Re: [RFC PATCH 5/8] qtest/libqos: Restrict CPU I/O instructions

[Laszlo Ersek <lersek@redhat.com>]

(+Peter, comment below)

On 03/15/21 00:29, Philippe Mathieu-Daudé wrote:
> Restrict CPU I/O instructions to architectures providing
> I/O bus.
> 
> Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
> ---
>  tests/qtest/libqos/fw_cfg.h | 3 +++
>  tests/qtest/libqos/fw_cfg.c | 2 ++
>  2 files changed, 5 insertions(+)
> 
> diff --git a/tests/qtest/libqos/fw_cfg.h b/tests/qtest/libqos/fw_cfg.h
> index c6a7cf8cf05..3bfb6d6d55b 100644
> --- a/tests/qtest/libqos/fw_cfg.h
> +++ b/tests/qtest/libqos/fw_cfg.h
> @@ -36,6 +36,8 @@ size_t qfw_cfg_get_file(QFWCFG *fw_cfg, const char *filename,
>  
>  QFWCFG *mm_fw_cfg_init(QTestState *qts, uint64_t base);
>  void mm_fw_cfg_uninit(QFWCFG *fw_cfg);
> +
> +#ifdef TARGET_HAS_IOPORT
>  QFWCFG *io_fw_cfg_init(QTestState *qts, uint16_t base);
>  void io_fw_cfg_uninit(QFWCFG *fw_cfg);
>  
> @@ -48,6 +50,7 @@ static inline void pc_fw_cfg_uninit(QFWCFG *fw_cfg)
>  {
>      io_fw_cfg_uninit(fw_cfg);
>  }
> +#endif /* TARGET_HAS_IOPORT */
>  
>  G_DEFINE_AUTOPTR_CLEANUP_FUNC(QFWCFG, mm_fw_cfg_uninit)
>  
> diff --git a/tests/qtest/libqos/fw_cfg.c b/tests/qtest/libqos/fw_cfg.c
> index 6b8e1babe51..db2b83f5212 100644
> --- a/tests/qtest/libqos/fw_cfg.c
> +++ b/tests/qtest/libqos/fw_cfg.c
> @@ -131,6 +131,7 @@ void mm_fw_cfg_uninit(QFWCFG *fw_cfg)
>      g_free(fw_cfg);
>  }
>  
> +#ifdef TARGET_HAS_IOPORT
>  static void io_fw_cfg_select(QFWCFG *fw_cfg, uint16_t key)
>  {
>      qtest_outw(fw_cfg->qts, fw_cfg->base, key);
> @@ -162,3 +163,4 @@ void io_fw_cfg_uninit(QFWCFG *fw_cfg)
>  {
>      g_free(fw_cfg);
>  }
> +#endif /* TARGET_HAS_IOPORT */
> 

I'm not sure the macro name is ideal; the PCI host on aarch64/"virt"
emulates IO Ports (it's possible to allocate PCI IO resources on
"virt"). From patch#3, TARGET_HAS_IOPORT does not seem to extend to arm64.

I guess the intent is OK in both patches #3 and #5.

Thanks
Laszlo