Re: [Qemu-devel] [PATCH 3/3] tcg: Avoid bouncing tb_lock between tb_gen_code() and tb_add_jump()

[Alex Bennée]

Sergey Fedorov <address@hidden> writes:

> From: Sergey Fedorov <address@hidden>
>
> Signed-off-by: Sergey Fedorov <address@hidden>
> Signed-off-by: Sergey Fedorov <address@hidden>
> ---
>  cpu-exec.c | 15 +++++++++------
>  1 file changed, 9 insertions(+), 6 deletions(-)
>
> diff --git a/cpu-exec.c b/cpu-exec.c
> index dd0bd5007701..54c935039592 100644
> --- a/cpu-exec.c
> +++ b/cpu-exec.c
> @@ -295,7 +295,8 @@ static TranslationBlock *tb_find_slow(CPUState *cpu,
>
>          /* mmap_lock is needed by tb_gen_code, and mmap_lock must be
>           * taken outside tb_lock. As system emulation is currently
> -         * single threaded the locks are NOPs.
> +         * single threaded the locks are NOPs. Both locks are to be
> +         * released at the end of tb_find_fast().
>           */
>          mmap_lock();
>          tb_lock();
> @@ -308,9 +309,6 @@ static TranslationBlock *tb_find_slow(CPUState *cpu,
>              /* if no translated code available, then translate it now */
>              tb = tb_gen_code(cpu, pc, cs_base, flags, 0);
>          }
> -
> -        tb_unlock();
> -        mmap_unlock();

Hmm pushing these outside of tb_find_slow() makes me uncomfortable. I
guess tb_find_fast/slow are intimately tied together but the idea of
taking locks which are the responsibility of the calling function to
clear seems ugly to me.

>      }
>
>      /* We add the TB in the virtual pc hash table for the fast lookup */
> @@ -354,10 +352,15 @@ static inline TranslationBlock *tb_find_fast(CPUState 
> *cpu,
>  #endif
>      /* See if we can patch the calling TB. */
>      if (*last_tb && !qemu_loglevel_mask(CPU_LOG_TB_NOCHAIN)) {
> -        tb_lock();
> +        if (!tb_lock_locked()) {
> +            tb_lock();
> +        }
>          tb_add_jump(*last_tb, tb_exit, tb);
> -        tb_unlock();
>      }
> +
> +    tb_lock_reset();
> +    mmap_lock_reset();
> +
>      return tb;
>  }


--
Alex Bennée