svsm_paging: Ensure page alignment during invalidation

[peterfang]

Boot memory regions aren't always page-aligned. For example, kernel_fs_end cannot always be page-aligned because the size of the FS area has to match that of its original image exactly. Make sure the memory region always has a page-aligned size in
invalidate_boot_memory_region(). The starting address is always checked during page invalidation so just assume the caller will do the right thing.

[joergroedel]

I see why this change is needed. Question(s): Can it cause problems elsewhere when we only enforce the alignment at this one call-site? Or would it be better to align the addresses in the page_state_change() platform call-backs?

[peterfang]

I see why this change is needed. Question(s): Can it cause problems elsewhere when we only enforce the alignment at this one call-site? Or would it be better to align the addresses in the page_state_change() platform call-backs?

Hi Joerg,

I checked all of the call sites for validate_physical_page_range() and validate_virtual_page_range() and this function is the only place right now where memory regions could be unaligned. The only other case that could cause this is if someone overrode the SVSM kernel region in QEMU to some odd, unaligned location in non-IGVM mode, but that doesn't sound like a normal case to me. page_state_change() is called after validate_physical_page_range() so for an unaligned memory region, the panic would be hit first.

[joergroedel]

Thanks, looks good the me now.

[joergroedel]

Will merge this after #577 and #578