arm: provide a PCS-compliant setjmp implementation

The previous setjmp-implementation (as a static inline function that
contained an 'asm volatile' sequence) was extremely fragile: (some
versions of) GCC optimised the set of registers.  One critical example
was the removal of 'r9' from the clobber list, if -ffixed-reg9 was
supplied.

To increase robustness and ensure PCS-compliant behaviour, the setjmp
and longjmp implementation are now in assembly and closely match what
one would expect to find in a libc implementation.

Signed-off-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Tested-by: Andy Yan <andy.yan@rock-chips.com>

1 files changed, 42 insertions, 0 deletions

diff --git a/arch/arm/lib/setjmp_aarch64.S b/arch/arm/lib/setjmp_aarch64.S
new file mode 100644
index 0000000000..b68edb86d6
--- /dev/null
+++ b/arch/arm/lib/setjmp_aarch64.S
@@ -0,0 +1,42 @@
+/*
+ * (C) 2017 Theobroma Systems Design und Consulting GmbH
+ *
+ * SPDX-License-Identifier:	GPL-2.0+
+ */
+
+#include <config.h>
+#include <asm/macro.h>
+#include <linux/linkage.h>
+
+.pushsection .text.setjmp, "ax"
+ENTRY(setjmp)
+	/* Preserve all callee-saved registers and the SP */
+	stp  x19, x20, [x0,#0]
+	stp  x21, x22, [x0,#16]
+	stp  x23, x24, [x0,#32]
+	stp  x25, x26, [x0,#48]
+	stp  x27, x28, [x0,#64]
+	stp  x29, x30, [x0,#80]
+	mov  x2, sp
+	str  x2, [x0, #96]
+	mov  x0, #0
+	ret
+ENDPROC(setjmp)
+.popsection
+
+.pushsection .text.longjmp, "ax"
+ENTRY(longjmp)
+	ldp  x19, x20, [x0,#0]
+	ldp  x21, x22, [x0,#16]
+	ldp  x23, x24, [x0,#32]
+	ldp  x25, x26, [x0,#48]
+	ldp  x27, x28, [x0,#64]
+	ldp  x29, x30, [x0,#80]
+	ldr  x2, [x0,#96]
+	mov  sp, x2
+	/* Move the return value in place, but return 1 if passed 0. */
+	adds x0, xzr, x1
+	csinc x0, x0, xzr, ne
+	ret
+ENDPROC(longjmp)
+.popsection