Re: [PATCH] migration: Fix the minus value for compressed_size

["Dr. David Alan Gilbert" <dgilbert@redhat.com>]

* Zhenzhong Duan (zhenzhong.duan@intel.com) wrote:
> When update_compress_thread_counts() is called first time, there is
> no data stream yet. We see compression_counters.compressed_size
> becomes minus value shortly.
> 
> Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
> ---
>  migration/ram.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/migration/ram.c b/migration/ram.c
> index dc1de9ddbc68..510db95cdc36 100644
> --- a/migration/ram.c
> +++ b/migration/ram.c
> @@ -1443,6 +1443,10 @@ static bool do_compress_ram_page(QEMUFile *f, z_stream *stream, RAMBlock *block,
>  static void
>  update_compress_thread_counts(const CompressParam *param, int bytes_xmit)
>  {
> +    if (bytes_xmit <= 0) {
> +        return;
> +    }

What's the call path where that happens? The only place I see
bytes_xmit being less than 0 is in compress_page_with_multi_thread where
it's initialised to -1 - but it's always updated before the call
to update_compress_thread_counts.

I wonder if the real problem is:

    compression_counters.compressed_size += bytes_xmit - 8;

Is bytes_xmit being less than 8 for some reason?

Dave

>      ram_transferred_add(bytes_xmit);
>  
>      if (param->zero_page) {
> -- 
> 2.25.1
> 
-- 
Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK