On Mon, Aug 19, 2013 at 11:59:25AM +0200, Laszlo Ersek wrote:
> On 08/13/13 00:43, Michael S. Tsirkin wrote:
> > Migration code assumes that each RAM block is a multiple of target page
> > size.
>
> Isn't that a valid assumption, considering the TARGET_PAGE_ALIGN() macro
> call in qemu_ram_alloc_from_ptr() [exec.c]?
>
> > We can fix this in a variety of ways, the simplest way is
> > exporting the required page size so callers can make regions
> > large enough.
> >
> > Signed-off-by: Michael S. Tsirkin <m...@redhat.com>
> > ---
> > arch_init.c | 6 ++++++
> > include/exec/memory.h | 1 +
> > 2 files changed, 7 insertions(+)
> >
> > diff --git a/arch_init.c b/arch_init.c
> > index 68a7ab7..c62778f 100644
> > --- a/arch_init.c
> > +++ b/arch_init.c
> > @@ -150,6 +150,12 @@ int qemu_read_default_config_files(bool userconfig)
> > return 0;
> > }
> >
> > +/* Smallest page size for migrated RAM. */
> > +uint64_t qemu_migration_page_size(void)
> > +{
> > + return TARGET_PAGE_SIZE;
> > +}
> > +
> > static inline bool is_zero_page(uint8_t *p)
> > {
> > return buffer_find_nonzero_offset(p, TARGET_PAGE_SIZE) ==
> > diff --git a/include/exec/memory.h b/include/exec/memory.h
> > index ebe0d24..6ffffa2 100644
> > --- a/include/exec/memory.h
> > +++ b/include/exec/memory.h
> > @@ -1055,6 +1055,7 @@ void *address_space_map(AddressSpace *as, hwaddr addr,
> > void address_space_unmap(AddressSpace *as, void *buffer, hwaddr len,
> > int is_write, hwaddr access_len);
> >
> > +extern uint64_t qemu_migration_page_size(void);
> >
> > #endif
>
> External linkage functions that are defined in "arch_init.c", and relate
> to migration -- for example, skipped_mig_bytes_transferred() -- are
> declared in "include/migration/migration.h". They seem to use
> TARGET_PAGE_SIZE too.
>
> What justifies declaring this new function in "include/exec/memory.h"?
>
> Thanks,
> Laszlo
Well, this one is a bit different as this just helps device
allocate the right amount of RAM.
It doesn't deal with migration directly.
--
MST
