On Fri, Mar 01, 2024 at 02:28:25AM +0000, Hao Xiang wrote:
> 1. Add a dedicated handler for MigrationOps::ram_save_target_page in
> multifd live migration.
> 2. Refactor ram_save_target_page_legacy so that the legacy and multifd
> handlers don't have internal functions calling into each other.
>
> Signed-off-by: Hao Xiang <hao.xiang@bytedance.com>
> Reviewed-by: Fabiano Rosas <farosas@suse.de>
> Message-Id: <20240226195654.934709-4-hao.xiang@bytedance.com>
> ---
> migration/ram.c | 43 ++++++++++++++++++++++++++++++-------------
> 1 file changed, 30 insertions(+), 13 deletions(-)
>
> diff --git a/migration/ram.c b/migration/ram.c
> index e1fa229acf..f9d6ea65cc 100644
> --- a/migration/ram.c
> +++ b/migration/ram.c
> @@ -1122,10 +1122,6 @@ static int save_zero_page(RAMState *rs, PageSearchStatus *pss,
> QEMUFile *file = pss->pss_channel;
> int len = 0;
>
> - if (migrate_zero_page_detection() == ZERO_PAGE_DETECTION_NONE) {
> - return 0;
> - }
We need to keep this to disable zero-page-detect on !multifd?
> -
> if (!buffer_is_zero(p, TARGET_PAGE_SIZE)) {
> return 0;
> }
> @@ -2045,7 +2041,6 @@ static bool save_compress_page(RAMState *rs, PageSearchStatus *pss,
> */
> static int ram_save_target_page_legacy(RAMState *rs, PageSearchStatus *pss)
> {
> - RAMBlock *block = pss->block;
> ram_addr_t offset = ((ram_addr_t)pss->page) << TARGET_PAGE_BITS;
> int res;
>
> @@ -2061,17 +2056,34 @@ static int ram_save_target_page_legacy(RAMState *rs, PageSearchStatus *pss)
> return 1;
> }
>
> + return ram_save_page(rs, pss);
> +}
> +
> +/**
> + * ram_save_target_page_multifd: send one target page to multifd workers
> + *
> + * Returns 1 if the page was queued, -1 otherwise.
> + *
> + * @rs: current RAM state
> + * @pss: data about the page we want to send
> + */
> +static int ram_save_target_page_multifd(RAMState *rs, PageSearchStatus *pss)
> +{
> + RAMBlock *block = pss->block;
> + ram_addr_t offset = ((ram_addr_t)pss->page) << TARGET_PAGE_BITS;
> +
> /*
> - * Do not use multifd in postcopy as one whole host page should be
> - * placed. Meanwhile postcopy requires atomic update of pages, so even
> - * if host page size == guest page size the dest guest during run may
> - * still see partially copied pages which is data corruption.
> + * Backward compatibility support. While using multifd live
We can also avoid mentioning "compatibility support" here - it's a
parameter, user can legally set it to anything.
> + * migration, we still need to handle zero page checking on the
> + * migration main thread.
> */
> - if (migrate_multifd() && !migration_in_postcopy()) {
> - return ram_save_multifd_page(block, offset);
> + if (migrate_zero_page_detection() == ZERO_PAGE_DETECTION_LEGACY) {
> + if (save_zero_page(rs, pss, offset)) {
> + return 1;
> + }
> }
>
> - return ram_save_page(rs, pss);
> + return ram_save_multifd_page(block, offset);
> }
>
> /* Should be called before sending a host page */
> @@ -2983,7 +2995,12 @@ static int ram_save_setup(QEMUFile *f, void *opaque)
> }
>
> migration_ops = g_malloc0(sizeof(MigrationOps));
> - migration_ops->ram_save_target_page = ram_save_target_page_legacy;
> +
> + if (migrate_multifd()) {
> + migration_ops->ram_save_target_page = ram_save_target_page_multifd;
> + } else {
> + migration_ops->ram_save_target_page = ram_save_target_page_legacy;
> + }
>
> bql_unlock();
> ret = multifd_send_sync_main();
> --
> 2.30.2
>
--
Peter Xu