arm: lib: fix push/pop-section directives

With the existing code, function symbols are defined in .text, and the
body is defined in .text.xxx. This causes (at least some version of) the
linker not to emit the function body into the final binary, since it's
part of a different section to the symbols being referenced. This of
course causes a wide variety of failures.

This change moves the push/pop-section directives before the function
symbols, and after any relate ENDPROC macro invocations, so that symbols
and bodies are all in the "pushed" sections, and thus the function bodies
are emitted into the binary.

This solves (at least) the boot problems currently seen on Tegra systems
that use SPL (i.e. all ARMv7 Tegras).

Fixes: 13b0a91a6d48 ("arm: lib: Split asm symbols into different .text subsections")
Cc: Marek Vasut <marex@denx.de>
Cc: Tom Warren <twarren@nvidia.com>
Cc: Simon Glass <sjg@chromium.org>
Cc: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
diff --git a/arch/arm/lib/lshrdi3.S b/arch/arm/lib/lshrdi3.S
index f710ccb..ead33e5 100644
--- a/arch/arm/lib/lshrdi3.S
+++ b/arch/arm/lib/lshrdi3.S
@@ -15,9 +15,9 @@
 #define ah r1
 #endif
 
+.pushsection .text.__lshldi3, "ax"
 ENTRY(__lshrdi3)
 ENTRY(__aeabi_llsr)
-.pushsection .text.__lshldi3, "ax"
 
 	subs	r3, r2, #32
 	rsb	ip, r2, #32
@@ -29,6 +29,6 @@
 	mov	ah, ah, lsr r2
 	ret	lr
 
-.popsection
 ENDPROC(__lshrdi3)
 ENDPROC(__aeabi_llsr)
+.popsection