[SPARC64]: Simplify Spitfire D-cache page flush.

It tries to batch up the tag loads and comparisons, and
then the stores.  And this is just complicated instead
of efficient.

Also, make the symbol of the Cheetah version more grepable.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

arch/sparc64/mm/ultra.S

@@ -144,42 +144,29 @@ __flush_icache_page:	/* %o0 = phys_page */
 
 #define DTAG_MASK 0x3
 
+	/* This routine is Spitfire specific so the hardcoded
+	 * D-cache size and line-size are OK.
+	 */
 	.align		64
 	.globl		__flush_dcache_page
 __flush_dcache_page:	/* %o0=kaddr, %o1=flush_icache */
 	sethi		%uhi(PAGE_OFFSET), %g1
 	sllx		%g1, 32, %g1
-	sub		%o0, %g1, %o0
-	clr		%o4
-	srlx		%o0, 11, %o0
-	sethi		%hi(1 << 14), %o2
-1:	ldxa		[%o4] ASI_DCACHE_TAG, %o3	! LSU	Group
-	add		%o4, (1 << 5), %o4		! IEU0
-	ldxa		[%o4] ASI_DCACHE_TAG, %g1	! LSU	Group
-	add		%o4, (1 << 5), %o4		! IEU0
-	ldxa		[%o4] ASI_DCACHE_TAG, %g2	! LSU	Group	o3 available
-	add		%o4, (1 << 5), %o4		! IEU0
-	andn		%o3, DTAG_MASK, %o3		! IEU1
-	ldxa		[%o4] ASI_DCACHE_TAG, %g3	! LSU	Group
-	add		%o4, (1 << 5), %o4		! IEU0
-	andn		%g1, DTAG_MASK, %g1		! IEU1
-	cmp		%o0, %o3			! IEU1	Group
-	be,a,pn		%xcc, dflush1			! CTI
-	 sub		%o4, (4 << 5), %o4		! IEU0	(Group)
-	cmp		%o0, %g1			! IEU1	Group
-	andn		%g2, DTAG_MASK, %g2		! IEU0
-	be,a,pn		%xcc, dflush2			! CTI
-	 sub		%o4, (3 << 5), %o4		! IEU0	(Group)
-	cmp		%o0, %g2			! IEU1	Group
-	andn		%g3, DTAG_MASK, %g3		! IEU0
-	be,a,pn		%xcc, dflush3			! CTI
-	 sub		%o4, (2 << 5), %o4		! IEU0	(Group)
-	cmp		%o0, %g3			! IEU1	Group
-	be,a,pn		%xcc, dflush4			! CTI
-	 sub		%o4, (1 << 5), %o4		! IEU0
-2:	cmp		%o4, %o2			! IEU1	Group
-	bne,pt		%xcc, 1b			! CTI
-	 nop						! IEU0
+	sub		%o0, %g1, %o0			! physical address
+	srlx		%o0, 11, %o0			! make D-cache TAG
+	sethi		%hi(1 << 14), %o2		! D-cache size
+	sub		%o2, (1 << 5), %o2		! D-cache line size
+1:	ldxa		[%o2] ASI_DCACHE_TAG, %o3	! load D-cache TAG
+	andcc		%o3, DTAG_MASK, %g0		! Valid?
+	be,pn		%xcc, 2f			! Nope, branch
+	 andn		%o3, DTAG_MASK, %o3		! Clear valid bits
+	cmp		%o3, %o0			! TAG match?
+	bne,pt		%xcc, 2f			! Nope, branch
+	 nop
+	stxa		%g0, [%o2] ASI_DCACHE_TAG	! Invalidate TAG
+	membar		#Sync
+2:	brnz,pt		%o2, 1b
+	 sub		%o2, (1 << 5), %o2		! D-cache line size
 
 	/* The I-cache does not snoop local stores so we
 	 * better flush that too when necessary.
@@ -189,20 +176,10 @@ __flush_dcache_page:	/* %o0=kaddr, %o1=flush_icache */
 	retl
 	 nop
 
-dflush1:stxa		%g0, [%o4] ASI_DCACHE_TAG
-	add		%o4, (1 << 5), %o4
-dflush2:stxa		%g0, [%o4] ASI_DCACHE_TAG
-	add		%o4, (1 << 5), %o4
-dflush3:stxa		%g0, [%o4] ASI_DCACHE_TAG
-	add		%o4, (1 << 5), %o4
-dflush4:stxa		%g0, [%o4] ASI_DCACHE_TAG
-	add		%o4, (1 << 5), %o4
-	membar		#Sync
-	ba,pt		%xcc, 2b
-	 nop
 #endif /* DCACHE_ALIASING_POSSIBLE */
 
-	.previous .text
+	.previous
+
 	.align		32
 __prefill_dtlb:
 	rdpr		%pstate, %g7
@@ -283,7 +260,7 @@ __cheetah_flush_tlb_pending:	/* 26 insns */
 	 wrpr		%g7, 0x0, %pstate
 
 #ifdef DCACHE_ALIASING_POSSIBLE
-flush_dcpage_cheetah: /* 11 insns */
+__cheetah_flush_dcache_page: /* 11 insns */
 	sethi		%uhi(PAGE_OFFSET), %g1
 	sllx		%g1, 32, %g1
 	sub		%o0, %g1, %o0
@@ -329,8 +306,8 @@ cheetah_patch_cachetlbops:
 #ifdef DCACHE_ALIASING_POSSIBLE
 	sethi		%hi(__flush_dcache_page), %o0
 	or		%o0, %lo(__flush_dcache_page), %o0
-	sethi		%hi(flush_dcpage_cheetah), %o1
-	or		%o1, %lo(flush_dcpage_cheetah), %o1
+	sethi		%hi(__cheetah_flush_dcache_page), %o1
+	or		%o1, %lo(__cheetah_flush_dcache_page), %o1
 	call		cheetah_patch_one
 	 mov		11, %o2
 #endif /* DCACHE_ALIASING_POSSIBLE */