sparc64: Fix return value of Niagara-2 memcpy.

It gets clobbered by the kernel's VISEntryHalf, so we have to save it
in a different register than the set clobbered by that macro.

The instance in glibc is OK and doesn't have this problem.

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

diff --git a/arch/sparc/lib/NG2memcpy.S b/arch/sparc/lib/NG2memcpy.S
index 03eadf6..2c20ad6 100644 (file)
--- a/arch/sparc/lib/NG2memcpy.S
+++ b/arch/sparc/lib/NG2memcpy.S
@@ -14,7 +14,7 @@
 #define FPRS_FEF  0x04
 #ifdef MEMCPY_DEBUG
 #define VISEntryHalf rd %fprs, %o5; wr %g0, FPRS_FEF, %fprs; \
-                    clr %g1; clr %g2; clr %g3; subcc %g0, %g0, %g0;
+                    clr %g1; clr %g2; clr %g3; clr %g5; subcc %g0, %g0, %g0;
 #define VISExitHalf and %o5, FPRS_FEF, %o5; wr %o5, 0x0, %fprs
 #else
 #define VISEntryHalf rd %fprs, %o5; wr %g0, FPRS_FEF, %fprs
@@ -182,13 +182,13 @@ FUNC_NAME:        /* %o0=dst, %o1=src, %o2=len */
        cmp             %g2, 0
        tne             %xcc, 5
        PREAMBLE
-       mov             %o0, GLOBAL_SPARE
+       mov             %o0, %o3
        cmp             %o2, 0
        be,pn           %XCC, 85f
-        or             %o0, %o1, %o3
+        or             %o0, %o1, GLOBAL_SPARE
        cmp             %o2, 16
        blu,a,pn        %XCC, 80f
-        or             %o3, %o2, %o3
+        or             GLOBAL_SPARE, %o2, GLOBAL_SPARE
 
        /* 2 blocks (128 bytes) is the minimum we can do the block
         * copy with.  We need to ensure that we'll iterate at least
@@ -202,7 +202,7 @@ FUNC_NAME:  /* %o0=dst, %o1=src, %o2=len */
         */
        cmp             %o2, (4 * 64)
        blu,pt          %XCC, 75f
-        andcc          %o3, 0x7, %g0
+        andcc          GLOBAL_SPARE, 0x7, %g0
 
        /* %o0: dst
         * %o1: src
@@ -404,13 +404,13 @@ FUNC_NAME:        /* %o0=dst, %o1=src, %o2=len */
         * over. If anything is left, we copy it one byte at a time.
         */
        brz,pt          %o2, 85f
-        sub            %o0, %o1, %o3
+        sub            %o0, %o1, GLOBAL_SPARE
        ba,a,pt         %XCC, 90f
 
        .align          64
 75: /* 16 < len <= 64 */
        bne,pn          %XCC, 75f
-        sub            %o0, %o1, %o3
+        sub            %o0, %o1, GLOBAL_SPARE
 
 72:
        andn            %o2, 0xf, %o4
@@ -420,9 +420,9 @@ FUNC_NAME:  /* %o0=dst, %o1=src, %o2=len */
        add             %o1, 0x08, %o1
        EX_LD(LOAD(ldx, %o1, %g1))
        sub             %o1, 0x08, %o1
-       EX_ST(STORE(stx, %o5, %o1 + %o3))
+       EX_ST(STORE(stx, %o5, %o1 + GLOBAL_SPARE))
        add             %o1, 0x8, %o1
-       EX_ST(STORE(stx, %g1, %o1 + %o3))
+       EX_ST(STORE(stx, %g1, %o1 + GLOBAL_SPARE))
        bgu,pt          %XCC, 1b
         add            %o1, 0x8, %o1
 73:    andcc           %o2, 0x8, %g0
@@ -430,14 +430,14 @@ FUNC_NAME:        /* %o0=dst, %o1=src, %o2=len */
         nop
        sub             %o2, 0x8, %o2
        EX_LD(LOAD(ldx, %o1, %o5))
-       EX_ST(STORE(stx, %o5, %o1 + %o3))
+       EX_ST(STORE(stx, %o5, %o1 + GLOBAL_SPARE))
        add             %o1, 0x8, %o1
 1:     andcc           %o2, 0x4, %g0
        be,pt           %XCC, 1f
         nop
        sub             %o2, 0x4, %o2
        EX_LD(LOAD(lduw, %o1, %o5))
-       EX_ST(STORE(stw, %o5, %o1 + %o3))
+       EX_ST(STORE(stw, %o5, %o1 + GLOBAL_SPARE))
        add             %o1, 0x4, %o1
 1:     cmp             %o2, 0
        be,pt           %XCC, 85f
@@ -454,11 +454,11 @@ FUNC_NAME:        /* %o0=dst, %o1=src, %o2=len */
 
 1:     subcc           %g1, 1, %g1
        EX_LD(LOAD(ldub, %o1, %o5))
-       EX_ST(STORE(stb, %o5, %o1 + %o3))
+       EX_ST(STORE(stb, %o5, %o1 + GLOBAL_SPARE))
        bgu,pt          %icc, 1b
         add            %o1, 1, %o1
 
-2:     add             %o1, %o3, %o0
+2:     add             %o1, GLOBAL_SPARE, %o0
        andcc           %o1, 0x7, %g1
        bne,pt          %icc, 8f
         sll            %g1, 3, %g1
@@ -468,16 +468,16 @@ FUNC_NAME:        /* %o0=dst, %o1=src, %o2=len */
         nop
        ba,a,pt         %xcc, 73b
 
-8:     mov             64, %o3
+8:     mov             64, GLOBAL_SPARE
        andn            %o1, 0x7, %o1
        EX_LD(LOAD(ldx, %o1, %g2))
-       sub             %o3, %g1, %o3
+       sub             GLOBAL_SPARE, %g1, GLOBAL_SPARE
        andn            %o2, 0x7, %o4
        sllx            %g2, %g1, %g2
 1:     add             %o1, 0x8, %o1
        EX_LD(LOAD(ldx, %o1, %g3))
        subcc           %o4, 0x8, %o4
-       srlx            %g3, %o3, %o5
+       srlx            %g3, GLOBAL_SPARE, %o5
        or              %o5, %g2, %o5
        EX_ST(STORE(stx, %o5, %o0))
        add             %o0, 0x8, %o0
@@ -489,32 +489,32 @@ FUNC_NAME:        /* %o0=dst, %o1=src, %o2=len */
        be,pn           %icc, 85f
         add            %o1, %g1, %o1
        ba,pt           %xcc, 90f
-        sub            %o0, %o1, %o3
+        sub            %o0, %o1, GLOBAL_SPARE
 
        .align          64
 80: /* 0 < len <= 16 */
-       andcc           %o3, 0x3, %g0
+       andcc           GLOBAL_SPARE, 0x3, %g0
        bne,pn          %XCC, 90f
-        sub            %o0, %o1, %o3
+        sub            %o0, %o1, GLOBAL_SPARE
 
 1:
        subcc           %o2, 4, %o2
        EX_LD(LOAD(lduw, %o1, %g1))
-       EX_ST(STORE(stw, %g1, %o1 + %o3))
+       EX_ST(STORE(stw, %g1, %o1 + GLOBAL_SPARE))
        bgu,pt          %XCC, 1b
         add            %o1, 4, %o1
 
 85:    retl
-        mov            EX_RETVAL(GLOBAL_SPARE), %o0
+        mov            EX_RETVAL(%o3), %o0
 
        .align          32
 90:
        subcc           %o2, 1, %o2
        EX_LD(LOAD(ldub, %o1, %g1))
-       EX_ST(STORE(stb, %g1, %o1 + %o3))
+       EX_ST(STORE(stb, %g1, %o1 + GLOBAL_SPARE))
        bgu,pt          %XCC, 90b
         add            %o1, 1, %o1
        retl
-        mov            EX_RETVAL(GLOBAL_SPARE), %o0
+        mov            EX_RETVAL(%o3), %o0
 
        .size           FUNC_NAME, .-FUNC_NAME