[Xen-devel] Re: [PATCH 2/2] xen: Introduce VGA sync dirty bitmap support

[Anthony PERARD <anthony.perard@xxxxxxxxxx>]

On Wed, 12 Jan 2011, Jan Kiszka wrote:

> Am 11.01.2011 16:37, anthony.perard@xxxxxxxxxx wrote:
> > From: Anthony PERARD <anthony.perard@xxxxxxxxxx>
> >
> > This patch introduces phys memory client for Xen.
> >
> > Only sync dirty_bitmap and set_memory are actually implemented.
> > migration_log will stay empty for the moment.
> >
> > Xen can only log one range for bit change, so only the range in the
> > first call will be synced.
> >
> > Signed-off-by: Anthony PERARD <anthony.perard@xxxxxxxxxx>
> > ---
> >  hw/vga.c   |    7 ++
> >  hw/xen.h   |    2 +
> >  xen-all.c  |  233 
> > ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
> >  xen-stub.c |   11 +++
> >  4 files changed, 253 insertions(+), 0 deletions(-)
> >
> > diff --git a/hw/vga.c b/hw/vga.c
> > index c057f4f..5f7a181 100644
> > --- a/hw/vga.c
> > +++ b/hw/vga.c
> > @@ -29,6 +29,7 @@
> >  #include "pixel_ops.h"
> >  #include "qemu-timer.h"
> >  #include "kvm.h"
> > +#include "xen.h"
> >
> >  //#define DEBUG_VGA
> >  //#define DEBUG_VGA_MEM
> > @@ -1599,6 +1600,9 @@ void vga_dirty_log_start(VGACommonState *s)
> >  {
> >      if (kvm_enabled() && s->map_addr)
> >          kvm_log_start(s->map_addr, s->map_end - s->map_addr);
> > +    if (xen_enabled() && s->map_addr) {
> > +        xen_log_start(s->map_addr, s->map_end - s->map_addr);
> > +    }
> >
> >      if (kvm_enabled() && s->lfb_vram_mapped) {
> >          kvm_log_start(isa_mem_base + 0xa0000, 0x8000);
> > @@ -1616,6 +1620,9 @@ void vga_dirty_log_stop(VGACommonState *s)
> >  {
> >      if (kvm_enabled() && s->map_addr)
> >     kvm_log_stop(s->map_addr, s->map_end - s->map_addr);
> > +    if (xen_enabled() && s->map_addr) {
> > +        xen_log_stop(s->map_addr, s->map_end - s->map_addr);
> > +    }
> >
> >      if (kvm_enabled() && s->lfb_vram_mapped) {
> >     kvm_log_stop(isa_mem_base + 0xa0000, 0x8000);
>
> This is probably the right time to make dirty_log_start/stop callbacks
> in CPUPhysMemoryClient as well. Would remove any KVM or Xen reference
> from vga code.
>
> We just need to think about how to deal with the quirks of its users:
> KVM requires the isa vram to be reported in two chunks, Xen can't handle
> more than one region at all. If Xen is able to filter out those events
> it can handle, we could replace kvm_log_start/stop with some
> cpu_notify_log_start/stop.

The first thing I think of, is to handle only the first call of
_log_start.

The other choice is to sync dirty bitmap only when a call to
xc_domain_add_to_physmap is working. This call is done inside the
set_memory client when the flags is IO_MEM_RAM.

-- 
Anthony PERARD

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel