Many instructions which load/store 128-bit values are supposed to raise #GP when the memory operand isn't 16-byte aligned. This includes: - Instructions explicitly requiring memory alignment (Exceptions Type 1 in the "AVX and SSE Instruction Exception Specification" section of the SDM) - Legacy SSE instructions that load/store 128-bit values (Exceptions Types 2 and 4).
This change sets MO_ALIGN_16 on 128-bit memory accesses that require 16-byte alignment. It adds cpu_record_sigbus and cpu_do_unaligned_access handlers that simulate a #GP exception in qemu-user and qemu-system, respectively. One minor behavior change apart from what is described above: Prior to this change, emulation of cvttps2pi and cvtps2pi instructions would incorrectly read 16 bytes of memory instead of 8. I double-checked on real hardware that these instructions only read 8 bytes (and do not have any address alignment requirements). Resolves: https://gitlab.com/qemu-project/qemu/-/issues/217 Signed-off-by: Ricky Zhou <ri...@rzhou.org> --- target/i386/tcg/excp_helper.c | 13 ++++++++ target/i386/tcg/helper-tcg.h | 28 +++++++++------- target/i386/tcg/sysemu/excp_helper.c | 8 +++++ target/i386/tcg/tcg-cpu.c | 2 ++ target/i386/tcg/translate.c | 49 ++++++++++++++++++---------- target/i386/tcg/user/excp_helper.c | 7 ++++ 6 files changed, 78 insertions(+), 29 deletions(-) diff --git a/target/i386/tcg/excp_helper.c b/target/i386/tcg/excp_helper.c index c1ffa1c0ef..7c3c8dc7fe 100644 --- a/target/i386/tcg/excp_helper.c +++ b/target/i386/tcg/excp_helper.c @@ -140,3 +140,16 @@ G_NORETURN void raise_exception_ra(CPUX86State *env, int exception_index, { raise_interrupt2(env, exception_index, 0, 0, 0, retaddr); } + +G_NORETURN void handle_unaligned_access(CPUX86State *env, vaddr vaddr, + MMUAccessType access_type, + uintptr_t retaddr) +{ + /* + * Unaligned accesses are currently only triggered by SSE/AVX + * instructions that impose alignment requirements on memory + * operands. These instructions raise #GP(0) upon accessing an + * unaligned address. + */ + raise_exception_ra(env, EXCP0D_GPF, retaddr); +} diff --git a/target/i386/tcg/helper-tcg.h b/target/i386/tcg/helper-tcg.h index 34167e2e29..cd1723389a 100644 --- a/target/i386/tcg/helper-tcg.h +++ b/target/i386/tcg/helper-tcg.h @@ -42,17 +42,6 @@ void x86_cpu_do_interrupt(CPUState *cpu); bool x86_cpu_exec_interrupt(CPUState *cpu, int int_req); #endif -/* helper.c */ -#ifdef CONFIG_USER_ONLY -void x86_cpu_record_sigsegv(CPUState *cs, vaddr addr, - MMUAccessType access_type, - bool maperr, uintptr_t ra); -#else -bool x86_cpu_tlb_fill(CPUState *cs, vaddr address, int size, - MMUAccessType access_type, int mmu_idx, - bool probe, uintptr_t retaddr); -#endif - void breakpoint_handler(CPUState *cs); /* n must be a constant to be efficient */ @@ -78,6 +67,23 @@ G_NORETURN void raise_exception_err_ra(CPUX86State *env, int exception_index, int error_code, uintptr_t retaddr); G_NORETURN void raise_interrupt(CPUX86State *nenv, int intno, int is_int, int error_code, int next_eip_addend); +G_NORETURN void handle_unaligned_access(CPUX86State *env, vaddr vaddr, + MMUAccessType access_type, + uintptr_t retaddr); +#ifdef CONFIG_USER_ONLY +void x86_cpu_record_sigsegv(CPUState *cs, vaddr addr, + MMUAccessType access_type, + bool maperr, uintptr_t ra); +void x86_cpu_record_sigbus(CPUState *cs, vaddr addr, + MMUAccessType access_type, uintptr_t ra); +#else +bool x86_cpu_tlb_fill(CPUState *cs, vaddr address, int size, + MMUAccessType access_type, int mmu_idx, + bool probe, uintptr_t retaddr); +G_NORETURN void x86_cpu_do_unaligned_access(CPUState *cs, vaddr vaddr, + MMUAccessType access_type, + int mmu_idx, uintptr_t retaddr); +#endif /* cc_helper.c */ extern const uint8_t parity_table[256]; diff --git a/target/i386/tcg/sysemu/excp_helper.c b/target/i386/tcg/sysemu/excp_helper.c index 48feba7e75..796dc2a1f3 100644 --- a/target/i386/tcg/sysemu/excp_helper.c +++ b/target/i386/tcg/sysemu/excp_helper.c @@ -439,3 +439,11 @@ bool x86_cpu_tlb_fill(CPUState *cs, vaddr addr, int size, } return true; } + +G_NORETURN void x86_cpu_do_unaligned_access(CPUState *cs, vaddr vaddr, + MMUAccessType access_type, + int mmu_idx, uintptr_t retaddr) +{ + X86CPU *cpu = X86_CPU(cs); + handle_unaligned_access(&cpu->env, vaddr, access_type, retaddr); +} diff --git a/target/i386/tcg/tcg-cpu.c b/target/i386/tcg/tcg-cpu.c index 6fdfdf9598..d3c2b8fb49 100644 --- a/target/i386/tcg/tcg-cpu.c +++ b/target/i386/tcg/tcg-cpu.c @@ -75,10 +75,12 @@ static const struct TCGCPUOps x86_tcg_ops = { #ifdef CONFIG_USER_ONLY .fake_user_interrupt = x86_cpu_do_interrupt, .record_sigsegv = x86_cpu_record_sigsegv, + .record_sigbus = x86_cpu_record_sigbus, #else .tlb_fill = x86_cpu_tlb_fill, .do_interrupt = x86_cpu_do_interrupt, .cpu_exec_interrupt = x86_cpu_exec_interrupt, + .do_unaligned_access = x86_cpu_do_unaligned_access, .debug_excp_handler = breakpoint_handler, .debug_check_breakpoint = x86_debug_check_breakpoint, #endif /* !CONFIG_USER_ONLY */ diff --git a/target/i386/tcg/translate.c b/target/i386/tcg/translate.c index b7972f0ff5..a98bcfca8f 100644 --- a/target/i386/tcg/translate.c +++ b/target/i386/tcg/translate.c @@ -2731,21 +2731,23 @@ static inline void gen_stq_env_A0(DisasContext *s, int offset) tcg_gen_qemu_st_i64(s->tmp1_i64, s->A0, s->mem_index, MO_LEUQ); } -static inline void gen_ldo_env_A0(DisasContext *s, int offset) +static inline void gen_ldo_env_A0(DisasContext *s, int offset, bool align) { int mem_index = s->mem_index; - tcg_gen_qemu_ld_i64(s->tmp1_i64, s->A0, mem_index, MO_LEUQ); + tcg_gen_qemu_ld_i64(s->tmp1_i64, s->A0, mem_index, + MO_LEUQ | (align ? MO_ALIGN_16 : 0)); tcg_gen_st_i64(s->tmp1_i64, cpu_env, offset + offsetof(ZMMReg, ZMM_Q(0))); tcg_gen_addi_tl(s->tmp0, s->A0, 8); tcg_gen_qemu_ld_i64(s->tmp1_i64, s->tmp0, mem_index, MO_LEUQ); tcg_gen_st_i64(s->tmp1_i64, cpu_env, offset + offsetof(ZMMReg, ZMM_Q(1))); } -static inline void gen_sto_env_A0(DisasContext *s, int offset) +static inline void gen_sto_env_A0(DisasContext *s, int offset, bool align) { int mem_index = s->mem_index; tcg_gen_ld_i64(s->tmp1_i64, cpu_env, offset + offsetof(ZMMReg, ZMM_Q(0))); - tcg_gen_qemu_st_i64(s->tmp1_i64, s->A0, mem_index, MO_LEUQ); + tcg_gen_qemu_st_i64(s->tmp1_i64, s->A0, mem_index, + MO_LEUQ | (align ? MO_ALIGN_16 : 0)); tcg_gen_addi_tl(s->tmp0, s->A0, 8); tcg_gen_ld_i64(s->tmp1_i64, cpu_env, offset + offsetof(ZMMReg, ZMM_Q(1))); tcg_gen_qemu_st_i64(s->tmp1_i64, s->tmp0, mem_index, MO_LEUQ); @@ -3054,7 +3056,7 @@ static const struct SSEOpHelper_epp sse_op_table6[256] = { [0x25] = SSE41_OP(pmovsxdq), [0x28] = SSE41_OP(pmuldq), [0x29] = SSE41_OP(pcmpeqq), - [0x2a] = SSE41_SPECIAL, /* movntqda */ + [0x2a] = SSE41_SPECIAL, /* movntdqa */ [0x2b] = SSE41_OP(packusdw), [0x30] = SSE41_OP(pmovzxbw), [0x31] = SSE41_OP(pmovzxbd), @@ -3194,17 +3196,17 @@ static void gen_sse(CPUX86State *env, DisasContext *s, int b, break; case 0x1e7: /* movntdq */ case 0x02b: /* movntps */ - case 0x12b: /* movntps */ + case 0x12b: /* movntpd */ if (mod == 3) goto illegal_op; gen_lea_modrm(env, s, modrm); - gen_sto_env_A0(s, offsetof(CPUX86State, xmm_regs[reg])); + gen_sto_env_A0(s, offsetof(CPUX86State, xmm_regs[reg]), true); break; case 0x3f0: /* lddqu */ if (mod == 3) goto illegal_op; gen_lea_modrm(env, s, modrm); - gen_ldo_env_A0(s, offsetof(CPUX86State, xmm_regs[reg])); + gen_ldo_env_A0(s, offsetof(CPUX86State, xmm_regs[reg]), false); break; case 0x22b: /* movntss */ case 0x32b: /* movntsd */ @@ -3273,7 +3275,9 @@ static void gen_sse(CPUX86State *env, DisasContext *s, int b, case 0x26f: /* movdqu xmm, ea */ if (mod != 3) { gen_lea_modrm(env, s, modrm); - gen_ldo_env_A0(s, offsetof(CPUX86State, xmm_regs[reg])); + gen_ldo_env_A0(s, offsetof(CPUX86State, xmm_regs[reg]), + /* movaps, movapd, movdqa */ + b == 0x028 || b == 0x128 || b == 0x16f); } else { rm = (modrm & 7) | REX_B(s); gen_op_movo(s, offsetof(CPUX86State, xmm_regs[reg]), @@ -3331,7 +3335,8 @@ static void gen_sse(CPUX86State *env, DisasContext *s, int b, case 0x212: /* movsldup */ if (mod != 3) { gen_lea_modrm(env, s, modrm); - gen_ldo_env_A0(s, offsetof(CPUX86State, xmm_regs[reg])); + gen_ldo_env_A0(s, offsetof(CPUX86State, xmm_regs[reg]), + !(s->prefix & PREFIX_VEX)); } else { rm = (modrm & 7) | REX_B(s); gen_op_movl(s, offsetof(CPUX86State, xmm_regs[reg].ZMM_L(0)), @@ -3373,7 +3378,8 @@ static void gen_sse(CPUX86State *env, DisasContext *s, int b, case 0x216: /* movshdup */ if (mod != 3) { gen_lea_modrm(env, s, modrm); - gen_ldo_env_A0(s, offsetof(CPUX86State, xmm_regs[reg])); + gen_ldo_env_A0(s, offsetof(CPUX86State, xmm_regs[reg]), + !(s->prefix & PREFIX_VEX)); } else { rm = (modrm & 7) | REX_B(s); gen_op_movl(s, offsetof(CPUX86State, xmm_regs[reg].ZMM_L(1)), @@ -3465,7 +3471,9 @@ static void gen_sse(CPUX86State *env, DisasContext *s, int b, case 0x27f: /* movdqu ea, xmm */ if (mod != 3) { gen_lea_modrm(env, s, modrm); - gen_sto_env_A0(s, offsetof(CPUX86State, xmm_regs[reg])); + gen_sto_env_A0(s, offsetof(CPUX86State, xmm_regs[reg]), + /* movaps, movapd, movdqa */ + b == 0x029 || b == 0x129 || b == 0x17f); } else { rm = (modrm & 7) | REX_B(s); gen_op_movo(s, offsetof(CPUX86State, xmm_regs[rm]), @@ -3621,7 +3629,11 @@ static void gen_sse(CPUX86State *env, DisasContext *s, int b, if (mod != 3) { gen_lea_modrm(env, s, modrm); op2_offset = offsetof(CPUX86State,xmm_t0); - gen_ldo_env_A0(s, op2_offset); + if ((b >> 8) & 1) { + gen_ldo_env_A0(s, op2_offset, true); + } else { + gen_ldq_env_A0(s, op2_offset); + } } else { rm = (modrm & 7) | REX_B(s); op2_offset = offsetof(CPUX86State,xmm_regs[rm]); @@ -3806,11 +3818,12 @@ static void gen_sse(CPUX86State *env, DisasContext *s, int b, tcg_gen_st16_tl(s->tmp0, cpu_env, op2_offset + offsetof(ZMMReg, ZMM_W(0))); break; - case 0x2a: /* movntqda */ - gen_ldo_env_A0(s, op1_offset); + case 0x2a: /* movntdqa */ + gen_ldo_env_A0(s, op1_offset, true); return; default: - gen_ldo_env_A0(s, op2_offset); + gen_ldo_env_A0(s, op2_offset, + !(s->prefix & PREFIX_VEX)); } } } else { @@ -4351,7 +4364,7 @@ static void gen_sse(CPUX86State *env, DisasContext *s, int b, } else { op2_offset = offsetof(CPUX86State,xmm_t0); gen_lea_modrm(env, s, modrm); - gen_ldo_env_A0(s, op2_offset); + gen_ldo_env_A0(s, op2_offset, !(s->prefix & PREFIX_VEX)); } } else { op1_offset = offsetof(CPUX86State,fpregs[reg].mmx); @@ -4469,7 +4482,7 @@ static void gen_sse(CPUX86State *env, DisasContext *s, int b, break; default: /* 128 bit access */ - gen_ldo_env_A0(s, op2_offset); + gen_ldo_env_A0(s, op2_offset, !(s->prefix & PREFIX_VEX)); break; } } else { diff --git a/target/i386/tcg/user/excp_helper.c b/target/i386/tcg/user/excp_helper.c index cd507e2a1b..b3bdb7831a 100644 --- a/target/i386/tcg/user/excp_helper.c +++ b/target/i386/tcg/user/excp_helper.c @@ -48,3 +48,10 @@ void x86_cpu_record_sigsegv(CPUState *cs, vaddr addr, cpu_loop_exit_restore(cs, ra); } + +void x86_cpu_record_sigbus(CPUState *cs, vaddr addr, + MMUAccessType access_type, uintptr_t ra) +{ + X86CPU *cpu = X86_CPU(cs); + handle_unaligned_access(&cpu->env, addr, access_type, ra); +} -- 2.37.2
