improve i386 inline syscall asm on non-broken compilers

we have to avoid using ebx unconditionally in asm constraints for
i386, because gcc 3 and 4 and possibly other simplistic compilers
(pcc?) implement PIC via making ebx a fixed-use register, and disallow
its use for anything else. rather than hard-coding knowledge of which
compilers work (at least gcc 5+ and clang), perform a configure test;
this should give us the good codegen on any new compilers we don't yet
know about.

swapping ebx and edx is kept for 1- and 2-arg syscalls because it
avoids having any spills/stack-frame at all in small functions. for
6-arg, if ebx is directly usable, the complex shuffling introduced in
commit c8798ef974d21c338a7d8d874a402978ffc6168e can be avoided, and
ebp can be loaded the same way ebx is in 5-arg syscalls for compilers
that don't support direct use of ebx.

arch/i386/syscall_arch.h

@@ -36,30 +36,49 @@ static inline long __syscall2(long n, long a1, long a2)
 static inline long __syscall3(long n, long a1, long a2, long a3)
 {
 	unsigned long __ret;
+#if !defined(__PIC__) || !defined(BROKEN_EBX_ASM)
+	__asm__ __volatile__ (SYSCALL_INSNS : "=a"(__ret) : "a"(n), "b"(a1), "c"(a2), "d"(a3) : "memory");
+#else
 	__asm__ __volatile__ (SYSCALL_INSNS_34 : "=a"(__ret) : "a"(n), "D"(a1), "c"(a2), "d"(a3) : "memory");
+#endif
 	return __ret;
 }
 
 static inline long __syscall4(long n, long a1, long a2, long a3, long a4)
 {
 	unsigned long __ret;
+#if !defined(__PIC__) || !defined(BROKEN_EBX_ASM)
+	__asm__ __volatile__ (SYSCALL_INSNS : "=a"(__ret) : "a"(n), "b"(a1), "c"(a2), "d"(a3), "S"(a4) : "memory");
+#else
 	__asm__ __volatile__ (SYSCALL_INSNS_34 : "=a"(__ret) : "a"(n), "D"(a1), "c"(a2), "d"(a3), "S"(a4) : "memory");
+#endif
 	return __ret;
 }
 
 static inline long __syscall5(long n, long a1, long a2, long a3, long a4, long a5)
 {
 	unsigned long __ret;
+#if !defined(__PIC__) || !defined(BROKEN_EBX_ASM)
+	__asm__ __volatile__ (SYSCALL_INSNS
+		: "=a"(__ret) : "a"(n), "b"(a1), "c"(a2), "d"(a3), "S"(a4), "D"(a5) : "memory");
+#else
 	__asm__ __volatile__ ("pushl %2 ; push %%ebx ; mov 4(%%esp),%%ebx ; " SYSCALL_INSNS " ; pop %%ebx ; add $4,%%esp"
 		: "=a"(__ret) : "a"(n), "g"(a1), "c"(a2), "d"(a3), "S"(a4), "D"(a5) : "memory");
+#endif
 	return __ret;
 }
 
 static inline long __syscall6(long n, long a1, long a2, long a3, long a4, long a5, long a6)
 {
-	unsigned long __ret, a1a6[2] = { a1, a6 };
+	unsigned long __ret;
+#if !defined(__PIC__) || !defined(BROKEN_EBX_ASM)
+	__asm__ __volatile__ ("pushl %7 ; push %%ebp ; mov 4(%%esp),%%ebp ; " SYSCALL_INSNS " ; pop %%ebp ; add $4,%%esp"
+		: "=a"(__ret) : "a"(n), "b"(a1), "c"(a2), "d"(a3), "S"(a4), "D"(a5), "g"(a6) : "memory");
+#else
+	unsigned long a1a6[2] = { a1, a6 };
 	__asm__ __volatile__ ("pushl %1 ; push %%ebx ; push %%ebp ; mov 8(%%esp),%%ebx ; mov 4(%%ebx),%%ebp ; mov (%%ebx),%%ebx ; " SYSCALL_INSNS " ; pop %%ebp ; pop %%ebx ; add $4,%%esp"
 		: "=a"(__ret) : "g"(&a1a6), "a"(n), "c"(a2), "d"(a3), "S"(a4), "D"(a5) : "memory");
+#endif
 	return __ret;
 }
 

configure

@@ -573,6 +573,20 @@ printf "using compiler runtime libraries: %s\n" "$LIBCC"
 SUBARCH=
 t="$CFLAGS_C99FSE $CPPFLAGS $CFLAGS"
 
+if test "$ARCH" = "i386" ; then
+printf "checking whether compiler can use ebx in PIC asm constraints... "
+cat > "$tmpc" <<EOF
+int foo(int x) { __asm__ ( "" : "+b"(x) ); return x; }
+EOF
+if $CC $CFLAGS_C99FSE $CPPFLAGS $CFLAGS -fPIC \
+  -c -o /dev/null "$tmpc" >/dev/null 2>&1 ; then
+printf "yes\n"
+else
+printf "no\n"
+CFLAGS_AUTO="$CFLAGS_AUTO -DBROKEN_EBX_ASM"
+fi
+fi
+
 if test "$ARCH" = "x86_64" ; then
 trycppif __ILP32__ "$t" && ARCH=x32
 fi