Re: [PATCH 4/6] migration/multifd: Zero page transmission on the multifd thread.

[Peter Xu <peterx@redhat.com>]

On Tue, Feb 06, 2024 at 11:19:06PM +0000, Hao Xiang wrote:
> This implements the zero page detection and handling on the multifd
> threads.
> 
> Signed-off-by: Hao Xiang <hao.xiang@bytedance.com>
> ---
>  migration/multifd.c | 62 +++++++++++++++++++++++++++++++++++++++++----
>  migration/multifd.h |  5 ++++
>  2 files changed, 62 insertions(+), 5 deletions(-)
> 
> diff --git a/migration/multifd.c b/migration/multifd.c
> index a20d0ed10e..c031f947c7 100644
> --- a/migration/multifd.c
> +++ b/migration/multifd.c
> @@ -11,6 +11,7 @@
>   */
>  
>  #include "qemu/osdep.h"
> +#include "qemu/cutils.h"
>  #include "qemu/rcu.h"
>  #include "exec/target_page.h"
>  #include "sysemu/sysemu.h"
> @@ -278,6 +279,12 @@ static void multifd_send_fill_packet(MultiFDSendParams *p)
>  
>          packet->offset[i] = cpu_to_be64(temp);
>      }
> +    for (i = 0; i < p->zero_num; i++) {
> +        /* there are architectures where ram_addr_t is 32 bit */
> +        uint64_t temp = p->zero[i];
> +
> +        packet->offset[p->normal_num + i] = cpu_to_be64(temp);
> +    }
>  }

Please be noted taht p->normal_num will be dropped very soon, see:

https://lore.kernel.org/all/20240202102857.110210-6-peterx@redhat.com/

Please use p->pages->num instead.

This patch also relies on some changes in previous patch.. IMHO we can
split the patch better in this way:

  - Patch 1: Add new parameter "zero-page-detection", support "none",
    "legacy".  You'll need to implement "none" here that we skip zero page
    by returning 0 in save_zero_page() if "none".

  - Patch 2: Add new "multifd" mode in above, implement it in the same
    patch completely.

  - Patch 3: introduce ram_save_target_page_multifd()

  - Patch 4: test case

If you want to add "zeros" accounting, that can be done as more patches on
top.

Thanks,

-- 
Peter Xu