On Mon, Oct 17, 2011 at 8:48 PM, Avi Kivity <a...@redhat.com> wrote:
> On 10/17/2011 09:13 PM, Blue Swirl wrote:
>> >
>> > - ret = cpu_physical_memory_get_dirty(page, VGA_DIRTY_FLAG);
>> > + ret = memory_region_get_dirty(&s->vram_mem, page, DIRTY_MEMORY_VGA);
>> > for (off = 0; off < TARGET_PAGE_SIZE * 4; off += TARGET_PAGE_SIZE) {
>> > - ret |= cpu_physical_memory_get_dirty(page24 + off,
>> > VGA_DIRTY_FLAG);
>> > - ret |= cpu_physical_memory_get_dirty(cpage + off, VGA_DIRTY_FLAG);
>> > + ret |= memory_region_get_dirty(&s->vram_mem, page24 + off,
>> > + DIRTY_MEMORY_VGA);
>> > + ret |= memory_region_get_dirty(&s->vram_mem, cpage + off,
>> > + DIRTY_MEMORY_VGA);
>>
>> Like memory_region_set_dirty(), this should be changed to take a range
>> so that the for loops and PAGE_SIZEs can be dropped.
>
> Agree. Lets' keep it outside this conversion patchset though.
Yes.
> What would the API look like? accept a range, return true if any of the
> pages in the range are dirty?
That would be 1:1 conversion but maybe better approach would be to
return the first dirty address (or zero if not found), then various
bitmap speedups may be available.
