[SPARC64]: Do not assume sun4v chips have load-twin/store-init support.

Check the cpu type in the OBP device tree before committing to
using the optimized Niagara memcpy and memset implementation.

If we don't recognize the cpu type, use a completely generic
version.

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

1 files changed, 160 insertions, 0 deletions

diff --git a/arch/sparc64/lib/GENbzero.S b/arch/sparc64/lib/GENbzero.S
new file mode 100644
index 00000000000..f9c71d64eba
--- /dev/null
+++ b/arch/sparc64/lib/GENbzero.S
@@ -0,0 +1,160 @@
+/* GENbzero.S: Generic sparc64 memset/clear_user.
+ *
+ * Copyright (C) 2007 David S. Miller (davem@davemloft.net)
+ */
+#include <asm/asi.h>
+
+#define EX_ST(x,y)		\
+98:	x,y;			\
+	.section .fixup;	\
+	.align 4;		\
+99:	retl;			\
+	 mov	%o1, %o0;	\
+	.section __ex_table;	\
+	.align 4;		\
+	.word 98b, 99b;		\
+	.text;			\
+	.align 4;
+
+	.align	32
+	.text
+
+	.globl		GENmemset
+	.type		GENmemset, #function
+GENmemset:		/* %o0=buf, %o1=pat, %o2=len */
+	and		%o1, 0xff, %o3
+	mov		%o2, %o1
+	sllx		%o3, 8, %g1
+	or		%g1, %o3, %o2
+	sllx		%o2, 16, %g1
+	or		%g1, %o2, %o2
+	sllx		%o2, 32, %g1
+	ba,pt		%xcc, 1f
+	 or		%g1, %o2, %o2
+
+	.globl		GENbzero
+	.type		GENbzero, #function
+GENbzero:
+	clr		%o2
+1:	brz,pn		%o1, GENbzero_return
+	 mov		%o0, %o3
+
+	/* %o5: saved %asi, restored at GENbzero_done
+	 * %o4:	store %asi to use
+	 */
+	rd		%asi, %o5
+	mov		ASI_P, %o4
+	wr		%o4, 0x0, %asi
+
+GENbzero_from_clear_user:
+	cmp		%o1, 15
+	bl,pn		%icc, GENbzero_tiny
+	 andcc		%o0, 0x7, %g1
+	be,pt		%xcc, 2f
+	 mov		8, %g2
+	sub		%g2, %g1, %g1
+	sub		%o1, %g1, %o1
+1:	EX_ST(stba %o2, [%o0 + 0x00] %asi)
+	subcc		%g1, 1, %g1
+	bne,pt		%xcc, 1b
+	 add		%o0, 1, %o0
+2:	cmp		%o1, 128
+	bl,pn		%icc, GENbzero_medium
+	 andcc		%o0, (64 - 1), %g1
+	be,pt		%xcc, GENbzero_pre_loop
+	 mov		64, %g2
+	sub		%g2, %g1, %g1
+	sub		%o1, %g1, %o1
+1:	EX_ST(stxa %o2, [%o0 + 0x00] %asi)
+	subcc		%g1, 8, %g1
+	bne,pt		%xcc, 1b
+	 add		%o0, 8, %o0
+
+GENbzero_pre_loop:
+	andn		%o1, (64 - 1), %g1
+	sub		%o1, %g1, %o1
+GENbzero_loop:
+	EX_ST(stxa %o2, [%o0 + 0x00] %asi)
+	EX_ST(stxa %o2, [%o0 + 0x08] %asi)
+	EX_ST(stxa %o2, [%o0 + 0x10] %asi)
+	EX_ST(stxa %o2, [%o0 + 0x18] %asi)
+	EX_ST(stxa %o2, [%o0 + 0x20] %asi)
+	EX_ST(stxa %o2, [%o0 + 0x28] %asi)
+	EX_ST(stxa %o2, [%o0 + 0x30] %asi)
+	EX_ST(stxa %o2, [%o0 + 0x38] %asi)
+	subcc		%g1, 64, %g1
+	bne,pt		%xcc, GENbzero_loop
+	 add		%o0, 64, %o0
+
+	membar		#Sync
+	wr		%o4, 0x0, %asi
+	brz,pn		%o1, GENbzero_done
+GENbzero_medium:
+	 andncc		%o1, 0x7, %g1
+	be,pn		%xcc, 2f
+	 sub		%o1, %g1, %o1
+1:	EX_ST(stxa %o2, [%o0 + 0x00] %asi)
+	subcc		%g1, 8, %g1
+	bne,pt		%xcc, 1b
+	 add		%o0, 8, %o0
+2:	brz,pt		%o1, GENbzero_done
+	 nop
+
+GENbzero_tiny:
+1:	EX_ST(stba %o2, [%o0 + 0x00] %asi)
+	subcc		%o1, 1, %o1
+	bne,pt		%icc, 1b
+	 add		%o0, 1, %o0
+
+	/* fallthrough */
+
+GENbzero_done:
+	wr		%o5, 0x0, %asi
+
+GENbzero_return:
+	retl
+	 mov		%o3, %o0
+	.size		GENbzero, .-GENbzero
+	.size		GENmemset, .-GENmemset
+
+	.globl		GENclear_user
+	.type		GENclear_user, #function
+GENclear_user:		/* %o0=buf, %o1=len */
+	rd		%asi, %o5
+	brz,pn		%o1, GENbzero_done
+	 clr		%o3
+	cmp		%o5, ASI_AIUS
+	bne,pn		%icc, GENbzero
+	 clr		%o2
+	ba,pt		%xcc, GENbzero_from_clear_user
+	 mov		ASI_AIUS, %o4
+	.size		GENclear_user, .-GENclear_user
+
+#define BRANCH_ALWAYS	0x10680000
+#define NOP		0x01000000
+#define GEN_DO_PATCH(OLD, NEW)	\
+	sethi	%hi(NEW), %g1; \
+	or	%g1, %lo(NEW), %g1; \
+	sethi	%hi(OLD), %g2; \
+	or	%g2, %lo(OLD), %g2; \
+	sub	%g1, %g2, %g1; \
+	sethi	%hi(BRANCH_ALWAYS), %g3; \
+	sll	%g1, 11, %g1; \
+	srl	%g1, 11 + 2, %g1; \
+	or	%g3, %lo(BRANCH_ALWAYS), %g3; \
+	or	%g3, %g1, %g3; \
+	stw	%g3, [%g2]; \
+	sethi	%hi(NOP), %g3; \
+	or	%g3, %lo(NOP), %g3; \
+	stw	%g3, [%g2 + 0x4]; \
+	flush	%g2;
+
+	.globl	generic_patch_bzero
+	.type	generic_patch_bzero,#function
+generic_patch_bzero:
+	GEN_DO_PATCH(memset, GENmemset)
+	GEN_DO_PATCH(__bzero, GENbzero)
+	GEN_DO_PATCH(__clear_user, GENclear_user)
+	retl
+	 nop
+	.size	generic_patch_bzero,.-generic_patch_bzero