o Let rtld(1) set up psABI user trap handlers prior to executing the

objects' init functions instead of doing the setup via a constructor
  in libc as the init functions may already depend on these handlers
  to be in place. This gets us rid of:
  - the undefined order in which libc constructors as __guard_setup()
    and jemalloc_constructor() are executed WRT __sparc_utrap_setup(),
  - the requirement to link libc last so __sparc_utrap_setup() gets
    called prior to constructors in other libraries (see r122883).
  For static binaries, crt1.o still sets up the user trap handlers.
o Move misplaced prototypes for MD functions in to the MD prototype
  section of rtld.h.
o Sprinkle nitems().

lib/libc/sparc64/sys/__sparc_utrap_setup.c

@@ -29,13 +29,11 @@
 #include <sys/cdefs.h>
 __FBSDID("$FreeBSD$");
 
-#include <sys/types.h>
+#include <sys/param.h>
 
 #include <machine/utrap.h>
 #include <machine/sysarch.h>
 
-#include <stdlib.h>
-
 #include "__sparc_utrap_private.h"
 
 static const struct sparc_utrap_args ua[] = {
@@ -47,10 +45,10 @@ static const struct sparc_utrap_args ua[] = {
 };
 
 static const struct sparc_utrap_install_args uia[] = {
-	{ sizeof (ua) / sizeof (*ua), ua }
+	{ nitems(ua), ua }
 };
 
-void __sparc_utrap_setup(void) __attribute__((constructor));
+void __sparc_utrap_setup(void);
 
 void
 __sparc_utrap_setup(void)

libexec/rtld-elf/aarch64/reloc.c

@@ -306,6 +306,13 @@ reloc_jmpslot(Elf_Addr *where, Elf_Addr target, const Obj_Entry *defobj,
 void
 ifunc_init(Elf_Auxinfo aux_info[__min_size(AT_COUNT)] __unused)
 {
+
+}
+
+void
+pre_init(void)
+{
+
 }
 
 /*

libexec/rtld-elf/amd64/reloc.c

@@ -486,6 +486,12 @@ ifunc_init(Elf_Auxinfo aux_info[__min_size(AT_COUNT)] __unused)
 	}
 }
 
+void
+pre_init(void)
+{
+
+}
+
 void
 allocate_initial_tls(Obj_Entry *objs)
 {

libexec/rtld-elf/arm/reloc.c

@@ -478,6 +478,13 @@ reloc_jmpslot(Elf_Addr *where, Elf_Addr target, const Obj_Entry *defobj,
 void
 ifunc_init(Elf_Auxinfo aux_info[__min_size(AT_COUNT)] __unused)
 {
+
+}
+
+void
+pre_init(void)
+{
+
 }
 
 void

libexec/rtld-elf/i386/reloc.c

@@ -472,6 +472,12 @@ ifunc_init(Elf_Auxinfo aux_info[__min_size(AT_COUNT)] __unused)
 	}
 }
 
+void
+pre_init(void)
+{
+
+}
+
 void
 allocate_initial_tls(Obj_Entry *objs)
 {

libexec/rtld-elf/mips/reloc.c

@@ -745,6 +745,13 @@ reloc_jmpslot(Elf_Addr *where, Elf_Addr target, const Obj_Entry *defobj,
 void
 ifunc_init(Elf_Auxinfo aux_info[__min_size(AT_COUNT)] __unused)
 {
+
+}
+
+void
+pre_init(void)
+{
+
 }
 
 void

libexec/rtld-elf/powerpc/reloc.c

@@ -628,6 +628,13 @@ init_pltgot(Obj_Entry *obj)
 void
 ifunc_init(Elf_Auxinfo aux_info[__min_size(AT_COUNT)] __unused)
 {
+
+}
+
+void
+pre_init(void)
+{
+
 }
 
 void

libexec/rtld-elf/powerpc64/reloc.c

@@ -533,6 +533,13 @@ init_pltgot(Obj_Entry *obj)
 void
 ifunc_init(Elf_Auxinfo aux_info[__min_size(AT_COUNT)] __unused)
 {
+
+}
+
+void
+pre_init(void)
+{
+
 }
 
 void

libexec/rtld-elf/riscv/reloc.c

@@ -368,6 +368,13 @@ reloc_non_plt(Obj_Entry *obj, Obj_Entry *obj_rtld, int flags,
 void
 ifunc_init(Elf_Auxinfo aux_info[__min_size(AT_COUNT)] __unused)
 {
+
+}
+
+void
+pre_init(void)
+{
+
 }
 
 void

libexec/rtld-elf/rtld.c

@@ -755,6 +755,12 @@ _rtld(Elf_Addr *sp, func_ptr_type *exit_proc, Obj_Entry **objp)
 	    obj_main->fini_array = (Elf_Addr)NULL;
     }
 
+    /*
+     * Execute MD initializers required before we call the objects'
+     * init functions.
+     */
+    pre_init();
+
     wlock_acquire(rtld_bind_lock, &lockstate);
     if (obj_main->crt_no_init)
 	preinit_main();

libexec/rtld-elf/rtld.h

@@ -374,8 +374,6 @@ void dump_Elf_Rela(Obj_Entry *, const Elf_Rela *, u_long);
 unsigned long elf_hash(const char *);
 const Elf_Sym *find_symdef(unsigned long, const Obj_Entry *,
   const Obj_Entry **, int, SymCache *, struct Struct_RtldLockState *);
-void ifunc_init(Elf_Auxinfo[__min_size(AT_COUNT)]);
-void init_pltgot(Obj_Entry *);
 void lockdflt_init(void);
 void digest_notes(Obj_Entry *, Elf_Addr, Elf_Addr);
 Obj_Entry *globallist_curr(const Obj_Entry *obj);
@@ -405,6 +403,9 @@ int reloc_plt(Obj_Entry *);
 int reloc_jmpslots(Obj_Entry *, int flags, struct Struct_RtldLockState *);
 int reloc_iresolve(Obj_Entry *, struct Struct_RtldLockState *);
 int reloc_gnu_ifunc(Obj_Entry *, int flags, struct Struct_RtldLockState *);
+void ifunc_init(Elf_Auxinfo[__min_size(AT_COUNT)]);
+void pre_init(void);
+void init_pltgot(Obj_Entry *);
 void allocate_initial_tls(Obj_Entry *);
 
 #endif /* } */

libexec/rtld-elf/sparc64/reloc.c

@@ -367,8 +367,7 @@ reloc_nonplt_object(Obj_Entry *obj, const Elf_Rela *rela, SymCache *cache,
 	 * Note: R_SPARC_TLS_TPOFF64 must be the numerically largest
 	 * relocation type.
 	 */
-	if (type >= sizeof(reloc_target_bitmask) /
-	    sizeof(*reloc_target_bitmask)) {
+	if (type >= nitems(reloc_target_bitmask)) {
 		_rtld_error("%s: Unsupported relocation type %d in non-PLT "
 		    "object\n", obj->path, type);
 		return (-1);
@@ -793,6 +792,16 @@ reloc_jmpslot(Elf_Addr *wherep, Elf_Addr target, const Obj_Entry *obj,
 void
 ifunc_init(Elf_Auxinfo aux_info[__min_size(AT_COUNT)] __unused)
 {
+
+}
+
+extern void __sparc_utrap_setup(void);
+
+void
+pre_init(void)
+{
+
+	__sparc_utrap_setup();
 }
 
 /*
@@ -823,6 +832,7 @@ init_pltgot(Obj_Entry *obj)
 static void
 install_plt(Elf_Word *pltgot, Elf_Addr proc)
 {
+
 	pltgot[0] = SAVE;
 	flush(pltgot, 0);
 	pltgot[1] = SETHI_l0 | HIVAL(proc, 42);