Re: [PATCH 3/3] plugins: avoid failing plugin when CPU is inited several times

[Nikolay Igotti]

Attached to the mail counter.c when running with attached test.c compiled to Linux standalone binary shows failing assert, unless the patch is applied.

вс, 10 мая 2020 г. в 02:00, Emilio G. Cota <address@hidden>:

On Mon, Apr 20, 2020 at 13:04:51 +0300, Nikolay Igotti wrote:
> In linux-user multithreaded scenarious CPU could be inited many times with
> the same id,
> so avoid assertions on already present hashtable entry.
>
> Signed-off-by: Nikolay Igotti <address@hidden>
> ---
>  plugins/core.c | 5 +----
>  1 file changed, 1 insertion(+), 4 deletions(-)
>
> diff --git a/plugins/core.c b/plugins/core.c
> index 51bfc94787..889cc6441a 100644
> --- a/plugins/core.c
> +++ b/plugins/core.c
> @@ -196,13 +196,10 @@ plugin_register_cb_udata(qemu_plugin_id_t id, enum
> qemu_plugin_event ev,
>
>  void qemu_plugin_vcpu_init_hook(CPUState *cpu)
>  {
> -    bool success;
> -
>      qemu_rec_mutex_lock(&plugin.lock);
>      plugin_cpu_update__locked(&cpu->cpu_index, NULL, NULL);
> -    success = g_hash_table_insert(plugin.cpu_ht, &cpu->cpu_index,
> +    g_hash_table_insert(plugin.cpu_ht, &cpu->cpu_index,
>                                    &cpu->cpu_index);
> -    g_assert(success);
>      qemu_rec_mutex_unlock(&plugin.lock);

Do you have a reproducer for this? I'd expect (1) the g_hash_table_remove
call in qemu_plugin_vcpu_exit_hook to clear this entry upon CPU exit,
and (2) no two live CPUs to have the same cpu_index. But maybe assumption
(2) is wrong, or simply (1) does not get called for some exiting CPUs,
in which case the right fix would be to make sure that it does get called
on CPU exit.

Thanks,

                Emilio