Re: [Qemu-devel] [PATCH 1/1] migration: fix ram decompression race deadlock

[Amit Shah]

Adding Liang Li to CC for his comments as the author of the feature.

On (Fri) 13 May 2016 [10:27:01], Denis V. Lunev wrote:
> From: Maxim Nestratov <mnestra...@virtuozzo.com>
> 
> There is a race in between do_data_decompress and start_decompression.
> 
> do_data_decompress()
>     while (!quit_decomp_thread) {
>         qemu_mutex_lock(&param->mutex);
>         while (!param->start && !quit_decomp_thread) {
>             qemu_cond_wait(&param->cond, &param->mutex);
>             ...
>             param->start = false;
>         }
>         qemu_mutex_unlock(&param->mutex);
>         [ preempted here, start_decompression() is executed ]
>     }
> 
> start_decompression()
> {
>     qemu_mutex_lock(&param->mutex);
>     param->start = true;
>     qemu_cond_signal(&param->cond);
>     qemu_mutex_unlock(&param->mutex);
> }
> 
> In this case do_data_decompress will never enter inner loop again and
> will eat 100% CPU. The patch fixes this problem by correcting while loop
> where we wait for condition only and other actions are moved out of it.
> 
> Signed-off-by: Maxim Nestratov <mnestra...@virtuozzo.com>
> Signed-off-by: Denis V. Lunev <d...@openvz.org>
> CC: Juan Quintela <quint...@redhat.com>
> CC: Amit Shah <amit.s...@redhat.com>
> ---
>  migration/ram.c | 22 +++++++++++-----------
>  1 file changed, 11 insertions(+), 11 deletions(-)
> 
> diff --git a/migration/ram.c b/migration/ram.c
> index 3f05738..579bfc0 100644
> --- a/migration/ram.c
> +++ b/migration/ram.c
> @@ -2193,18 +2193,18 @@ static void *do_data_decompress(void *opaque)
>          qemu_mutex_lock(&param->mutex);
>          while (!param->start && !quit_decomp_thread) {
>              qemu_cond_wait(&param->cond, &param->mutex);
> -            pagesize = TARGET_PAGE_SIZE;
> -            if (!quit_decomp_thread) {
> -                /* uncompress() will return failed in some case, especially
> -                 * when the page is dirted when doing the compression, it's
> -                 * not a problem because the dirty page will be retransferred
> -                 * and uncompress() won't break the data in other pages.
> -                 */
> -                uncompress((Bytef *)param->des, &pagesize,
> -                           (const Bytef *)param->compbuf, param->len);
> -            }
> -            param->start = false;
>          }
> +        pagesize = TARGET_PAGE_SIZE;
> +        if (!quit_decomp_thread) {
> +           /* uncompress() will return failed in some case, especially
> +            * when the page is dirted when doing the compression, it's
> +            * not a problem because the dirty page will be retransferred
> +            * and uncompress() won't break the data in other pages.
> +            */
> +            uncompress((Bytef *)param->des, &pagesize,
> +                       (const Bytef *)param->compbuf, param->len);
> +        }
> +        param->start = false;
>          qemu_mutex_unlock(&param->mutex);
>      }
>  
> -- 
> 2.1.4
> 

                Amit