[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH 66/67] target/arm: Move singlestep check from ge
From: |
Peter Maydell |
Subject: |
Re: [Qemu-devel] [PATCH 66/67] target/arm: Move singlestep check from gen_jmp to gen_goto_tb |
Date: |
Fri, 26 Jul 2019 19:13:16 +0100 |
On Fri, 26 Jul 2019 at 18:51, Richard Henderson
<address@hidden> wrote:
>
> We miss quite a number of single-step events by having
> the check in the wrong place.
>
> Signed-off-by: Richard Henderson <address@hidden>
> ---
> target/arm/translate.c | 16 ++++++----------
> 1 file changed, 6 insertions(+), 10 deletions(-)
>
> diff --git a/target/arm/translate.c b/target/arm/translate.c
> index c2b8b86fd2..9ae9b23823 100644
> --- a/target/arm/translate.c
> +++ b/target/arm/translate.c
> @@ -2740,7 +2740,10 @@ static void gen_goto_ptr(void)
> */
> static void gen_goto_tb(DisasContext *s, int n, target_ulong dest)
> {
> - if (use_goto_tb(s, dest)) {
> + if (unlikely(is_singlestepping(s))) {
> + gen_set_pc_im(s, dest);
> + gen_singlestep_exception(s);
> + } else if (use_goto_tb(s, dest)) {
> tcg_gen_goto_tb(n);
> gen_set_pc_im(s, dest);
> tcg_gen_exit_tb(s->base.tb, n);
> @@ -2751,16 +2754,9 @@ static void gen_goto_tb(DisasContext *s, int n,
> target_ulong dest)
> s->base.is_jmp = DISAS_NORETURN;
> }
>
> -static inline void gen_jmp (DisasContext *s, uint32_t dest)
> +static inline void gen_jmp(DisasContext *s, uint32_t dest)
> {
> - if (unlikely(is_singlestepping(s))) {
> - /* An indirect jump so that we still trigger the debug exception. */
> - if (s->thumb)
> - dest |= 1;
> - gen_bx_im(s, dest);
> - } else {
> - gen_goto_tb(s, 0, dest);
> - }
> + gen_goto_tb(s, 0, dest);
> }
>
> static inline void gen_mulxy(TCGv_i32 t0, TCGv_i32 t1, int x, int y)
I haven't tested but I'm not sure this change is right.
The idea of the current code is that we handle generating
the actual singlestep exception in arm_tr_tb_stop() at the
end, rather than in the process of generating code for the
insn. This change means we now do a gen_singlestep_exception()
in the gen_jmp()/gen_goto_tb() part of the code, but we haven't
removed the handling of it in arm_tr_tb_stop(), so we're now
doing this in two places. Why doesn't the current design work?
And if we need to do something different for the route to
"change the PC via gen_jmp()" why don't we need to do it
also when we change the PC via eg gen_bx_im() ?
(Incidentally, the only places other than gen_jmp()
which call gen_goto_tb() are:
* the end-of-TB handling of DISAS_NEXT/DISAS_TOO_MANY
* four places for barrier insns where we use a
gen_goto_tb to end the TB -- this isn't consistent
with how we end the TB for other situations...)
thanks
-- PMM
- [Qemu-devel] [PATCH 58/67] target/arm: Convert T16, push and pop, (continued)
- [Qemu-devel] [PATCH 58/67] target/arm: Convert T16, push and pop, Richard Henderson, 2019/07/26
- [Qemu-devel] [PATCH 57/67] target/arm: Convert T16, nop hints, Richard Henderson, 2019/07/26
- [Qemu-devel] [PATCH 65/67] target/arm: Clean up disas_thumb_insn, Richard Henderson, 2019/07/26
- [Qemu-devel] [PATCH 56/67] target/arm: Convert T16, Reverse bytes, Richard Henderson, 2019/07/26
- [Qemu-devel] [PATCH 54/67] target/arm: Convert T16, extract, Richard Henderson, 2019/07/26
- [Qemu-devel] [PATCH 52/67] target/arm: Convert T16 add, compare, move (two high registers), Richard Henderson, 2019/07/26
- [Qemu-devel] [PATCH 63/67] target/arm: Convert T16, Unconditional branch, Richard Henderson, 2019/07/26
- [Qemu-devel] [PATCH 60/67] target/arm: Convert T16, Miscellaneous 16-bit instructions, Richard Henderson, 2019/07/26
- [Qemu-devel] [PATCH 62/67] target/arm: Convert T16, load (literal), Richard Henderson, 2019/07/26
- [Qemu-devel] [PATCH 66/67] target/arm: Move singlestep check from gen_jmp to gen_goto_tb, Richard Henderson, 2019/07/26
- Re: [Qemu-devel] [PATCH 66/67] target/arm: Move singlestep check from gen_jmp to gen_goto_tb,
Peter Maydell <=
- Re: [Qemu-devel] [PATCH 00/67] target/arm: Convert aa32 base isa to decodetree, no-reply, 2019/07/26