-----Original Message-----
From: Richard Henderson
Sent: Friday, July 3, 2020 2:54 AM
To: Wu, Wentong <wentong.wu@intel.com>; qemu-devel@nongnu.org
Cc: qemu-trivial@nongnu.org; marex@denx.de; crwulff@gmail.com; peter.maydell@linaro.org
Subject: Re: [PATCH 3/3] target/nios2: Use gen_io_start around wrctl instruction
On 6/29/20 9:05 AM, Wentong Wu wrote:
> wrctl instruction on nios2 target will cause checking cpu interrupt
> but tcg_handle_interrupt() will call cpu_abort() if the CPU gets an
> interrupt while it's not in 'can do IO'
> state, so add gen_io_start around wrctl instruction. Also at the same
> time, end the onging TB with DISAS_UPDATE.
>
> ---
> target/nios2/translate.c | 5 +++++
> 1 file changed, 5 insertions(+)
>
> diff --git a/target/nios2/translate.c b/target/nios2/translate.c index
> 83c10eb2..51347ada 100644
> --- a/target/nios2/translate.c
> +++ b/target/nios2/translate.c
> @@ -32,6 +32,7 @@
> #include "exec/cpu_ldst.h"
> #include "exec/translator.h"
> #include "qemu/qemu-print.h"
> +#include "exec/gen-icount.h"
>
> /* is_jmp field values */
> #define DISAS_JUMP DISAS_TARGET_0 /* only pc was modified dynamically */
> @@ -518,7 +519,11 @@ static void wrctl(DisasContext *dc, uint32_t code, uint32_t flags)
> /* If interrupts were enabled using WRCTL, trigger them. */ #if
> !defined(CONFIG_USER_ONLY)
> if ((instr.imm5 + CR_BASE) == CR_STATUS) {
> + if (tb_cflags(dc->tb) & CF_USE_ICOUNT) {
> + gen_io_start();
> + }
> gen_helper_check_interrupts(dc->cpu_env);
> + dc->is_jmp = DISAS_UPDATE;
> }
> #endif
This isn't right. Not so much the gen_io_start portion, but the entire existence of helper_check_interrupt.
The correct way to acknowledge interrupts after changing an interrupt mask bit is to exit the TB back to the cpu main loop.
Which you are doing here with DISAS_UPDATE, so that part is fine. (Although you could merge that into the switch statement above.)
Looking at nios_pic_cpu_handler, there are two other bugs:
1) Get rid of env->irq_pending and use cpu_interrupt/cpu_reset_interrupt instead
IMHO, we need env->irq_pending, because if ((env->regs[CR_STATUS] & CR_STATUS_PIE) == False), it can indicate there is interrupt pending, we can exit to main loop to handle the interrupt at once if guest code enable irq later.
2) Do not check env->regs[CR_STATUS] & CR_STATUS_PIE. That variable does not belong to the pic and should not be checked there. The check belongs in nios2_cpu_exec_interrupt, and is in fact already there.
But that will cause some assert happen if device trigger irq with env->regs[CR_STATUS] & CR_STATUS_PIE = False
BR