|
|
@@ -351,7 +351,25 @@ void radix__early_init_mmu_secondary(void)
|
|
|
void radix__setup_initial_memory_limit(phys_addr_t first_memblock_base,
|
|
|
phys_addr_t first_memblock_size)
|
|
|
{
|
|
|
- /* Finally limit subsequent allocations */
|
|
|
+ /* We don't currently support the first MEMBLOCK not mapping 0
|
|
|
+ * physical on those processors
|
|
|
+ */
|
|
|
+ BUG_ON(first_memblock_base != 0);
|
|
|
+ /*
|
|
|
+ * We limit the allocation that depend on ppc64_rma_size
|
|
|
+ * to first_memblock_size. We also clamp it to 1GB to
|
|
|
+ * avoid some funky things such as RTAS bugs.
|
|
|
+ *
|
|
|
+ * On radix config we really don't have a limitation
|
|
|
+ * on real mode access. But keeping it as above works
|
|
|
+ * well enough.
|
|
|
+ */
|
|
|
+ ppc64_rma_size = min_t(u64, first_memblock_size, 0x40000000);
|
|
|
+ /*
|
|
|
+ * Finally limit subsequent allocations. We really don't want
|
|
|
+ * to limit the memblock allocations to rma_size. FIXME!! should
|
|
|
+ * we even limit at all ?
|
|
|
+ */
|
|
|
memblock_set_current_limit(first_memblock_base + first_memblock_size);
|
|
|
}
|
|
|
|