[SPARC64]: Fill dead cycles on trap entry with real work.

As we save trap state onto the stack, the store buffer fills up
mid-way through and we stall for several cycles as the store buffer
trickles out to the L2 cache.  Meanwhile we can do some privileged
register reads and other calculations, essentially for free.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

arch/sparc64/kernel/etrap.S

@@ -98,37 +98,40 @@ etrap_save:	save	%g2, -STACK_BIAS, %sp
 		stxa	%g3, [%l4] ASI_DMMU
 		sethi	%hi(KERNBASE), %l4
 		flush	%l4
-		wr	%g0, ASI_AIUS, %asi
-2:		wrpr	%g0, 0x0, %tl
-		mov	%g4, %l4
+		mov	ASI_AIUS, %l7
+2:		mov	%g4, %l4
 		mov	%g5, %l5
-
-		mov	%g7, %l2
+		add	%g7, 4, %l2
 		wrpr	%g0, ETRAP_PSTATE1, %pstate
 		stx	%g1, [%sp + PTREGS_OFF + PT_V9_G1]
 		stx	%g2, [%sp + PTREGS_OFF + PT_V9_G2]
+		sllx	%l7, 24, %l7
 		stx	%g3, [%sp + PTREGS_OFF + PT_V9_G3]
+		rdpr	%cwp, %l0
 		stx	%g4, [%sp + PTREGS_OFF + PT_V9_G4]
 		stx	%g5, [%sp + PTREGS_OFF + PT_V9_G5]
 		stx	%g6, [%sp + PTREGS_OFF + PT_V9_G6]
-
 		stx	%g7, [%sp + PTREGS_OFF + PT_V9_G7]
+		or	%l7, %l0, %l7
+		sethi	%hi(TSTATE_RMO | TSTATE_PEF), %l0
+		or	%l7, %l0, %l7
+		wrpr	%l2, %tnpc
+		wrpr	%l7, (TSTATE_PRIV | TSTATE_IE), %tstate
 		stx	%i0, [%sp + PTREGS_OFF + PT_V9_I0]
 		stx	%i1, [%sp + PTREGS_OFF + PT_V9_I1]
 		stx	%i2, [%sp + PTREGS_OFF + PT_V9_I2]
 		stx	%i3, [%sp + PTREGS_OFF + PT_V9_I3]
 		stx	%i4, [%sp + PTREGS_OFF + PT_V9_I4]
 		stx	%i5, [%sp + PTREGS_OFF + PT_V9_I5]
-
 		stx	%i6, [%sp + PTREGS_OFF + PT_V9_I6]
-		stx	%i7, [%sp + PTREGS_OFF + PT_V9_I7]
-		wrpr	%g0, ETRAP_PSTATE2, %pstate
 		mov	%l6, %g6
+		stx	%i7, [%sp + PTREGS_OFF + PT_V9_I7]
 		LOAD_PER_CPU_BASE(%g5, %g6, %g4, %g3, %l1)
-		jmpl	%l2 + 0x4, %g0
-		 ldx	[%g6 + TI_TASK], %g4
+		ldx	[%g6 + TI_TASK], %g4
+		done
 
-3:		ldub	[%l6 + TI_FPDEPTH], %l5
+3:		mov	ASI_P, %l7
+		ldub	[%l6 + TI_FPDEPTH], %l5
 		add	%l6, TI_FPSAVED + 1, %l4
 		srl	%l5, 1, %l3
 		add	%l5, 2, %l5