[libunwind][RISC-V] Rewrite testcase with C as possible.

[kito-cheng]

Summary

Fix #60472

The testcase is writen in all inline asm but it seems not well
maintained for the CFI directive, of cause we can fix that, but this
patch also contain another issue is it use s0 and s1 without
store/restore.

This patch proposed another way to testing that, use inline asm to
generate dummy def and use, so compiler will generate store/restore for
the vector register, and then generate the CFI directives.

Also check __riscv_vector as the testcase guard, because the testcase
will read vlenb which is only available when V or zve* extensions is
present.

[asb]

LGTM in that the reasoning seems sensible, but this also isn't an area I've done a huge amount of work in. If you have a high degree of confidence about the change I think it's good to commit - if not, best to wait for another review.

[MaskRay]

libunwind/test/unwind_scalable_vectors.pass.cpp

29

Add noinline and make it static

31

"#" can be ""

libunwind/test/unwind_scalable_vectors.pass.cpp

 // -*- C++ -*-
 //===----------------------------------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 // REQUIRES: linux && target={{riscv64-.+}}
 
 #undef NDEBUG
 #include <assert.h>
 #include <libunwind.h>
 
-// Check correct unwinding of frame with VLENB-sized objects (vector registers):
-// 1. Save return address (ra) in temporary register.
-// 2. Load VLENB (vector length in bytes) and substract it from current stack
-//    pointer (sp) - equivalent to one vector register on stack frame.
-// 3. Set DWARF cannonical frame address (CFA) to "sp + vlenb" expresssion so it
-//    can be correctly unwinded.
-// 4. Call stepper() function and check that 2 unwind steps are successful -
-//    from stepper() into foo() and from foo() into main().
-// 5. Restore stack pointer and return address.
-__attribute__((naked)) static void foo() {
-  __asm__(".cfi_startproc\n"
-          "mv s0, ra\n"
-          "csrr  s1, vlenb\n"
-          "sub sp, sp, s1\n"
-          "# .cfi_def_cfa_expression sp + vlenb\n"
-          ".cfi_escape 0x0f, 0x07, 0x72, 0x00, 0x92, 0xa2, 0x38, 0x00, 0x22\n"
-          "call stepper\n"
-          "add sp, sp, s1\n"
-          "mv ra, s0\n"
-          "ret\n"
-          ".cfi_endproc\n");
-}
-
-extern "C" void stepper() {
+#ifdef __riscv_vector
+__attribute__((noinline)) extern "C" void stepper() {
   unw_cursor_t cursor;
   unw_context_t uc;
   unw_getcontext(&uc);
   unw_init_local(&cursor, &uc);
   // Stepping into foo() should succeed.
   assert(unw_step(&cursor) > 0);
   // Stepping past foo() should succeed, too.
   assert(unw_step(&cursor) > 0);
 }
 
+// Check correct unwinding of frame with VLENB-sized objects (vector registers).
+__attribute__((noinline)) static void foo() {
+  __rvv_int32m1_t v;
+  asm volatile("" : "=vr"(v)); // Dummy inline asm to def v.
+  stepper();                   // def-use of v has cross the function, so that
+                               // will triger spill/reload to/from the stack.
+  asm volatile("" ::"vr"(v));  // Dummy inline asm to use v.
+}
+
 int main() { foo(); }
+#else
+int main() { return 0; }
+#endif