git: 2e38975e5aa4 - stable/13 - Add an implementation of .mcount on arm64

The branch stable/13 has been updated by andrew:

URL: https://cgit.FreeBSD.org/src/commit/?id=2e38975e5aa41c576b87e5766b51d28fed0034ee

commit 2e38975e5aa41c576b87e5766b51d28fed0034ee
Author:     Andrew Turner <andrew@FreeBSD.org>
AuthorDate: 2022-03-22 15:46:15 +0000
Commit:     Andrew Turner <andrew@FreeBSD.org>
CommitDate: 2022-04-19 09:18:02 +0000

    Add an implementation of .mcount on arm64
    
    To support cc -pg on arm64 we need to implement .mcount. As clang and
    gcc think it is function like it just needs to load the arguments
    to _mcount and call it.
    
    On gcc the first argument is passed in x0, however this is missing on
    clang so we need to load it from the stack. As it's the caller return
    address this will be at a known location.
    
    PR:             262709
    Reviewed by:    emaste (earlier version)
    Sponsored by:   The FreeBSD Foundation
    Differential Revision: https://reviews.freebsd.org/D34634
    
    (cherry picked from commit 16d5f9a1649e6a3e5841c6c41d6af9430ed9f064)
---
 lib/libc/aarch64/Symbol.map |  1 +
 sys/arm64/include/profile.h | 45 ++++++++++++++++++++++++++++++++++++++++++++-
 2 files changed, 45 insertions(+), 1 deletion(-)

diff --git a/lib/libc/aarch64/Symbol.map b/lib/libc/aarch64/Symbol.map
index d9121798140c..573dc0e6df0e 100644
--- a/lib/libc/aarch64/Symbol.map
+++ b/lib/libc/aarch64/Symbol.map
@@ -11,6 +11,7 @@ FBSD_1.0 {
 	/* PSEUDO syscalls */
 	_exit;
 
+	.mcount;
 	_setjmp;
 	_longjmp;
 	fabs;
diff --git a/sys/arm64/include/profile.h b/sys/arm64/include/profile.h
index afbfbdb44c0b..0b117d0ff442 100644
--- a/sys/arm64/include/profile.h
+++ b/sys/arm64/include/profile.h
@@ -72,8 +72,51 @@ void	mcount(uintfptr_t frompc, uintfptr_t selfpc);
 
 typedef __uintfptr_t    uintfptr_t;
 
-#define	_MCOUNT_DECL	void mcount
+#define	_MCOUNT_DECL \
+static void _mcount(uintfptr_t frompc, uintfptr_t selfpc) __used; \
+static void _mcount
+
+#ifdef __GNUCLIKE_ASM
+/*
+ * Call into _mcount. On arm64 the .mcount is a function so callers will
+ * handle caller saved registers. As we don't directly touch any callee
+ * saved registers we can just load the two arguments and use a tail call
+ * into the MI _mcount function.
+ *
+ * When building with gcc frompc will be in x0, however this is not the
+ * case on clang. As such we need to load it from the stack. As long as
+ * the caller follows the ABI this will load the correct value.
+ */
+#define	MCOUNT __asm(					\
+"	.text					\n"	\
+"	.align	6				\n"	\
+"	.type	.mcount,#function		\n"	\
+"	.globl	.mcount				\n"	\
+"	.mcount:				\n"	\
+"	.cfi_startproc				\n"	\
+	/* Load the caller return address as frompc */	\
+"	ldr	x0, [x29, #8]			\n"	\
+	/* Use our return address as selfpc */		\
+"	mov	x1, lr				\n"	\
+"	b	_mcount				\n"	\
+"	.cfi_endproc				\n"	\
+"	.size	.mcount, . - .mcount		\n"	\
+	);
+#if 0
+/*
+ * If clang passed frompc correctly we could implement it like this, however
+ * all clang versions we care about would need to be fixed before we could
+ * make this change.
+ */
+void
+mcount(uintfptr_t frompc)
+{
+	_mcount(frompc, __builtin_return_address(0));
+}
+#endif
+#else
 #define	MCOUNT
+#endif
 
 #endif /* _KERNEL */