[SPARC64]: Fix userland FPU state corruption.

We need to use stricter memory barriers around the block
load and store instructions we use to save and restore the
FPU register file.

Signed-off-by: David S. Miller <davem@davemloft.net>

2 files changed, 25 insertions, 21 deletions

diff --git a/arch/sparc64/kernel/entry.S b/arch/sparc64/kernel/entry.S
index f685035dbdb8..11a848402fb1 100644
--- a/arch/sparc64/kernel/entry.S
+++ b/arch/sparc64/kernel/entry.S
@@ -33,7 +33,7 @@
 	/* This is trivial with the new code... */
 	.globl		do_fpdis
 do_fpdis:
-	sethi		%hi(TSTATE_PEF), %g4					! IEU0
+	sethi		%hi(TSTATE_PEF), %g4
 	rdpr		%tstate, %g5
 	andcc		%g5, %g4, %g0
 	be,pt		%xcc, 1f
@@ -50,18 +50,18 @@ do_fpdis:
 	add		%g0, %g0, %g0
 	ba,a,pt		%xcc, rtrap_clr_l6
 
-1:	ldub		[%g6 + TI_FPSAVED], %g5					! Load	Group
-	wr		%g0, FPRS_FEF, %fprs					! LSU	Group+4bubbles
-	andcc		%g5, FPRS_FEF, %g0					! IEU1	Group
-	be,a,pt		%icc, 1f						! CTI
-	 clr		%g7							! IEU0
-	ldx		[%g6 + TI_GSR], %g7					! Load	Group
-1:	andcc		%g5, FPRS_DL, %g0					! IEU1
-	bne,pn		%icc, 2f						! CTI
-	 fzero		%f0							! FPA
-	andcc		%g5, FPRS_DU, %g0					! IEU1  Group
-	bne,pn		%icc, 1f						! CTI
-	 fzero		%f2							! FPA
+1:	ldub		[%g6 + TI_FPSAVED], %g5
+	wr		%g0, FPRS_FEF, %fprs
+	andcc		%g5, FPRS_FEF, %g0
+	be,a,pt		%icc, 1f
+	 clr		%g7
+	ldx		[%g6 + TI_GSR], %g7
+1:	andcc		%g5, FPRS_DL, %g0
+	bne,pn		%icc, 2f
+	 fzero		%f0
+	andcc		%g5, FPRS_DU, %g0
+	bne,pn		%icc, 1f
+	 fzero		%f2
 	faddd		%f0, %f2, %f4
 	fmuld		%f0, %f2, %f6
 	faddd		%f0, %f2, %f8
@@ -104,8 +104,10 @@ do_fpdis:
 	add		%g6, TI_FPREGS + 0xc0, %g2
 	faddd		%f0, %f2, %f8
 	fmuld		%f0, %f2, %f10
-	ldda		[%g1] ASI_BLK_S, %f32	! grrr, where is ASI_BLK_NUCLEUS 8-(
+	membar		#Sync
+	ldda		[%g1] ASI_BLK_S, %f32
 	ldda		[%g2] ASI_BLK_S, %f48
+	membar		#Sync
 	faddd		%f0, %f2, %f12
 	fmuld		%f0, %f2, %f14
 	faddd		%f0, %f2, %f16
@@ -116,7 +118,6 @@ do_fpdis:
 	fmuld		%f0, %f2, %f26
 	faddd		%f0, %f2, %f28
 	fmuld		%f0, %f2, %f30
-	membar		#Sync
 	b,pt		%xcc, fpdis_exit
 	 nop
 2:	andcc		%g5, FPRS_DU, %g0
@@ -133,8 +134,10 @@ do_fpdis:
 	add		%g6, TI_FPREGS + 0x40, %g2
 	faddd		%f32, %f34, %f36
 	fmuld		%f32, %f34, %f38
-	ldda		[%g1] ASI_BLK_S, %f0	! grrr, where is ASI_BLK_NUCLEUS 8-(
+	membar		#Sync
+	ldda		[%g1] ASI_BLK_S, %f0
 	ldda		[%g2] ASI_BLK_S, %f16
+	membar		#Sync
 	faddd		%f32, %f34, %f40
 	fmuld		%f32, %f34, %f42
 	faddd		%f32, %f34, %f44
@@ -147,7 +150,6 @@ do_fpdis:
 	fmuld		%f32, %f34, %f58
 	faddd		%f32, %f34, %f60
 	fmuld		%f32, %f34, %f62
-	membar		#Sync
 	ba,pt		%xcc, fpdis_exit
 	 nop
 3:	mov		SECONDARY_CONTEXT, %g3
@@ -158,7 +160,8 @@ do_fpdis:
 	stxa		%g2, [%g3] ASI_DMMU
 	membar		#Sync
 	mov		0x40, %g2
-	ldda		[%g1] ASI_BLK_S, %f0		! grrr, where is ASI_BLK_NUCLEUS 8-(
+	membar		#Sync
+	ldda		[%g1] ASI_BLK_S, %f0
 	ldda		[%g1 + %g2] ASI_BLK_S, %f16
 	add		%g1, 0x80, %g1
 	ldda		[%g1] ASI_BLK_S, %f32
diff --git a/arch/sparc64/kernel/rtrap.S b/arch/sparc64/kernel/rtrap.S
index ecfb42a69a44..090dcca00d2a 100644
--- a/arch/sparc64/kernel/rtrap.S
+++ b/arch/sparc64/kernel/rtrap.S
@@ -312,32 +312,33 @@ kern_fpucheck:	ldub			[%g6 + TI_FPDEPTH], %l5
 		wr			%g1, FPRS_FEF, %fprs
 		ldx			[%o1 + %o5], %g1
 		add			%g6, TI_XFSR, %o1
-		membar			#StoreLoad | #LoadLoad
 		sll			%o0, 8, %o2
 		add			%g6, TI_FPREGS, %o3
 		brz,pn			%l6, 1f
 		 add			%g6, TI_FPREGS+0x40, %o4
 
+		membar			#Sync
 		ldda			[%o3 + %o2] ASI_BLK_P, %f0
 		ldda			[%o4 + %o2] ASI_BLK_P, %f16
+		membar			#Sync
 1:		andcc			%l2, FPRS_DU, %g0
 		be,pn			%icc, 1f
 		 wr			%g1, 0, %gsr
 		add			%o2, 0x80, %o2
+		membar			#Sync
 		ldda			[%o3 + %o2] ASI_BLK_P, %f32
 		ldda			[%o4 + %o2] ASI_BLK_P, %f48
-
 1:		membar			#Sync
 		ldx			[%o1 + %o5], %fsr
 2:		stb			%l5, [%g6 + TI_FPDEPTH]
 		ba,pt			%xcc, rt_continue
 		 nop
 5:		wr			%g0, FPRS_FEF, %fprs
-		membar			#StoreLoad | #LoadLoad
 		sll			%o0, 8, %o2
 
 		add			%g6, TI_FPREGS+0x80, %o3
 		add			%g6, TI_FPREGS+0xc0, %o4
+		membar			#Sync
 		ldda			[%o3 + %o2] ASI_BLK_P, %f32
 		ldda			[%o4 + %o2] ASI_BLK_P, %f48
 		membar			#Sync