csu: Implement _start using as to satisfy unwinders on i386

The right unwinding stop indicator should be CFI-undefined PC.
https://dwarfstd.org/doc/Dwarf3.pdf - page 118:
If a Return Address register is defined in the virtual unwind table,
and its rule is undefined (for example, by DW_CFA_undefined), then
there is no return address and no call address, and the virtual
unwind of stack activations is complete.

The hack localizing _start1 symbol removed.

Reviewed by:		kib
Differential Revision:	https://reviews.freebsd.org/D40624

lib/csu/Makefile.inc

@@ -42,9 +42,6 @@ CLEANFILES+=	crti_s.o ${CRT1SRC:C/.[S|c]$/.o/}
 
 crt1.o:	${CRT1SRC:C/.[S|c]$/.o/} ${CRT1OBJS} ${CRT1OBJ}
 	${LD} ${_LDFLAGS} -o ${.TARGET} -r ${.ALLSRC:M*.o}
-.if ${MACHINE_ARCH} == "i386"
-	${OBJCOPY} --localize-symbol _start1 ${.TARGET}
-.endif
 
 gcrt1_c.o: ${CRT1SRC}
 	${CC} ${CFLAGS} -DGCRT -c -o ${.TARGET} ${.CURDIR}/${CRT1SRC}
@@ -57,9 +54,6 @@ Scrt1_c.o: ${CRT1SRC}
 
 Scrt1.o: Scrt1_c.o ${CRT1OBJS} ${CRT1OBJ}
 	${LD} ${_LDFLAGS} -o ${.TARGET} -r ${.ALLSRC:M*.o}
-.if ${MACHINE_ARCH} == "i386"
-	${OBJCOPY} --localize-symbol _start1 ${.TARGET}
-.endif
 
 crtbegin.o: crtbegin.c
 crtbeginS.o: crtbegin.c

lib/csu/i386/Makefile

@@ -4,6 +4,7 @@
 
 CFLAGS+=	-I${.CURDIR}
 
-CRT1OBJS+=	crt1_s.o
+CRT1SRC=	crt1_s.S
+CRT1OBJ=	crt1_c.o
 
 .include <bsd.lib.mk>

lib/csu/i386/crt1_c.c

@@ -29,25 +29,6 @@
 #include <sys/cdefs.h>
 __FBSDID("$FreeBSD$");
 
-#include "libc_private.h"
 #include "csu_common.h"
 
-void _start(char *, ...);
-void _start1(void (*)(void), int, char *[]) __dead2;
-
-/* The entry function, C part. */
-void
-_start1(void (*cleanup)(void), int argc, char *argv[])
-{
-	char **env;
-
-	env = argv + argc + 1;
-#ifdef GCRT
-	__libc_start1_gcrt(argc, argv, env, cleanup, main, &eprol, &etext);
-__asm__("eprol:");
-#else
-	__libc_start1(argc, argv, env, cleanup, main);
-#endif
-}
-
-__asm(".hidden	_start1");
+void _start(char *, ...) __dead2;

lib/csu/i386/crt1_s.S

@@ -31,19 +31,63 @@ __FBSDID("$FreeBSD$");
 	.type	_start, @function
 _start:
 	.cfi_startproc
+	.cfi_undefined %eip
+	popl	%esi		# Pop argc
+	.cfi_def_cfa_offset -4
+	movl	%esp,%edi	# argv starts at stack top
 	xorl	%ebp,%ebp
 	pushl	%ebp
-	.cfi_def_cfa_offset 4
+	.cfi_def_cfa_offset 0
 	movl	%esp,%ebp
-	.cfi_offset %ebp,-8
+	.cfi_offset %ebp,-4
 	.cfi_def_cfa_register %ebp
 	andl	$0xfffffff0,%esp # align stack
-	leal	8(%ebp),%eax
-	subl	$4,%esp
-	pushl	%eax		# argv
-	pushl	4(%ebp)		# argc
+
+#ifdef GCRT
+	subl	$4,%esp		# Align stack for 7 arguments
+	pushl	$etext
+	pushl	$eprol
+eprol:
+#else
+	subl	$12,%esp	# Align stack for 5 arguments
+#endif /* GCRT */
+
+#ifdef PIC
+	calll	1f
+1:	popl	%ebx
+	addl	$_GLOBAL_OFFSET_TABLE_+[.-1b],%ebx
+	leal	main@GOTOFF(%ebx),%eax
+	pushl	%eax
+#else
+	pushl	$main
+#endif /* PIC */
+
 	pushl	%edx		# rtld cleanup
-	call	_start1
+	/* env = argv + argc + 1 */
+	movl	%edi,%eax	# env = argv
+	movl	%esi,%ecx
+	shll	$2,%ecx		# argc * 4
+	addl	%ecx,%eax	# env += argc
+	addl	$4,%eax		# env += 1
+	pushl	%eax		# env
+	pushl	%edi		# argv
+	pushl	%esi		# argc
+
+#ifdef GCRT
+	/*
+	 * __libc_start1_gcrt(argc, argv, env, cleanup, main, &eprol, &etext);
+	 */
+	calll	__libc_start1_gcrt
+#else
+	/*
+	 * __libc_start1(argc, argv, env, cleanup, main);
+	 */
+#ifdef PIC
+	calll	__libc_start1@PLT
+#else
+	calll	__libc_start1
+#endif
+#endif /* GCRT */
 	int3
 	.cfi_endproc
 	.size	_start, . - _start