diff --git a/arch/sparc64/kernel/head.S b/arch/sparc64/kernel/head.S
index 5c11529..7725952 100644
--- a/arch/sparc64/kernel/head.S
+++ b/arch/sparc64/kernel/head.S
@@ -653,33 +653,54 @@
 	 restore
 sparc64_boot_end:
 
-#include "ktlb.S"
-#include "tsb.S"
 #include "etrap.S"
 #include "rtrap.S"
 #include "winfixup.S"
 #include "entry.S"
 #include "sun4v_tlb_miss.S"
 #include "sun4v_ivec.S"
+#include "ktlb.S"
+#include "tsb.S"
 
 /*
  * The following skip makes sure the trap table in ttable.S is aligned
  * on a 32K boundary as required by the v9 specs for TBA register.
  *
  * We align to a 32K boundary, then we have the 32K kernel TSB,
- * then the 32K aligned trap table.
+ * the 64K kernel 4MB TSB, and then the 32K aligned trap table.
  */
 1:
 	.skip	0x4000 + _start - 1b
 
+! 0x0000000000408000
+
 	.globl	swapper_tsb
 swapper_tsb:
 	.skip	(32 * 1024)
 
-! 0x0000000000408000
+	.globl	swapper_4m_tsb
+swapper_4m_tsb:
+	.skip	(64 * 1024)
 
+! 0x0000000000420000
+
+	/* Some care needs to be exercised if you try to move the
+	 * location of the trap table relative to other things.  For
+	 * one thing there are br* instructions in some of the
+	 * trap table entires which branch back to code in ktlb.S
+	 * Those instructions can only handle a signed 16-bit
+	 * displacement.
+	 *
+	 * There is a binutils bug (bugzilla #4558) which causes
+	 * the relocation overflow checks for such instructions to
+	 * not be done correctly.  So bintuils will not notice the
+	 * error and will instead write junk into the relocation and
+	 * you'll have an unbootable kernel.
+	 */
 #include "ttable.S"
 
+! 0x0000000000428000
+
 #include "systbls.S"
 
 	.data
diff --git a/arch/sparc64/mm/init.c b/arch/sparc64/mm/init.c
index 087cbf0..97af4311 100644
--- a/arch/sparc64/mm/init.c
+++ b/arch/sparc64/mm/init.c
@@ -61,8 +61,11 @@
 unsigned long kpte_linear_bitmap[KPTE_BITMAP_BYTES / sizeof(unsigned long)];
 
 #ifndef CONFIG_DEBUG_PAGEALLOC
-/* A special kernel TSB for 4MB and 256MB linear mappings.  */
-struct tsb swapper_4m_tsb[KERNEL_TSB4M_NENTRIES];
+/* A special kernel TSB for 4MB and 256MB linear mappings.
+ * Space is allocated for this right after the trap table
+ * in arch/sparc64/kernel/head.S
+ */
+extern struct tsb swapper_4m_tsb[KERNEL_TSB4M_NENTRIES];
 #endif
 
 #define MAX_BANKS	32
diff --git a/include/asm-sparc64/tsb.h b/include/asm-sparc64/tsb.h
index ab55ffc..76e4299 100644
--- a/include/asm-sparc64/tsb.h
+++ b/include/asm-sparc64/tsb.h
@@ -271,7 +271,7 @@
 #define KERN_TSB4M_LOOKUP_TL1(TAG, REG1, REG2, REG3, REG4, OK_LABEL) \
 	sethi		%hi(swapper_4m_tsb), REG1; \
 	or		REG1, %lo(swapper_4m_tsb), REG1; \
-	and		TAG, (KERNEL_TSB_NENTRIES - 1), REG2; \
+	and		TAG, (KERNEL_TSB4M_NENTRIES - 1), REG2; \
 	sllx		REG2, 4, REG2; \
 	add		REG1, REG2, REG2; \
 	KTSB_LOAD_QUAD(REG2, REG3); \