Tested with linux guest. Not sure how to check actual performance affect of this. Checked with the previously send traceevent that the kvm ioctl to start/stop dirty logging is being called. (KVM_SET_USER_MEMORY_REGION).
Signed-off-by: Alon Levy <al...@redhat.com> --- These are the two patches to queue instead of the previously squashed patch. hw/qxl.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/hw/qxl.c b/hw/qxl.c index 8777ba9..5a7be60 100644 --- a/hw/qxl.c +++ b/hw/qxl.c @@ -932,6 +932,7 @@ static void qxl_enter_vga_mode(PCIQXLDevice *d) qemu_spice_create_host_primary(&d->ssd); d->mode = QXL_MODE_VGA; memset(&d->ssd.dirty, 0, sizeof(d->ssd.dirty)); + vga_dirty_log_start(&d->vga); } static void qxl_exit_vga_mode(PCIQXLDevice *d) @@ -940,6 +941,7 @@ static void qxl_exit_vga_mode(PCIQXLDevice *d) return; } trace_qxl_exit_vga_mode(d->id); + vga_dirty_log_stop(&d->vga); qxl_destroy_primary(d, QXL_SYNC); } -- 1.7.10.1