rv64: use pc-relative addressing for globals

When extern was added, the rv64 code switched regular globals from
using la (a pseudo-instruction that uses pc-relative addressing
with auipc + addi) to lui + addi (absolute addressing). Presumably,
this was done to match gcc's output with -fPIC and -fno-PIC
respectively.  However, the key difference here is actually the
directive `%option pic` vs `%option nopic`, which determines whether
la uses the GOT or not.

This broke linking as PIE (for instance in compilers with
--enable-default-pie), even when there are no global references
outside the executable. The other architectures always use pc-relative
addressing, so rv64 should do the same.

It turns out there are more specific pseudo-instructions lla and
lga we can use instead. lla uses auipc+addi, and lga uses auipc+ld
from the GOT.

lga does not allow an offset. If you try to use one, it will assemble
without error, but fail at link time with

(.text+0x10): dangerous relocation: The addend isn't allowed for R_RISCV_GOT_HI20

For now, just add a check for this. A proper fix involves some
changes to rv64/isel like is done for amd64.

arm64/emit.c

@@ -292,12 +292,13 @@ loadaddr(Con *c, char *rn, E *e)
 			    "\tadd\tR, R, #:tprel_lo12_nc:SO\n";
 		break;
 	case SExt:
+		assert(c->bits.i == 0);
 		if (T.apple)
-			s = "\tadrp\tR, S@gotpageO\n"
-			    "\tldr\tR, [R, S@gotpageoffO]\n";
+			s = "\tadrp\tR, S@gotpage\n"
+			    "\tldr\tR, [R, S@gotpageoff]\n";
 		else
-			s = "\tadrp\tR, :got:SO\n"
-			    "\tldr\tR, [R, #:got_lo12:SO]\n";
+			s = "\tadrp\tR, :got:S\n"
+			    "\tldr\tR, [R, #:got_lo12:S]\n";
 		break;
 	}
 

arm64/isel.c

@@ -78,9 +78,8 @@ fixarg(Ref *pr, int k, int phi, Fn *fn)
 	switch (rtype(r0)) {
 	case RCon:
 		c = &fn->con[r0.val];
-		if (T.apple
-		&& c->type == CAddr
-		&& (c->sym.type & SThr)) {
+		if (c->type == CAddr && ((c->sym.type & SExt)
+		 || (T.apple && (c->sym.type & SThr)))) {
 			r1 = newtmp("isel", Kl, fn);
 			*pr = r1;
 			if (c->bits.i) {
@@ -91,16 +90,19 @@ fixarg(Ref *pr, int k, int phi, Fn *fn)
 				emit(Oadd, Kl, r1, r2, r3);
 				r1 = r2;
 			}
+			if (T.apple && (c->sym.type & SThr)) {
 				emit(Ocopy, Kl, r1, TMP(R0), R);
 				r1 = newtmp("isel", Kl, fn);
 				r2 = newtmp("isel", Kl, fn);
 				emit(Ocall, 0, R, r1, CALL(33));
 				emit(Ocopy, Kl, TMP(R0), r2, R);
 				emit(Oload, Kl, r1, r2, R);
+				r1 = r2;
+			}
 			cc = *c;
 			cc.bits.i = 0;
 			r3 = newcon(&cc, fn);
-			emit(Ocopy, Kl, r2, r3, R);
+			emit(Ocopy, Kl, r1, r3, R);
 			break;
 		}
 		if (KBASE(k) == 0 && phi)

rv64/emit.c

@@ -133,9 +133,13 @@ emitaddr(Con *c, FILE *f)
 {
 	assert((c->sym.type & ~SExt) == SGlo);
 	fputs(str(c->sym.id), f);
-	if (c->bits.i)
+	if (c->bits.i) {
+		/* TODO: fix isel to ensure no offset for SGlo */
+		if (c->sym.type & SExt)
+			die("extern with offset is not supported");
 		fprintf(f, "+%"PRIi64, c->bits.i);
 	}
+}
 
 static void
 emitf(char *s, Ins *i, Fn *fn, FILE *f)
@@ -233,14 +237,8 @@ loadaddr(Con *c, char *rn, FILE *f)
 
 	switch (c->sym.type) {
 	case SGlo:
-		fprintf(f, "\tlui %s, %%hi(", rn);
-		emitaddr(c, f);
-		fprintf(f, ")\n\taddi %s, %s, %%lo(", rn, rn);
-		emitaddr(c, f);
-		fputs(")\n", f);
-		break;
 	case SExt:
-		fprintf(f, "\tla %s, ", rn);
+		fprintf(f, "\t%s %s, ", c->sym.type == SExt ? "lga" : "lla", rn);
 		emitaddr(c, f);
 		fputc('\n', f);
 		break;