x86-64, NUMA: Remove custom phys_to_nid() implementation

phys_to_nid() maps physical address to NUMA node id.  This is
implemented by building perfect hash in compute_hash_shift() during
initialization.

However, with SPARSE memory model, the nid is encoded in page flags.
The perfect hash implementation was for DISCONTIG memory model which
got removed years ago by b263295dbf (x86: 64-bit, make sparsemem
vmemmap the only memory model).

So, the perfect hash ends up being used only during initialization
when the core SPARSE code already provides perfectly acceptable
generic early_pfn_to_nid() implementation.

Drop phys_to_nid() and use the generic ealry_pfn_to_nid() instead.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Christoph Lameter <cl@linux-foundation.org>
Acked-by: Yinghai Lu <yinghai@kernel.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Thomas Gleixner <tglx@linutronix.de>

1 files changed, 0 insertions, 4 deletions

diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index cc6c53a95bfd..b034814bf975 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -1703,10 +1703,6 @@ config ARCH_ENABLE_MEMORY_HOTREMOVE
         def_bool y
         depends on MEMORY_HOTPLUG
 
-config HAVE_ARCH_EARLY_PFN_TO_NID
-        def_bool X86_64
-        depends on NUMA
-
 config USE_PERCPU_NUMA_NODE_ID
         def_bool y
         depends on NUMA