Re: [PATCH] target/s390x: Exit tb after executing ex_value

[Sven Schnelle <svens@linux.ibm.com>]

Hi Richard,

Richard Henderson <richard.henderson@linaro.org> writes:

> When EXECUTE sets ex_value to interrupt the constructed instruction,
> we implicitly disable interrupts so that the value is not corrupted.
> Exit to the main loop after execution, so that we re-evaluate any
> pending interrupts.
>
> Reported-by: Sven Schnelle <svens@linux.ibm.com>
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> ---
>
> Hi Sven.  Will you test this vs your testcase?  Thanks,

Of course, i'm happy if someone fixes this so i don't have to. :-)

Unfortunately it doesn't fix the issue:

exec_tb_exit tb:(nil) flags=0x0
exec_tb tb:0x3ff35c66f00 pc=0x400
exec_tb tb:0x3ff3410f300 pc=0x1edf7f8
tcg_handle_interrupt: 2
exec_tb_exit tb:0x3ff340d2d00 flags=0x3
ignoring irq during EX
ignoring irq during EX
exec_tb tb:0x3ff340d2d00 pc=0x1edf810

writing dc->base.is_jmp to the qemu log shows:

s390x_tr_translate_insn: is_jmp: 3
s390x_tr_translate_insn: is_jmp: 3
s390x_tr_translate_insn: is_jmp: 3
s390x_tr_translate_insn: is_jmp: 3
s390x_tr_translate_insn: is_jmp: 3
s390x_tr_translate_insn: is_jmp: 3
[..]

So is_jump is always 3, which is DISAS_TARGET_0. I think the
if (dc->base.is_jmp == DISAS_NEXT) condition therefore never matches.