Steve Sistare <steven.sist...@oracle.com> writes:
> CPR preserves memory in place, so there is no need to track dirty memory.
> By skipping it, CPR can support devices that do not support tracking.
>
> Signed-off-by: Steve Sistare <steven.sist...@oracle.com>
> ---
> system/memory.c | 11 +++++++++++
> 1 file changed, 11 insertions(+)
>
> diff --git a/system/memory.c b/system/memory.c
> index b7548bf112..aef584e638 100644
> --- a/system/memory.c
> +++ b/system/memory.c
> @@ -27,6 +27,7 @@
>
> #include "exec/memory-internal.h"
> #include "exec/ram_addr.h"
> +#include "migration/misc.h"
> #include "sysemu/kvm.h"
> #include "sysemu/runstate.h"
> #include "sysemu/tcg.h"
> @@ -2947,6 +2948,11 @@ bool memory_global_dirty_log_start(unsigned int flags,
> Error **errp)
>
> assert(flags && !(flags & (~GLOBAL_DIRTY_MASK)));
>
> + /* CPR preserves memory in place, so no need to track dirty memory */
> + if (migrate_mode() != MIG_MODE_NORMAL) {
> + return true;
> + }
How this interacts with DIRTY_RATE and DIRTY_LIMIT? The former at least
seems to never overlap with CPR, right? I'm wondering whether this check
would be more appropriate up in ram.c along with the similar
migrate_background_snapshot() check.
(I wish we had made the global_dirty_log_change() function a bit more
flexible. It would have been a nice place to put this and the snapshot
check. Not worth the risk of changing it now...)
Also, not tracking dirty memory implies also not doing the bitmap sync?
We skip it for bg_snapshot, but not for CPR.
> +
> if (vmstate_change) {
> /* If there is postponed stop(), operate on it first */
> postponed_stop_flags &= ~flags;
> @@ -3021,6 +3027,11 @@ static void memory_vm_change_state_handler(void
> *opaque, bool running,
>
> void memory_global_dirty_log_stop(unsigned int flags)
> {
> + /* CPR preserves memory in place, so no need to track dirty memory */
> + if (migrate_mode() != MIG_MODE_NORMAL) {
> + return;
> + }
> +
> if (!runstate_is_running()) {
> /* Postpone the dirty log stop, e.g., to when VM starts again */
> if (vmstate_change) {
