[PATCH] m68k: broken constraints on mulu.l

Too permissive constraint on mulu.l - the first argument should not be
an a-register.  Fixed by replacing "g" with "dm"; with older gcc we got
lucky and it had never attempted mulu.l %a0, %d1:%d0.  These days it
does, with predictable objections from as(1).

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Cc: Roman Zippel <zippel@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

diff --git a/arch/m68k/math-emu/multi_arith.h b/arch/m68k/math-emu/multi_arith.h
index 02251e5afd891f83bcfc304ed8f34a975f2cb809..4ad0ca918e2e3e08426c5dd012adec1c29781b9e 100644 (file)
--- a/arch/m68k/math-emu/multi_arith.h
+++ b/arch/m68k/math-emu/multi_arith.h
@@ -366,7 +366,7 @@ static inline void fp_submant(struct fp_ext *dest, struct fp_ext *src1,
 
 #define fp_mul64(desth, destl, src1, src2) ({                          \
        asm ("mulu.l %2,%1:%0" : "=d" (destl), "=d" (desth)             \
-               : "g" (src1), "0" (src2));                              \
+               : "dm" (src1), "0" (src2));                             \
 })
 #define fp_div64(quot, rem, srch, srcl, div)                           \
        asm ("divu.l %2,%1:%0" : "=d" (quot), "=d" (rem)                \