On 02/08/2024 14:33, Julien Grall wrote:
Hi,
Hi Julien,
On 02/08/2024 13:14, Ayan Kumar Halder wrote:
Moved init_domheap_mappings(), map_domain_page_global(),
unmap_domain_page_global(), map_domain_page(), unmap_domain_page(),
domain_page_map_to_mfn() to MMU specific folder.
On the top level, we have added stubs which will invoke the
MMU specific equivalent of these functions.
Signed-off-by: Ayan Kumar Halder <[email protected]>
---
xen/arch/arm/domain_page.c | 161 +----------------------
Looking at what's left in domain_page.c, we seem to have just a series
of stubs:
bool init_domheap_mappings(unsigned int cpu)
{
return init_domheap_mappings_mm(cpu);
}
void *map_domain_page_global(mfn_t mfn)
{
return map_domain_page_global_mm(mfn);
}
void unmap_domain_page_global(const void *ptr)
{
return unmap_domain_page_global_mm(ptr);
}
/* Map a page of domheap memory */
void *map_domain_page(mfn_t mfn)
{
return map_domain_page_mm(mfn);
}
/* Release a mapping taken with map_domain_page() */
void unmap_domain_page(const void *ptr)
{
return unmap_domain_page_mm(ptr);
}
mfn_t domain_page_map_to_mfn(const void *ptr)
{
return domain_page_map_to_mfn_mm(ptr);
}
The indirection seems unnecessary to me. What about renaming
arch/arm/domain_page.c to arch/arm/mmu/domain_page.c? Then, for the
MMU you can implement unmap_domain_page() & co rather than suffixed
_mm ones.
Makes sense. I will do this way.
- Ayan
Cheers,
