add with a single rounding operation. Emulate this + * by performing the first multiply in round-to-odd, then doing + * the second multiply as fused multiply-add, and rounding to + * float32 all in one step. + */ + t64 = float64_mul(e1r, e2r, s_odd); + t64 = float64r32_muladd(e1c, e2c, t64, 0, s_std); + + /* This conversion is exact, because we've already rounded. */ + t32 = float64_to_float32(t64, s_std); + + /* The final accumulation step is not fused. */ + return float32_add(sum, t32, s_std); +} + +void HELPER(sme_fmopa_h)(void *vza, void *vzn, void *vzm, void *vpn, + void *vpm, void *vst, uint32_t desc) +{ + intptr_t row, col, oprsz = simd_maxsz(desc); + uint32_t neg = simd_data(desc) * 0x80008000u; + uint16_t *pn = vpn, *pm = vpm; + float_status fpst_odd, fpst_std; + + /* + * Make a copy of float_status because this operation does not + * update the cumulative fp exception status. It also produces + * default nans. Make a second copy with round-to-odd -- see above. + */ + fpst_std = *(float_status *)vst; + set_default_nan_mode(true, &fpst_std); + fpst_odd = fpst_std; + set_float_rounding_mode(float_round_to_odd, &fpst_odd); + + for (row = 0; row < oprsz; ) { + uint16_t prow = pn[H2(row >> 4)]; + do { + void *vza_row = vza + tile_vslice_offset(row); + uint32_t n = *(uint32_t *)(vzn + H1_4(row)); + + n = f16mop_adj_pair(n, prow, neg); + + for (col = 0; col < oprsz; ) { + uint16_t pcol = pm[H2(col >> 4)]; + do { + if (prow & pcol & 0b0101) { + uint32_t *a = vza_row + H1_4(col); + uint32_t m = *(uint32_t *)(vzm + H1_4(col)); + + m = f16mop_adj_pair(m, pcol, 0); + *a = f16_dotadd(*a, n, m, &fpst_std, &fpst_odd); + + col += 4; + pcol >>= 4; + } + } while (col & 15); + } + row += 4; + prow >>= 4; + } while (row & 15); + } +} + void HELPER(sme_bfmopa)(void *vza, void *vzn, void *vzm, void *vpn, void *vpm, uint32_t desc) { diff --git a/target/arm/translate-sme.c b/target/arm/translate-sme.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sme.c +++ b/target/arm/translate-sme.c @@ -XXX,XX +XXX,XX @@ static bool do_outprod_fpst(DisasContext *s, arg_op *a, MemOp esz, return true; } +TRANS_FEAT(FMOPA_h, aa64_sme, do_outprod_fpst, a, MO_32, gen_helper_sme_fmopa_h) TRANS_FEAT(FMOPA_s, aa64_sme, do_outprod_fpst, a, MO_32, gen_helper_sme_fmopa_s) TRANS_FEAT(FMOPA_d, aa64_sme_f64f64, do_outprod_fpst, a, MO_64, gen_helper_sme_fmopa_d) -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> This is SMOPA, SUMOPA, USMOPA_s, UMOPA, for both Int8 and Int16. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-28-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sme.h | 16 ++++++++ target/arm/sme.decode | 10 +++++ target/arm/sme_helper.c | 82 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-sme.c | 10 +++++ 4 files changed, 118 insertions(+) diff --git a/target/arm/helper-sme.h b/target/arm/helper-sme.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sme.h +++ b/target/arm/helper-sme.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_7(sme_fmopa_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_6(sme_bfmopa, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sme_smopa_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sme_umopa_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sme_sumopa_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sme_usmopa_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sme_smopa_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sme_umopa_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sme_sumopa_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sme_usmopa_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) diff --git a/target/arm/sme.decode b/target/arm/sme.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme.decode +++ b/target/arm/sme.decode @@ -XXX,XX +XXX,XX @@ FMOPA_d 10000000 110 ..... ... ... ..... . 0 ... @op_64 BFMOPA 10000001 100 ..... ... ... ..... . 00 .. @op_32 FMOPA_h 10000001 101 ..... ... ... ..... . 00 .. @op_32 + +SMOPA_s 1010000 0 10 0 ..... ... ... ..... . 00 .. @op_32 +SUMOPA_s 1010000 0 10 1 ..... ... ... ..... . 00 .. @op_32 +USMOPA_s 1010000 1 10 0 ..... ... ... ..... . 00 .. @op_32 +UMOPA_s 1010000 1 10 1 ..... ... ... ..... . 00 .. @op_32 + +SMOPA_d 1010000 0 11 0 ..... ... ... ..... . 0 ... @op_64 +SUMOPA_d 1010000 0 11 1 ..... ... ... ..... . 0 ... @op_64 +USMOPA_d 1010000 1 11 0 ..... ... ... ..... . 0 ... @op_64 +UMOPA_d 1010000 1 11 1 ..... ... ... ..... . 0 ... @op_64 diff --git a/target/arm/sme_helper.c b/target/arm/sme_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme_helper.c +++ b/target/arm/sme_helper.c @@ -XXX,XX +XXX,XX @@ void HELPER(sme_bfmopa)(void *vza, void *vzn, void *vzm, void *vpn, } while (row & 15); } } + +typedef uint64_t IMOPFn(uint64_t, uint64_t, uint64_t, uint8_t, bool); + +static inline void do_imopa(uint64_t *za, uint64_t *zn, uint64_t *zm, + uint8_t *pn, uint8_t *pm, + uint32_t desc, IMOPFn *fn) +{ + intptr_t row, col, oprsz = simd_oprsz(desc) / 8; + bool neg = simd_data(desc); + + for (row = 0; row < oprsz; ++row) { + uint8_t pa = pn[H1(row)]; + uint64_t *za_row = &za[tile_vslice_index(row)]; + uint64_t n = zn[row]; + + for (col = 0; col < oprsz; ++col) { + uint8_t pb = pm[H1(col)]; + uint64_t *a = &za_row[col]; + + *a = fn(n, zm[col], *a, pa & pb, neg); + } + } +} + +#define DEF_IMOP_32(NAME, NTYPE, MTYPE) \ +static uint64_t NAME(uint64_t n, uint64_t m, uint64_t a, uint8_t p, bool neg) \ +{ \ + uint32_t sum0 = 0, sum1 = 0; \ + /* Apply P to N as a mask, making the inactive elements 0. */ \ + n &= expand_pred_b(p); \ + sum0 += (NTYPE)(n >> 0) * (MTYPE)(m >> 0); \ + sum0 += (NTYPE)(n >> 8) * (MTYPE)(m >> 8); \ + sum0 += (NTYPE)(n >> 16) * (MTYPE)(m >> 16); \ + sum0 += (NTYPE)(n >> 24) * (MTYPE)(m >> 24); \ + sum1 += (NTYPE)(n >> 32) * (MTYPE)(m >> 32); \ + sum1 += (NTYPE)(n >> 40) * (MTYPE)(m >> 40); \ + sum1 += (NTYPE)(n >> 48) * (MTYPE)(m >> 48); \ + sum1 += (NTYPE)(n >> 56) * (MTYPE)(m >> 56); \ + if (neg) { \ + sum0 = (uint32_t)a - sum0, sum1 = (uint32_t)(a >> 32) - sum1; \ + } else { \ + sum0 = (uint32_t)a + sum0, sum1 = (uint32_t)(a >> 32) + sum1; \ + } \ + return ((uint64_t)sum1 << 32) | sum0; \ +} + +#define DEF_IMOP_64(NAME, NTYPE, MTYPE) \ +static uint64_t NAME(uint64_t n, uint64_t m, uint64_t a, uint8_t p, bool neg) \ +{ \ + uint64_t sum = 0; \ + /* Apply P to N as a mask, making the inactive elements 0. */ \ + n &= expand_pred_h(p); \ + sum += (NTYPE)(n >> 0) * (MTYPE)(m >> 0); \ + sum += (NTYPE)(n >> 16) * (MTYPE)(m >> 16); \ + sum += (NTYPE)(n >> 32) * (MTYPE)(m >> 32); \ + sum += (NTYPE)(n >> 48) * (MTYPE)(m >> 48); \ + return neg ? a - sum : a + sum; \ +} + +DEF_IMOP_32(smopa_s, int8_t, int8_t) +DEF_IMOP_32(umopa_s, uint8_t, uint8_t) +DEF_IMOP_32(sumopa_s, int8_t, uint8_t) +DEF_IMOP_32(usmopa_s, uint8_t, int8_t) + +DEF_IMOP_64(smopa_d, int16_t, int16_t) +DEF_IMOP_64(umopa_d, uint16_t, uint16_t) +DEF_IMOP_64(sumopa_d, int16_t, uint16_t) +DEF_IMOP_64(usmopa_d, uint16_t, int16_t) + +#define DEF_IMOPH(NAME) \ + void HELPER(sme_##NAME)(void *vza, void *vzn, void *vzm, void *vpn, \ + void *vpm, uint32_t desc) \ + { do_imopa(vza, vzn, vzm, vpn, vpm, desc, NAME); } + +DEF_IMOPH(smopa_s) +DEF_IMOPH(umopa_s) +DEF_IMOPH(sumopa_s) +DEF_IMOPH(usmopa_s) +DEF_IMOPH(smopa_d) +DEF_IMOPH(umopa_d) +DEF_IMOPH(sumopa_d) +DEF_IMOPH(usmopa_d) diff --git a/target/arm/translate-sme.c b/target/arm/translate-sme.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sme.c +++ b/target/arm/translate-sme.c @@ -XXX,XX +XXX,XX @@ TRANS_FEAT(FMOPA_d, aa64_sme_f64f64, do_outprod_fpst, a, MO_64, gen_helper_sme_f /* TODO: FEAT_EBF16 */ TRANS_FEAT(BFMOPA, aa64_sme, do_outprod, a, MO_32, gen_helper_sme_bfmopa) + +TRANS_FEAT(SMOPA_s, aa64_sme, do_outprod, a, MO_32, gen_helper_sme_smopa_s) +TRANS_FEAT(UMOPA_s, aa64_sme, do_outprod, a, MO_32, gen_helper_sme_umopa_s) +TRANS_FEAT(SUMOPA_s, aa64_sme, do_outprod, a, MO_32, gen_helper_sme_sumopa_s) +TRANS_FEAT(USMOPA_s, aa64_sme, do_outprod, a, MO_32, gen_helper_sme_usmopa_s) + +TRANS_FEAT(SMOPA_d, aa64_sme_i16i64, do_outprod, a, MO_64, gen_helper_sme_smopa_d) +TRANS_FEAT(UMOPA_d, aa64_sme_i16i64, do_outprod, a, MO_64, gen_helper_sme_umopa_d) +TRANS_FEAT(SUMOPA_d, aa64_sme_i16i64, do_outprod, a, MO_64, gen_helper_sme_sumopa_d) +TRANS_FEAT(USMOPA_d, aa64_sme_i16i64, do_outprod, a, MO_64, gen_helper_sme_usmopa_d) -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> This is an SVE instruction that operates using the SVE vector length but that it is present only if SME is implemented. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-29-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/sve.decode | 20 +++++++++++++ target/arm/translate-sve.c | 57 ++++++++++++++++++++++++++++++++++++++ 2 files changed, 77 insertions(+) diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ BFMLALT_zzxw 01100100 11 1 ..... 0100.1 ..... ..... @rrxr_3a esz=2 ### SVE2 floating-point bfloat16 dot-product (indexed) BFDOT_zzxz 01100100 01 1 ..... 010000 ..... ..... @rrxr_2 esz=2 + +### SVE broadcast predicate element + +&psel esz pd pn pm rv imm +%psel_rv 16:2 !function=plus_12 +%psel_imm_b 22:2 19:2 +%psel_imm_h 22:2 20:1 +%psel_imm_s 22:2 +%psel_imm_d 23:1 +@psel ........ .. . ... .. .. pn:4 . pm:4 . pd:4 \ + &psel rv=%psel_rv + +PSEL 00100101 .. 1 ..1 .. 01 .... 0 .... 0 .... \ + @psel esz=0 imm=%psel_imm_b +PSEL 00100101 .. 1 .10 .. 01 .... 0 .... 0 .... \ + @psel esz=1 imm=%psel_imm_h +PSEL 00100101 .. 1 100 .. 01 .... 0 .... 0 .... \ + @psel esz=2 imm=%psel_imm_s +PSEL 00100101 .1 1 000 .. 01 .... 0 .... 0 .... \ + @psel esz=3 imm=%psel_imm_d diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool do_BFMLAL_zzxw(DisasContext *s, arg_rrxr_esz *a, bool sel) TRANS_FEAT(BFMLALB_zzxw, aa64_sve_bf16, do_BFMLAL_zzxw, a, false) TRANS_FEAT(BFMLALT_zzxw, aa64_sve_bf16, do_BFMLAL_zzxw, a, true) + +static bool trans_PSEL(DisasContext *s, arg_psel *a) +{ + int vl = vec_full_reg_size(s); + int pl = pred_gvec_reg_size(s); + int elements = vl >> a->esz; + TCGv_i64 tmp, didx, dbit; + TCGv_ptr ptr; + + if (!dc_isar_feature(aa64_sme, s)) { + return false; + } + if (!sve_access_check(s)) { + return true; + } + + tmp = tcg_temp_new_i64(); + dbit = tcg_temp_new_i64(); + didx = tcg_temp_new_i64(); + ptr = tcg_temp_new_ptr(); + + /* Compute the predicate element. */ + tcg_gen_addi_i64(tmp, cpu_reg(s, a->rv), a->imm); + if (is_power_of_2(elements)) { + tcg_gen_andi_i64(tmp, tmp, elements - 1); + } else { + tcg_gen_remu_i64(tmp, tmp, tcg_constant_i64(elements)); + } + + /* Extract the predicate byte and bit indices. */ + tcg_gen_shli_i64(tmp, tmp, a->esz); + tcg_gen_andi_i64(dbit, tmp, 7); + tcg_gen_shri_i64(didx, tmp, 3); + if (HOST_BIG_ENDIAN) { + tcg_gen_xori_i64(didx, didx, 7); + } + + /* Load the predicate word. */ + tcg_gen_trunc_i64_ptr(ptr, didx); + tcg_gen_add_ptr(ptr, ptr, cpu_env); + tcg_gen_ld8u_i64(tmp, ptr, pred_full_reg_offset(s, a->pm)); + + /* Extract the predicate bit and replicate to MO_64. */ + tcg_gen_shr_i64(tmp, tmp, dbit); + tcg_gen_andi_i64(tmp, tmp, 1); + tcg_gen_neg_i64(tmp, tmp); + + /* Apply to either copy the source, or write zeros. */ + tcg_gen_gvec_ands(MO_64, pred_full_reg_offset(s, a->pd), + pred_full_reg_offset(s, a->pn), tmp, pl, pl); + + tcg_temp_free_i64(tmp); + tcg_temp_free_i64(dbit); + tcg_temp_free_i64(didx); + tcg_temp_free_ptr(ptr); + return true; +} -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> This is an SVE instruction that operates using the SVE vector length but that it is present only if SME is implemented. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-30-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 2 ++ target/arm/sve.decode | 1 + target/arm/sve_helper.c | 16 ++++++++++++++++ target/arm/translate-sve.c | 2 ++ 4 files changed, 21 insertions(+) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(sve_revh_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(sve_revw_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sme_revd_q, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_4(sve_rbit_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(sve_rbit_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(sve_rbit_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ REVB 00000101 .. 1001 00 100 ... ..... ..... @rd_pg_rn REVH 00000101 .. 1001 01 100 ... ..... ..... @rd_pg_rn REVW 00000101 .. 1001 10 100 ... ..... ..... @rd_pg_rn RBIT 00000101 .. 1001 11 100 ... ..... ..... @rd_pg_rn +REVD 00000101 00 1011 10 100 ... ..... ..... @rd_pg_rn_e0 # SVE vector splice (predicated, destructive) SPLICE 00000101 .. 101 100 100 ... ..... ..... @rdn_pg_rm 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 @@ DO_ZPZ_D(sve_revh_d, uint64_t, hswap64) DO_ZPZ_D(sve_revw_d, uint64_t, wswap64) +void HELPER(sme_revd_q)(void *vd, void *vn, void *vg, uint32_t desc) +{ + intptr_t i, opr_sz = simd_oprsz(desc) / 8; + uint64_t *d = vd, *n = vn; + uint8_t *pg = vg; + + for (i = 0; i < opr_sz; i += 2) { + if (pg[H1(i)] & 1) { + uint64_t n0 = n[i + 0]; + uint64_t n1 = n[i + 1]; + d[i + 0] = n1; + d[i + 1] = n0; + } + } +} + DO_ZPZ(sve_rbit_b, uint8_t, H1, revbit8) DO_ZPZ(sve_rbit_h, uint16_t, H1_2, revbit16) DO_ZPZ(sve_rbit_s, uint32_t, H1_4, revbit32) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ TRANS_FEAT(REVH, aa64_sve, gen_gvec_ool_arg_zpz, revh_fns[a->esz], a, 0) TRANS_FEAT(REVW, aa64_sve, gen_gvec_ool_arg_zpz, a->esz == 3 ? gen_helper_sve_revw_d : NULL, a, 0) +TRANS_FEAT(REVD, aa64_sme, gen_gvec_ool_arg_zpz, gen_helper_sme_revd_q, a, 0) + TRANS_FEAT(SPLICE, aa64_sve, gen_gvec_ool_arg_zpzz, gen_helper_sve_splice, a, a->esz) -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> This is an SVE instruction that operates using the SVE vector length but that it is present only if SME is implemented. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-31-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper.h | 18 +++++++ target/arm/sve.decode | 5 ++ target/arm/translate-sve.c | 102 +++++++++++++++++++++++++++++++++++++ target/arm/vec_helper.c | 24 +++++++++ 4 files changed, 149 insertions(+) diff --git a/target/arm/helper.h b/target/arm/helper.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper.h +++ b/target/arm/helper.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_6(gvec_bfmlal, TCG_CALL_NO_RWG, DEF_HELPER_FLAGS_6(gvec_bfmlal_idx, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_sclamp_b, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_sclamp_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_sclamp_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_sclamp_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_5(gvec_uclamp_b, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_uclamp_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_uclamp_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_uclamp_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) + #ifdef TARGET_AARCH64 #include "helper-a64.h" #include "helper-sve.h" diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ PSEL 00100101 .. 1 100 .. 01 .... 0 .... 0 .... \ @psel esz=2 imm=%psel_imm_s PSEL 00100101 .1 1 000 .. 01 .... 0 .... 0 .... \ @psel esz=3 imm=%psel_imm_d + +### SVE clamp + +SCLAMP 01000100 .. 0 ..... 110000 ..... ..... @rda_rn_rm +UCLAMP 01000100 .. 0 ..... 110001 ..... ..... @rda_rn_rm diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool trans_PSEL(DisasContext *s, arg_psel *a) tcg_temp_free_ptr(ptr); return true; } + +static void gen_sclamp_i32(TCGv_i32 d, TCGv_i32 n, TCGv_i32 m, TCGv_i32 a) +{ + tcg_gen_smax_i32(d, a, n); + tcg_gen_smin_i32(d, d, m); +} + +static void gen_sclamp_i64(TCGv_i64 d, TCGv_i64 n, TCGv_i64 m, TCGv_i64 a) +{ + tcg_gen_smax_i64(d, a, n); + tcg_gen_smin_i64(d, d, m); +} + +static void gen_sclamp_vec(unsigned vece, TCGv_vec d, TCGv_vec n, + TCGv_vec m, TCGv_vec a) +{ + tcg_gen_smax_vec(vece, d, a, n); + tcg_gen_smin_vec(vece, d, d, m); +} + +static void gen_sclamp(unsigned vece, uint32_t d, uint32_t n, uint32_t m, + uint32_t a, uint32_t oprsz, uint32_t maxsz) +{ + static const TCGOpcode vecop[] = { + INDEX_op_smin_vec, INDEX_op_smax_vec, 0 + }; + static const GVecGen4 ops[4] = { + { .fniv = gen_sclamp_vec, + .fno = gen_helper_gvec_sclamp_b, + .opt_opc = vecop, + .vece = MO_8 }, + { .fniv = gen_sclamp_vec, + .fno = gen_helper_gvec_sclamp_h, + .opt_opc = vecop, + .vece = MO_16 }, + { .fni4 = gen_sclamp_i32, + .fniv = gen_sclamp_vec, + .fno = gen_helper_gvec_sclamp_s, + .opt_opc = vecop, + .vece = MO_32 }, + { .fni8 = gen_sclamp_i64, + .fniv = gen_sclamp_vec, + .fno = gen_helper_gvec_sclamp_d, + .opt_opc = vecop, + .vece = MO_64, + .prefer_i64 = TCG_TARGET_REG_BITS == 64 } + }; + tcg_gen_gvec_4(d, n, m, a, oprsz, maxsz, &ops[vece]); +} + +TRANS_FEAT(SCLAMP, aa64_sme, gen_gvec_fn_arg_zzzz, gen_sclamp, a) + +static void gen_uclamp_i32(TCGv_i32 d, TCGv_i32 n, TCGv_i32 m, TCGv_i32 a) +{ + tcg_gen_umax_i32(d, a, n); + tcg_gen_umin_i32(d, d, m); +} + +static void gen_uclamp_i64(TCGv_i64 d, TCGv_i64 n, TCGv_i64 m, TCGv_i64 a) +{ + tcg_gen_umax_i64(d, a, n); + tcg_gen_umin_i64(d, d, m); +} + +static void gen_uclamp_vec(unsigned vece, TCGv_vec d, TCGv_vec n, + TCGv_vec m, TCGv_vec a) +{ + tcg_gen_umax_vec(vece, d, a, n); + tcg_gen_umin_vec(vece, d, d, m); +} + +static void gen_uclamp(unsigned vece, uint32_t d, uint32_t n, uint32_t m, + uint32_t a, uint32_t oprsz, uint32_t maxsz) +{ + static const TCGOpcode vecop[] = { + INDEX_op_umin_vec, INDEX_op_umax_vec, 0 + }; + static const GVecGen4 ops[4] = { + { .fniv = gen_uclamp_vec, + .fno = gen_helper_gvec_uclamp_b, + .opt_opc = vecop, + .vece = MO_8 }, + { .fniv = gen_uclamp_vec, + .fno = gen_helper_gvec_uclamp_h, + .opt_opc = vecop, + .vece = MO_16 }, + { .fni4 = gen_uclamp_i32, + .fniv = gen_uclamp_vec, + .fno = gen_helper_gvec_uclamp_s, + .opt_opc = vecop, + .vece = MO_32 }, + { .fni8 = gen_uclamp_i64, + .fniv = gen_uclamp_vec, + .fno = gen_helper_gvec_uclamp_d, + .opt_opc = vecop, + .vece = MO_64, + .prefer_i64 = TCG_TARGET_REG_BITS == 64 } + }; + tcg_gen_gvec_4(d, n, m, a, oprsz, maxsz, &ops[vece]); +} + +TRANS_FEAT(UCLAMP, aa64_sme, gen_gvec_fn_arg_zzzz, gen_uclamp, a) diff --git a/target/arm/vec_helper.c b/target/arm/vec_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/vec_helper.c +++ b/target/arm/vec_helper.c @@ -XXX,XX +XXX,XX @@ void HELPER(gvec_bfmlal_idx)(void *vd, void *vn, void *vm, } clear_tail(d, opr_sz, simd_maxsz(desc)); } + +#define DO_CLAMP(NAME, TYPE) \ +void HELPER(NAME)(void *d, void *n, void *m, void *a, uint32_t desc) \ +{ \ + intptr_t i, opr_sz = simd_oprsz(desc); \ + for (i = 0; i < opr_sz; i += sizeof(TYPE)) { \ + TYPE aa = *(TYPE *)(a + i); \ + TYPE nn = *(TYPE *)(n + i); \ + TYPE mm = *(TYPE *)(m + i); \ + TYPE dd = MIN(MAX(aa, nn), mm); \ + *(TYPE *)(d + i) = dd; \ + } \ + clear_tail(d, opr_sz, simd_maxsz(desc)); \ +} + +DO_CLAMP(gvec_sclamp_b, int8_t) +DO_CLAMP(gvec_sclamp_h, int16_t) +DO_CLAMP(gvec_sclamp_s, int32_t) +DO_CLAMP(gvec_sclamp_d, int64_t) + +DO_CLAMP(gvec_uclamp_b, uint8_t) +DO_CLAMP(gvec_uclamp_h, uint16_t) +DO_CLAMP(gvec_uclamp_s, uint32_t) +DO_CLAMP(gvec_uclamp_d, uint64_t) -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> We can handle both exception entry and exception return by hooking into aarch64_sve_change_el. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-32-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper.c | 15 +++++++++++++-- 1 file changed, 13 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 @@ void aarch64_sve_change_el(CPUARMState *env, int old_el, return; } + old_a64 = old_el ? arm_el_is_aa64(env, old_el) : el0_a64; + new_a64 = new_el ? arm_el_is_aa64(env, new_el) : el0_a64; + + /* + * Both AArch64.TakeException and AArch64.ExceptionReturn + * invoke ResetSVEState when taking an exception from, or + * returning to, AArch32 state when PSTATE.SM is enabled. + */ + if (old_a64 != new_a64 && FIELD_EX64(env->svcr, SVCR, SM)) { + arm_reset_sve_state(env); + return; + } + /* * DDI0584A.d sec 3.2: "If SVE instructions are disabled or trapped * at ELx, or not available because the EL is in AArch32 state, then @@ -XXX,XX +XXX,XX @@ void aarch64_sve_change_el(CPUARMState *env, int old_el, * we already have the correct register contents when encountering the * vq0->vq0 transition between EL0->EL1. */ - old_a64 = old_el ? arm_el_is_aa64(env, old_el) : el0_a64; old_len = (old_a64 && !sve_exception_el(env, old_el) ? sve_vqm1_for_el(env, old_el) : 0); - new_a64 = new_el ? arm_el_is_aa64(env, new_el) : el0_a64; new_len = (new_a64 && !sve_exception_el(env, new_el) ? sve_vqm1_for_el(env, new_el) : 0); -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Note that SME remains effectively disabled for user-only, because we do not yet set CPACR_EL1.SMEN. This needs to wait until the kernel ABI is implemented. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-33-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- docs/system/arm/emulation.rst | 4 ++++ target/arm/cpu64.c | 11 +++++++++++ 2 files changed, 15 insertions(+) diff --git a/docs/system/arm/emulation.rst b/docs/system/arm/emulation.rst index XXXXXXX..XXXXXXX 100644 --- a/docs/system/arm/emulation.rst +++ b/docs/system/arm/emulation.rst @@ -XXX,XX +XXX,XX @@ the following architecture extensions: - FEAT_SHA512 (Advanced SIMD SHA512 instructions) - FEAT_SM3 (Advanced SIMD SM3 instructions) - FEAT_SM4 (Advanced SIMD SM4 instructions) +- FEAT_SME (Scalable Matrix Extension) +- FEAT_SME_FA64 (Full A64 instruction set in Streaming SVE mode) +- FEAT_SME_F64F64 (Double-precision floating-point outer product instructions) +- FEAT_SME_I16I64 (16-bit to 64-bit integer widening outer product instructions) - FEAT_SPECRES (Speculation restriction instructions) - FEAT_SSBS (Speculative Store Bypass Safe) - FEAT_TLBIOS (TLB invalidate instructions in Outer Shareable domain) 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_AA64PFR1, MTE, 3); /* FEAT_MTE3 */ t = FIELD_DP64(t, ID_AA64PFR1, RAS_FRAC, 0); /* FEAT_RASv1p1 + FEAT_DoubleFault */ + t = FIELD_DP64(t, ID_AA64PFR1, SME, 1); /* FEAT_SME */ t = FIELD_DP64(t, ID_AA64PFR1, CSV2_FRAC, 0); /* FEAT_CSV2_2 */ cpu->isar.id_aa64pfr1 = t; @@ -XXX,XX +XXX,XX @@ static void aarch64_max_initfn(Object *obj) t = FIELD_DP64(t, ID_AA64DFR0, PMUVER, 5); /* FEAT_PMUv3p4 */ cpu->isar.id_aa64dfr0 = t; + t = cpu->isar.id_aa64smfr0; + t = FIELD_DP64(t, ID_AA64SMFR0, F32F32, 1); /* FEAT_SME */ + t = FIELD_DP64(t, ID_AA64SMFR0, B16F32, 1); /* FEAT_SME */ + t = FIELD_DP64(t, ID_AA64SMFR0, F16F32, 1); /* FEAT_SME */ + t = FIELD_DP64(t, ID_AA64SMFR0, I8I32, 0xf); /* FEAT_SME */ + t = FIELD_DP64(t, ID_AA64SMFR0, F64F64, 1); /* FEAT_SME_F64F64 */ + t = FIELD_DP64(t, ID_AA64SMFR0, I16I64, 0xf); /* FEAT_SME_I16I64 */ + t = FIELD_DP64(t, ID_AA64SMFR0, FA64, 1); /* FEAT_SME_FA64 */ + cpu->isar.id_aa64smfr0 = t; + /* Replicate the same data to the 32-bit id registers. */ aa32_max_features(cpu); -- 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: 20220708151540.18136-34-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- linux-user/aarch64/target_cpu.h | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/linux-user/aarch64/target_cpu.h b/linux-user/aarch64/target_cpu.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/aarch64/target_cpu.h +++ b/linux-user/aarch64/target_cpu.h @@ -XXX,XX +XXX,XX @@ static inline void cpu_clone_regs_parent(CPUARMState *env, unsigned flags) static inline void cpu_set_tls(CPUARMState *env, target_ulong newtls) { - /* Note that AArch64 Linux keeps the TLS pointer in TPIDR; this is + /* + * Note that AArch64 Linux keeps the TLS pointer in TPIDR; this is * different from AArch32 Linux, which uses TPIDRRO. */ env->cp15.tpidr_el[0] = newtls; + /* TPIDR2_EL0 is cleared with CLONE_SETTLS. */ + env->cp15.tpidr2_el0 = 0; } static inline abi_ulong get_sp_from_cpustate(CPUARMState *state) -- 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: 20220708151540.18136-35-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- linux-user/aarch64/cpu_loop.c | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/linux-user/aarch64/cpu_loop.c b/linux-user/aarch64/cpu_loop.c index XXXXXXX..XXXXXXX 100644 --- a/linux-user/aarch64/cpu_loop.c +++ b/linux-user/aarch64/cpu_loop.c @@ -XXX,XX +XXX,XX @@ void cpu_loop(CPUARMState *env) switch (trapnr) { case EXCP_SWI: + /* + * On syscall, PSTATE.ZA is preserved, along with the ZA matrix. + * PSTATE.SM is cleared, per SMSTOP, which does ResetSVEState. + */ + if (FIELD_EX64(env->svcr, SVCR, SM)) { + env->svcr = FIELD_DP64(env->svcr, SVCR, SM, 0); + arm_rebuild_hflags(env); + arm_reset_sve_state(env); + } ret = do_syscall(env, env->xregs[8], env->xregs[0], -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Make sure to zero the currently reserved fields. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-36-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- linux-user/aarch64/signal.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/linux-user/aarch64/signal.c b/linux-user/aarch64/signal.c index XXXXXXX..XXXXXXX 100644 --- a/linux-user/aarch64/signal.c +++ b/linux-user/aarch64/signal.c @@ -XXX,XX +XXX,XX @@ struct target_extra_context { struct target_sve_context { struct target_aarch64_ctx head; uint16_t vl; - uint16_t reserved[3]; + uint16_t flags; + uint16_t reserved[2]; /* The actual SVE data immediately follows. It is laid out * according to TARGET_SVE_SIG_{Z,P}REG_OFFSET, based off of * the original struct pointer. @@ -XXX,XX +XXX,XX @@ struct target_sve_context { #define TARGET_SVE_SIG_CONTEXT_SIZE(VQ) \ (TARGET_SVE_SIG_PREG_OFFSET(VQ, 17)) +#define TARGET_SVE_SIG_FLAG_SM 1 + struct target_rt_sigframe { struct target_siginfo info; struct target_ucontext uc; @@ -XXX,XX +XXX,XX @@ static void target_setup_sve_record(struct target_sve_context *sve, { int i, j; + memset(sve, 0, sizeof(*sve)); __put_user(TARGET_SVE_MAGIC, &sve->head.magic); __put_user(size, &sve->head.size); __put_user(vq * TARGET_SVE_VQ_BYTES, &sve->vl); + if (FIELD_EX64(env->svcr, SVCR, SM)) { + __put_user(TARGET_SVE_SIG_FLAG_SM, &sve->flags); + } /* Note that SVE regs are stored as a byte stream, with each byte element * at a subsequent address. This corresponds to a little-endian store -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Fold the return value setting into the goto, so each point of failure need not do both. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-37-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- linux-user/aarch64/signal.c | 26 +++++++++++--------------- 1 file changed, 11 insertions(+), 15 deletions(-) diff --git a/linux-user/aarch64/signal.c b/linux-user/aarch64/signal.c index XXXXXXX..XXXXXXX 100644 --- a/linux-user/aarch64/signal.c +++ b/linux-user/aarch64/signal.c @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, struct target_sve_context *sve = NULL; uint64_t extra_datap = 0; bool used_extra = false; - bool err = false; int vq = 0, sve_size = 0; target_restore_general_frame(env, sf); @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, switch (magic) { case 0: if (size != 0) { - err = true; - goto exit; + goto err; } if (used_extra) { ctx = NULL; @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, case TARGET_FPSIMD_MAGIC: if (fpsimd || size != sizeof(struct target_fpsimd_context)) { - err = true; - goto exit; + goto err; } fpsimd = (struct target_fpsimd_context *)ctx; break; @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, break; } } - err = true; - goto exit; + goto err; case TARGET_EXTRA_MAGIC: if (extra || size != sizeof(struct target_extra_context)) { - err = true; - goto exit; + goto err; } __get_user(extra_datap, &((struct target_extra_context *)ctx)->datap); @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, /* Unknown record -- we certainly didn't generate it. * Did we in fact get out of sync? */ - err = true; - goto exit; + goto err; } ctx = (void *)ctx + size; } @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, if (fpsimd) { target_restore_fpsimd_record(env, fpsimd); } else { - err = true; + goto err; } /* SVE data, if present, overwrites FPSIMD data. */ if (sve) { target_restore_sve_record(env, sve, vq); } - - exit: unlock_user(extra, extra_datap, 0); - return err; + return 0; + + err: + unlock_user(extra, extra_datap, 0); + return 1; } static abi_ulong get_sigframe(struct target_sigaction *ka, -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> In parse_user_sigframe, the kernel rejects duplicate sve records, or records that are smaller than the header. We were silently allowing these cases to pass, dropping the record. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-38-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- linux-user/aarch64/signal.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/linux-user/aarch64/signal.c b/linux-user/aarch64/signal.c index XXXXXXX..XXXXXXX 100644 --- a/linux-user/aarch64/signal.c +++ b/linux-user/aarch64/signal.c @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, break; case TARGET_SVE_MAGIC: + if (sve || size < sizeof(struct target_sve_context)) { + goto err; + } if (cpu_isar_feature(aa64_sve, env_archcpu(env))) { vq = sve_vq(env); sve_size = QEMU_ALIGN_UP(TARGET_SVE_SIG_CONTEXT_SIZE(vq), 16); - if (!sve && size == sve_size) { + if (size == sve_size) { sve = (struct target_sve_context *)ctx; break; } -- 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: 20220708151540.18136-39-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- linux-user/aarch64/signal.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/linux-user/aarch64/signal.c b/linux-user/aarch64/signal.c index XXXXXXX..XXXXXXX 100644 --- a/linux-user/aarch64/signal.c +++ b/linux-user/aarch64/signal.c @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, __get_user(extra_size, &((struct target_extra_context *)ctx)->size); extra = lock_user(VERIFY_READ, extra_datap, extra_size, 0); + if (!extra) { + return 1; + } break; default: -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Move the checks out of the parsing loop and into the restore function. This more closely mirrors the code structure in the kernel, and is slightly clearer. Reject rather than silently skip incorrect VL and SVE record sizes, bringing our checks in to line with those the kernel does. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-40-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- linux-user/aarch64/signal.c | 51 +++++++++++++++++++++++++------------ 1 file changed, 35 insertions(+), 16 deletions(-) diff --git a/linux-user/aarch64/signal.c b/linux-user/aarch64/signal.c index XXXXXXX..XXXXXXX 100644 --- a/linux-user/aarch64/signal.c +++ b/linux-user/aarch64/signal.c @@ -XXX,XX +XXX,XX @@ static void target_restore_fpsimd_record(CPUARMState *env, } } -static void target_restore_sve_record(CPUARMState *env, - struct target_sve_context *sve, int vq) +static bool target_restore_sve_record(CPUARMState *env, + struct target_sve_context *sve, + int size) { - int i, j; + int i, j, vl, vq; - /* Note that SVE regs are stored as a byte stream, with each byte element + if (!cpu_isar_feature(aa64_sve, env_archcpu(env))) { + return false; + } + + __get_user(vl, &sve->vl); + vq = sve_vq(env); + + /* Reject mismatched VL. */ + if (vl != vq * TARGET_SVE_VQ_BYTES) { + return false; + } + + /* Accept empty record -- used to clear PSTATE.SM. */ + if (size <= sizeof(*sve)) { + return true; + } + + /* Reject non-empty but incomplete record. */ + if (size < TARGET_SVE_SIG_CONTEXT_SIZE(vq)) { + return false; + } + + /* + * Note that SVE regs are stored as a byte stream, with each byte element * at a subsequent address. This corresponds to a little-endian load * of our 64-bit hunks. */ @@ -XXX,XX +XXX,XX @@ static void target_restore_sve_record(CPUARMState *env, } } } + return true; } static int target_restore_sigframe(CPUARMState *env, @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, struct target_sve_context *sve = NULL; uint64_t extra_datap = 0; bool used_extra = false; - int vq = 0, sve_size = 0; + int sve_size = 0; target_restore_general_frame(env, sf); @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, if (sve || size < sizeof(struct target_sve_context)) { goto err; } - if (cpu_isar_feature(aa64_sve, env_archcpu(env))) { - vq = sve_vq(env); - sve_size = QEMU_ALIGN_UP(TARGET_SVE_SIG_CONTEXT_SIZE(vq), 16); - if (size == sve_size) { - sve = (struct target_sve_context *)ctx; - break; - } - } - goto err; + sve = (struct target_sve_context *)ctx; + sve_size = size; + break; case TARGET_EXTRA_MAGIC: if (extra || size != sizeof(struct target_extra_context)) { @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, } /* SVE data, if present, overwrites FPSIMD data. */ - if (sve) { - target_restore_sve_record(env, sve, vq); + if (sve && !target_restore_sve_record(env, sve, sve_size)) { + goto err; } unlock_user(extra, extra_datap, 0); return 0; -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Set the SM bit in the SVE record on signal delivery, create the ZA record. Restore SM and ZA state according to the records present on return. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-41-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- linux-user/aarch64/signal.c | 167 +++++++++++++++++++++++++++++++++--- 1 file changed, 154 insertions(+), 13 deletions(-) diff --git a/linux-user/aarch64/signal.c b/linux-user/aarch64/signal.c index XXXXXXX..XXXXXXX 100644 --- a/linux-user/aarch64/signal.c +++ b/linux-user/aarch64/signal.c @@ -XXX,XX +XXX,XX @@ struct target_sve_context { #define TARGET_SVE_SIG_FLAG_SM 1 +#define TARGET_ZA_MAGIC 0x54366345 + +struct target_za_context { + struct target_aarch64_ctx head; + uint16_t vl; + uint16_t reserved[3]; + /* The actual ZA data immediately follows. */ +}; + +#define TARGET_ZA_SIG_REGS_OFFSET \ + QEMU_ALIGN_UP(sizeof(struct target_za_context), TARGET_SVE_VQ_BYTES) +#define TARGET_ZA_SIG_ZAV_OFFSET(VQ, N) \ + (TARGET_ZA_SIG_REGS_OFFSET + (VQ) * TARGET_SVE_VQ_BYTES * (N)) +#define TARGET_ZA_SIG_CONTEXT_SIZE(VQ) \ + TARGET_ZA_SIG_ZAV_OFFSET(VQ, VQ * TARGET_SVE_VQ_BYTES) + struct target_rt_sigframe { struct target_siginfo info; struct target_ucontext uc; @@ -XXX,XX +XXX,XX @@ static void target_setup_end_record(struct target_aarch64_ctx *end) } static void target_setup_sve_record(struct target_sve_context *sve, - CPUARMState *env, int vq, int size) + CPUARMState *env, int size) { - int i, j; + int i, j, vq = sve_vq(env); memset(sve, 0, sizeof(*sve)); __put_user(TARGET_SVE_MAGIC, &sve->head.magic); @@ -XXX,XX +XXX,XX @@ static void target_setup_sve_record(struct target_sve_context *sve, } } +static void target_setup_za_record(struct target_za_context *za, + CPUARMState *env, int size) +{ + int vq = sme_vq(env); + int vl = vq * TARGET_SVE_VQ_BYTES; + int i, j; + + memset(za, 0, sizeof(*za)); + __put_user(TARGET_ZA_MAGIC, &za->head.magic); + __put_user(size, &za->head.size); + __put_user(vl, &za->vl); + + if (size == TARGET_ZA_SIG_CONTEXT_SIZE(0)) { + return; + } + assert(size == TARGET_ZA_SIG_CONTEXT_SIZE(vq)); + + /* + * Note that ZA vectors are stored as a byte stream, + * with each byte element at a subsequent address. + */ + for (i = 0; i < vl; ++i) { + uint64_t *z = (void *)za + TARGET_ZA_SIG_ZAV_OFFSET(vq, i); + for (j = 0; j < vq * 2; ++j) { + __put_user_e(env->zarray[i].d[j], z + j, le); + } + } +} + static void target_restore_general_frame(CPUARMState *env, struct target_rt_sigframe *sf) { @@ -XXX,XX +XXX,XX @@ static void target_restore_fpsimd_record(CPUARMState *env, static bool target_restore_sve_record(CPUARMState *env, struct target_sve_context *sve, - int size) + int size, int *svcr) { - int i, j, vl, vq; + int i, j, vl, vq, flags; + bool sm; - if (!cpu_isar_feature(aa64_sve, env_archcpu(env))) { + __get_user(vl, &sve->vl); + __get_user(flags, &sve->flags); + + sm = flags & TARGET_SVE_SIG_FLAG_SM; + + /* The cpu must support Streaming or Non-streaming SVE. */ + if (sm + ? !cpu_isar_feature(aa64_sme, env_archcpu(env)) + : !cpu_isar_feature(aa64_sve, env_archcpu(env))) { return false; } - __get_user(vl, &sve->vl); - vq = sve_vq(env); + /* + * Note that we cannot use sve_vq() because that depends on the + * current setting of PSTATE.SM, not the state to be restored. + */ + vq = sve_vqm1_for_el_sm(env, 0, sm) + 1; /* Reject mismatched VL. */ if (vl != vq * TARGET_SVE_VQ_BYTES) { @@ -XXX,XX +XXX,XX @@ static bool target_restore_sve_record(CPUARMState *env, return false; } + *svcr = FIELD_DP64(*svcr, SVCR, SM, sm); + /* * Note that SVE regs are stored as a byte stream, with each byte element * at a subsequent address. This corresponds to a little-endian load @@ -XXX,XX +XXX,XX @@ static bool target_restore_sve_record(CPUARMState *env, return true; } +static bool target_restore_za_record(CPUARMState *env, + struct target_za_context *za, + int size, int *svcr) +{ + int i, j, vl, vq; + + if (!cpu_isar_feature(aa64_sme, env_archcpu(env))) { + return false; + } + + __get_user(vl, &za->vl); + vq = sme_vq(env); + + /* Reject mismatched VL. */ + if (vl != vq * TARGET_SVE_VQ_BYTES) { + return false; + } + + /* Accept empty record -- used to clear PSTATE.ZA. */ + if (size <= TARGET_ZA_SIG_CONTEXT_SIZE(0)) { + return true; + } + + /* Reject non-empty but incomplete record. */ + if (size < TARGET_ZA_SIG_CONTEXT_SIZE(vq)) { + return false; + } + + *svcr = FIELD_DP64(*svcr, SVCR, ZA, 1); + + for (i = 0; i < vl; ++i) { + uint64_t *z = (void *)za + TARGET_ZA_SIG_ZAV_OFFSET(vq, i); + for (j = 0; j < vq * 2; ++j) { + __get_user_e(env->zarray[i].d[j], z + j, le); + } + } + return true; +} + static int target_restore_sigframe(CPUARMState *env, struct target_rt_sigframe *sf) { struct target_aarch64_ctx *ctx, *extra = NULL; struct target_fpsimd_context *fpsimd = NULL; struct target_sve_context *sve = NULL; + struct target_za_context *za = NULL; uint64_t extra_datap = 0; bool used_extra = false; int sve_size = 0; + int za_size = 0; + int svcr = 0; target_restore_general_frame(env, sf); @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, sve_size = size; break; + case TARGET_ZA_MAGIC: + if (za || size < sizeof(struct target_za_context)) { + goto err; + } + za = (struct target_za_context *)ctx; + za_size = size; + break; + case TARGET_EXTRA_MAGIC: if (extra || size != sizeof(struct target_extra_context)) { goto err; @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, } /* SVE data, if present, overwrites FPSIMD data. */ - if (sve && !target_restore_sve_record(env, sve, sve_size)) { + if (sve && !target_restore_sve_record(env, sve, sve_size, &svcr)) { goto err; } + if (za && !target_restore_za_record(env, za, za_size, &svcr)) { + goto err; + } + if (env->svcr != svcr) { + env->svcr = svcr; + arm_rebuild_hflags(env); + } unlock_user(extra, extra_datap, 0); return 0; @@ -XXX,XX +XXX,XX @@ static void target_setup_frame(int usig, struct target_sigaction *ka, .total_size = offsetof(struct target_rt_sigframe, uc.tuc_mcontext.__reserved), }; - int fpsimd_ofs, fr_ofs, sve_ofs = 0, vq = 0, sve_size = 0; + int fpsimd_ofs, fr_ofs, sve_ofs = 0, za_ofs = 0; + int sve_size = 0, za_size = 0; struct target_rt_sigframe *frame; struct target_rt_frame_record *fr; abi_ulong frame_addr, return_addr; @@ -XXX,XX +XXX,XX @@ static void target_setup_frame(int usig, struct target_sigaction *ka, &layout); /* SVE state needs saving only if it exists. */ - if (cpu_isar_feature(aa64_sve, env_archcpu(env))) { - vq = sve_vq(env); - sve_size = QEMU_ALIGN_UP(TARGET_SVE_SIG_CONTEXT_SIZE(vq), 16); + if (cpu_isar_feature(aa64_sve, env_archcpu(env)) || + cpu_isar_feature(aa64_sme, env_archcpu(env))) { + sve_size = QEMU_ALIGN_UP(TARGET_SVE_SIG_CONTEXT_SIZE(sve_vq(env)), 16); sve_ofs = alloc_sigframe_space(sve_size, &layout); } + if (cpu_isar_feature(aa64_sme, env_archcpu(env))) { + /* ZA state needs saving only if it is enabled. */ + if (FIELD_EX64(env->svcr, SVCR, ZA)) { + za_size = TARGET_ZA_SIG_CONTEXT_SIZE(sme_vq(env)); + } else { + za_size = TARGET_ZA_SIG_CONTEXT_SIZE(0); + } + za_ofs = alloc_sigframe_space(za_size, &layout); + } if (layout.extra_ofs) { /* Reserve space for the extra end marker. The standard end marker @@ -XXX,XX +XXX,XX @@ static void target_setup_frame(int usig, struct target_sigaction *ka, target_setup_end_record((void *)frame + layout.extra_end_ofs); } if (sve_ofs) { - target_setup_sve_record((void *)frame + sve_ofs, env, vq, sve_size); + target_setup_sve_record((void *)frame + sve_ofs, env, sve_size); + } + if (za_ofs) { + target_setup_za_record((void *)frame + za_ofs, env, za_size); } /* Set up the stack frame for unwinding. */ @@ -XXX,XX +XXX,XX @@ static void target_setup_frame(int usig, struct target_sigaction *ka, env->btype = 2; } + /* + * Invoke the signal handler with both SM and ZA disabled. + * When clearing SM, ResetSVEState, per SMSTOP. + */ + if (FIELD_EX64(env->svcr, SVCR, SM)) { + arm_reset_sve_state(env); + } + if (env->svcr) { + env->svcr = 0; + arm_rebuild_hflags(env); + } + if (info) { tswap_siginfo(&frame->info, info); env->xregs[1] = frame_addr + offsetof(struct target_rt_sigframe, info); -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Add "sve" to the sve prctl functions, to distinguish them from the coming "sme" prctls with similar names. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-42-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- linux-user/aarch64/target_prctl.h | 8 ++++---- linux-user/syscall.c | 12 ++++++------ 2 files changed, 10 insertions(+), 10 deletions(-) diff --git a/linux-user/aarch64/target_prctl.h b/linux-user/aarch64/target_prctl.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/aarch64/target_prctl.h +++ b/linux-user/aarch64/target_prctl.h @@ -XXX,XX +XXX,XX @@ #ifndef AARCH64_TARGET_PRCTL_H #define AARCH64_TARGET_PRCTL_H -static abi_long do_prctl_get_vl(CPUArchState *env) +static abi_long do_prctl_sve_get_vl(CPUArchState *env) { ARMCPU *cpu = env_archcpu(env); if (cpu_isar_feature(aa64_sve, cpu)) { @@ -XXX,XX +XXX,XX @@ static abi_long do_prctl_get_vl(CPUArchState *env) } return -TARGET_EINVAL; } -#define do_prctl_get_vl do_prctl_get_vl +#define do_prctl_sve_get_vl do_prctl_sve_get_vl -static abi_long do_prctl_set_vl(CPUArchState *env, abi_long arg2) +static abi_long do_prctl_sve_set_vl(CPUArchState *env, abi_long arg2) { /* * We cannot support either PR_SVE_SET_VL_ONEXEC or PR_SVE_VL_INHERIT. @@ -XXX,XX +XXX,XX @@ static abi_long do_prctl_set_vl(CPUArchState *env, abi_long arg2) } return -TARGET_EINVAL; } -#define do_prctl_set_vl do_prctl_set_vl +#define do_prctl_sve_set_vl do_prctl_sve_set_vl static abi_long do_prctl_reset_keys(CPUArchState *env, abi_long arg2) { diff --git a/linux-user/syscall.c b/linux-user/syscall.c index XXXXXXX..XXXXXXX 100644 --- a/linux-user/syscall.c +++ b/linux-user/syscall.c @@ -XXX,XX +XXX,XX @@ static abi_long do_prctl_inval1(CPUArchState *env, abi_long arg2) #ifndef do_prctl_set_fp_mode #define do_prctl_set_fp_mode do_prctl_inval1 #endif -#ifndef do_prctl_get_vl -#define do_prctl_get_vl do_prctl_inval0 +#ifndef do_prctl_sve_get_vl +#define do_prctl_sve_get_vl do_prctl_inval0 #endif -#ifndef do_prctl_set_vl -#define do_prctl_set_vl do_prctl_inval1 +#ifndef do_prctl_sve_set_vl +#define do_prctl_sve_set_vl do_prctl_inval1 #endif #ifndef do_prctl_reset_keys #define do_prctl_reset_keys do_prctl_inval1 @@ -XXX,XX +XXX,XX @@ static abi_long do_prctl(CPUArchState *env, abi_long option, abi_long arg2, case PR_SET_FP_MODE: return do_prctl_set_fp_mode(env, arg2); case PR_SVE_GET_VL: - return do_prctl_get_vl(env); + return do_prctl_sve_get_vl(env); case PR_SVE_SET_VL: - return do_prctl_set_vl(env, arg2); + return do_prctl_sve_set_vl(env, arg2); case PR_PAC_RESET_KEYS: if (arg3 || arg4 || arg5) { return -TARGET_EINVAL; -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> These prctl set the Streaming SVE vector length, which may be completely different from the Normal SVE vector length. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-43-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- linux-user/aarch64/target_prctl.h | 54 +++++++++++++++++++++++++++++++ linux-user/syscall.c | 16 +++++++++ 2 files changed, 70 insertions(+) diff --git a/linux-user/aarch64/target_prctl.h b/linux-user/aarch64/target_prctl.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/aarch64/target_prctl.h +++ b/linux-user/aarch64/target_prctl.h @@ -XXX,XX +XXX,XX @@ static abi_long do_prctl_sve_get_vl(CPUArchState *env) { ARMCPU *cpu = env_archcpu(env); if (cpu_isar_feature(aa64_sve, cpu)) { + /* PSTATE.SM is always unset on syscall entry. */ return sve_vq(env) * 16; } return -TARGET_EINVAL; @@ -XXX,XX +XXX,XX @@ static abi_long do_prctl_sve_set_vl(CPUArchState *env, abi_long arg2) && arg2 >= 0 && arg2 <= 512 * 16 && !(arg2 & 15)) { uint32_t vq, old_vq; + /* PSTATE.SM is always unset on syscall entry. */ old_vq = sve_vq(env); /* @@ -XXX,XX +XXX,XX @@ static abi_long do_prctl_sve_set_vl(CPUArchState *env, abi_long arg2) } #define do_prctl_sve_set_vl do_prctl_sve_set_vl +static abi_long do_prctl_sme_get_vl(CPUArchState *env) +{ + ARMCPU *cpu = env_archcpu(env); + if (cpu_isar_feature(aa64_sme, cpu)) { + return sme_vq(env) * 16; + } + return -TARGET_EINVAL; +} +#define do_prctl_sme_get_vl do_prctl_sme_get_vl + +static abi_long do_prctl_sme_set_vl(CPUArchState *env, abi_long arg2) +{ + /* + * We cannot support either PR_SME_SET_VL_ONEXEC or PR_SME_VL_INHERIT. + * Note the kernel definition of sve_vl_valid allows for VQ=512, + * i.e. VL=8192, even though the architectural maximum is VQ=16. + */ + if (cpu_isar_feature(aa64_sme, env_archcpu(env)) + && arg2 >= 0 && arg2 <= 512 * 16 && !(arg2 & 15)) { + int vq, old_vq; + + old_vq = sme_vq(env); + + /* + * Bound the value of vq, so that we know that it fits into + * the 4-bit field in SMCR_EL1. Because PSTATE.SM is cleared + * on syscall entry, we are not modifying the current SVE + * vector length. + */ + vq = MAX(arg2 / 16, 1); + vq = MIN(vq, 16); + env->vfp.smcr_el[1] = + FIELD_DP64(env->vfp.smcr_el[1], SMCR, LEN, vq - 1); + + /* Delay rebuilding hflags until we know if ZA must change. */ + vq = sve_vqm1_for_el_sm(env, 0, true) + 1; + + if (vq != old_vq) { + /* + * PSTATE.ZA state is cleared on any change to SVL. + * We need not call arm_rebuild_hflags because PSTATE.SM was + * cleared on syscall entry, so this hasn't changed VL. + */ + env->svcr = FIELD_DP64(env->svcr, SVCR, ZA, 0); + arm_rebuild_hflags(env); + } + return vq * 16; + } + return -TARGET_EINVAL; +} +#define do_prctl_sme_set_vl do_prctl_sme_set_vl + static abi_long do_prctl_reset_keys(CPUArchState *env, abi_long arg2) { ARMCPU *cpu = env_archcpu(env); diff --git a/linux-user/syscall.c b/linux-user/syscall.c index XXXXXXX..XXXXXXX 100644 --- a/linux-user/syscall.c +++ b/linux-user/syscall.c @@ -XXX,XX +XXX,XX @@ abi_long do_arch_prctl(CPUX86State *env, int code, abi_ulong addr) #ifndef PR_SET_SYSCALL_USER_DISPATCH # define PR_SET_SYSCALL_USER_DISPATCH 59 #endif +#ifndef PR_SME_SET_VL +# define PR_SME_SET_VL 63 +# define PR_SME_GET_VL 64 +# define PR_SME_VL_LEN_MASK 0xffff +# define PR_SME_VL_INHERIT (1 << 17) +#endif #include "target_prctl.h" @@ -XXX,XX +XXX,XX @@ static abi_long do_prctl_inval1(CPUArchState *env, abi_long arg2) #ifndef do_prctl_set_unalign #define do_prctl_set_unalign do_prctl_inval1 #endif +#ifndef do_prctl_sme_get_vl +#define do_prctl_sme_get_vl do_prctl_inval0 +#endif +#ifndef do_prctl_sme_set_vl +#define do_prctl_sme_set_vl do_prctl_inval1 +#endif static abi_long do_prctl(CPUArchState *env, abi_long option, abi_long arg2, abi_long arg3, abi_long arg4, abi_long arg5) @@ -XXX,XX +XXX,XX @@ static abi_long do_prctl(CPUArchState *env, abi_long option, abi_long arg2, return do_prctl_sve_get_vl(env); case PR_SVE_SET_VL: return do_prctl_sve_set_vl(env, arg2); + case PR_SME_GET_VL: + return do_prctl_sme_get_vl(env); + case PR_SME_SET_VL: + return do_prctl_sme_set_vl(env, arg2); case PR_PAC_RESET_KEYS: if (arg3 || arg4 || arg5) { return -TARGET_EINVAL; -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> There's no reason to set CPACR_EL1.ZEN if SVE disabled. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-44-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/cpu.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) 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_cpu_reset(DeviceState *dev) /* and to the FP/Neon instructions */ env->cp15.cpacr_el1 = FIELD_DP64(env->cp15.cpacr_el1, CPACR_EL1, FPEN, 3); - /* and to the SVE instructions */ - env->cp15.cpacr_el1 = FIELD_DP64(env->cp15.cpacr_el1, - CPACR_EL1, ZEN, 3); - /* with reasonable vector length */ + /* and to the SVE instructions, with default vector length */ if (cpu_isar_feature(aa64_sve, cpu)) { + env->cp15.cpacr_el1 = FIELD_DP64(env->cp15.cpacr_el1, + CPACR_EL1, ZEN, 3); env->vfp.zcr_el[1] = cpu->sve_default_vq - 1; } /* -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Enable SME, TPIDR2_EL0, and FA64 if supported by the cpu. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-45-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/cpu.c | 11 +++++++++++ 1 file changed, 11 insertions(+) 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_cpu_reset(DeviceState *dev) CPACR_EL1, ZEN, 3); env->vfp.zcr_el[1] = cpu->sve_default_vq - 1; } + /* and for SME instructions, with default vector length, and TPIDR2 */ + if (cpu_isar_feature(aa64_sme, cpu)) { + env->cp15.sctlr_el[1] |= SCTLR_EnTP2; + env->cp15.cpacr_el1 = FIELD_DP64(env->cp15.cpacr_el1, + CPACR_EL1, SMEN, 3); + env->vfp.smcr_el[1] = cpu->sme_default_vq - 1; + if (cpu_isar_feature(aa64_sme_fa64, cpu)) { + env->vfp.smcr_el[1] = FIELD_DP64(env->vfp.smcr_el[1], + SMCR, FA64, 1); + } + } /* * Enable 48-bit address space (TODO: take reserved_va into account). * Enable TBI0 but not TBI1. -- 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: 20220708151540.18136-46-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- linux-user/elfload.c | 20 ++++++++++++++++++++ 1 file changed, 20 insertions(+) diff --git a/linux-user/elfload.c b/linux-user/elfload.c index XXXXXXX..XXXXXXX 100644 --- a/linux-user/elfload.c +++ b/linux-user/elfload.c @@ -XXX,XX +XXX,XX @@ enum { ARM_HWCAP2_A64_RNG = 1 << 16, ARM_HWCAP2_A64_BTI = 1 << 17, ARM_HWCAP2_A64_MTE = 1 << 18, + ARM_HWCAP2_A64_ECV = 1 << 19, + ARM_HWCAP2_A64_AFP = 1 << 20, + ARM_HWCAP2_A64_RPRES = 1 << 21, + ARM_HWCAP2_A64_MTE3 = 1 << 22, + ARM_HWCAP2_A64_SME = 1 << 23, + ARM_HWCAP2_A64_SME_I16I64 = 1 << 24, + ARM_HWCAP2_A64_SME_F64F64 = 1 << 25, + ARM_HWCAP2_A64_SME_I8I32 = 1 << 26, + ARM_HWCAP2_A64_SME_F16F32 = 1 << 27, + ARM_HWCAP2_A64_SME_B16F32 = 1 << 28, + ARM_HWCAP2_A64_SME_F32F32 = 1 << 29, + ARM_HWCAP2_A64_SME_FA64 = 1 << 30, }; #define ELF_HWCAP get_elf_hwcap() @@ -XXX,XX +XXX,XX @@ static uint32_t get_elf_hwcap2(void) GET_FEATURE_ID(aa64_rndr, ARM_HWCAP2_A64_RNG); GET_FEATURE_ID(aa64_bti, ARM_HWCAP2_A64_BTI); GET_FEATURE_ID(aa64_mte, ARM_HWCAP2_A64_MTE); + GET_FEATURE_ID(aa64_sme, (ARM_HWCAP2_A64_SME | + ARM_HWCAP2_A64_SME_F32F32 | + ARM_HWCAP2_A64_SME_B16F32 | + ARM_HWCAP2_A64_SME_F16F32 | + ARM_HWCAP2_A64_SME_I8I32)); + GET_FEATURE_ID(aa64_sme_f64f64, ARM_HWCAP2_A64_SME_F64F64); + GET_FEATURE_ID(aa64_sme_i16i64, ARM_HWCAP2_A64_SME_I16I64); + GET_FEATURE_ID(aa64_sme_fa64, ARM_HWCAP2_A64_SME_FA64); return hwcaps; } -- 2.25.1 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