On 12/11/2014 21:07, Paolo Bonzini wrote:
>
>
> On 08/10/2014 12:55, Leon Alrae wrote:
>> case OPC_LUI:
>> - tcg_gen_movi_tl(cpu_gpr[rt], imm << 16);
>> - MIPS_DEBUG("lui %s, " TARGET_FMT_lx, regnames[rt], uimm);
>> + if (rs != 0 && (ctx->insn_flags & ISA_MIPS32R6)) {
>> + /* OPC_AUI */
>> + tcg_gen_addi_tl(cpu_gpr[rt], cpu_gpr[rs], imm << 16);
>> + tcg_gen_ext32s_tl(cpu_gpr[rt], cpu_gpr[rt]);
>> + MIPS_DEBUG("aui %s, %s, %04x", regnames[rt], regnames[rs], imm);
>> + } else {
>> + tcg_gen_movi_tl(cpu_gpr[rt], imm << 16);
>> + MIPS_DEBUG("lui %s, " TARGET_FMT_lx, regnames[rt], uimm);
>> + }
>> break;
>
> Coverity reported a
>
> gen_logic_imm(ctx, OPC_LUI, rs, -1, imm);
>
> where the -1 probably has to become zero now.
This line is from microMIPS decoder. We don't support microMIPS R6 in
QEMU so I think this isn't an issue at the moment as (ctx->insn_flags &
ISA_MIPS32R6) should always be false on a CPU with existing
implementation of microMIPS. Nevertheless, I agree this should be 0 for
safety, I'll correct it in 2.3.
Thanks for pointing this out.
Regards,
Leon
