lrdp-v2: Fixe vmemmap offset when kaslr is enabled for arm64

KASLR randomizes the address of the linear kernel mapping, which
causes the starting address of vmemmap to be similarly randomized.
Use the value of memstart_addr in calculations instead of phys_offset.

Change-Id: I905c0f50e011f85871ca2406b9a5d78afe4f0dc6

linux-ramdump-parser-v2/mm.py

@@ -152,22 +152,27 @@ def get_vmemmap(ramdump):
     spsize = ramdump.sizeof('struct page')
     vmemmap_size = bitops.align((1 << (va_bits - page_shift)) * spsize,
                                 pud_size)
-    pfn_offset = (ramdump.phys_offset >> page_shift)
-    offset = pfn_offset * spsize
+
+    memstart_addr = ramdump.read_s64('memstart_addr')
+    page_section_mask = ~((1 << 18) - 1)
+    memstart_offset = (memstart_addr >> page_shift) & page_section_mask
+    memstart_offset *= spsize
+
     if (ramdump.kernel_version < (3, 18, 31)):
-        vmalloc_end = ramdump.page_offset - pud_size - vmemmap_size
         # vmalloc_end = 0xFFFFFFBC00000000
+        vmemmap = ramdump.page_offset - pud_size - vmemmap_size
     elif (ramdump.kernel_version < (4, 9, 0)):
         # for version >= 3.18.31,
         # vmemmap is shifted to base addr (0x80000000) pfn.
-        vmalloc_end = ramdump.page_offset - pud_size - vmemmap_size - offset
+        vmemmap = (ramdump.page_offset - pud_size - vmemmap_size -
+                   memstart_offset)
     else:
         # for version >= 4.9.0,
+        # vmemmap_size = ( 1 << (39 - 12 - 1 + 6))
         struct_page_max_shift = 6
-        #vmemmap_size = ( 1 << (39 - 12 - 1 + 6))
         vmemmap_size = ( 1 << (va_bits - page_shift - 1 + struct_page_max_shift))
-        vmalloc_end = ramdump.page_offset - vmemmap_size - offset
-    return vmalloc_end
+        vmemmap = ramdump.page_offset - vmemmap_size - memstart_offset
+    return vmemmap
 
 
 def page_to_pfn_vmemmap(ramdump, page):