On 6/14/22 02:05, Song Gao wrote:
@@ -172,17 +173,20 @@ static void loongarch_cpu_do_interrupt(CPUState *cs)
update_badinstr = 0;
break;
case EXCCODE_ADEM:
+ case EXCCODE_BCE:
case EXCCODE_SYS:
case EXCCODE_BRK:
+ case EXCCODE_INE:
+ case EXCCODE_IPE:
+ case EXCCODE_FPE:
+ env->badvaddr = env->pc;
+ QEMU_FALLTHROUGH;
This is incorrect still.
(1) env->badaddr (in this patch renamed badvaddr) is actually unused prior to this patch
and should go away. It seems to have been copied from RISC-V? The correct LoongArch
variable is env->CSR_BADV (see raise_mmu_exception in tlb_helper.c).
(2) EXCCODE_ADEM is on the wrong side of this FALLTHROUGH. This is the exception raised
by TLB faults, and should retain the BADV address of the fault, not the faulting instruction.
Also, this patch is trying to do too many things at once. Please split it into smaller
logical changes. Any bug fixes for the system code, for instance raising EXCCODE_BCE
instead of EXCCODE_ADEM for helper_asrtle_d should be completely separated.
r~
