add aarch64 port

This adds complete aarch64 target support including bigendian subarch.

Some of the long double math functions are known to be broken otherwise
interfaces should be fully functional, but at this point consider this
port experimental.

Initial work on this port was done by Sireesh Tripurari and Kevin Bortis.

3 files changed, 115 insertions, 0 deletions

diff --git a/src/ldso/aarch64/dlsym.s b/src/ldso/aarch64/dlsym.s
new file mode 100644
index 00000000..be2dce52
--- /dev/null
+++ b/src/ldso/aarch64/dlsym.s
@@ -0,0 +1,5 @@
+.global dlsym
+.type dlsym,%function
+dlsym:
+	mov x2,x30
+	b __dlsym
diff --git a/src/ldso/aarch64/start.s b/src/ldso/aarch64/start.s
new file mode 100644
index 00000000..41d1d1e2
--- /dev/null
+++ b/src/ldso/aarch64/start.s
@@ -0,0 +1,18 @@
+.global _dlstart
+_dlstart:
+	ldr x0,[sp]
+	add x1,sp,#8
+	bl __dynlink
+	mov x1,sp
+	ldr x2,[x1],#8
+1:	sub x2,x2,1
+	ldr x3,[x1],#8
+	cmn x3,#1
+	b.eq 1b
+	add x2,x2,1
+	str x3,[x1,#-8]!
+	str x2,[x1,#-8]!
+	mov sp,x1
+	mov x1,x0
+	mov x0,#0
+	blr x1
diff --git a/src/ldso/aarch64/tlsdesc.s b/src/ldso/aarch64/tlsdesc.s
new file mode 100644
index 00000000..32064bd7
--- /dev/null
+++ b/src/ldso/aarch64/tlsdesc.s
@@ -0,0 +1,92 @@
+// long __tlsdesc_static(long *a)
+// {
+// 	return a[1];
+// }
+.global __tlsdesc_static
+.type __tlsdesc_static,@function
+__tlsdesc_static:
+	ldr x0,[x0,#8]
+	ret
+
+// long __tlsdesc_dynamic(long *a)
+// {
+// 	struct {size_t modidx,off;} *p = (void*)a[1];
+// 	size_t *dtv = *(size_t**)(tp + 16 - 8);
+// 	if (p->modidx <= dtv[0])
+// 		return dtv[p->modidx] + p->off - tp;
+// 	return __tls_get_addr(p) - tp;
+// }
+.global __tlsdesc_dynamic
+.type __tlsdesc_dynamic,@function
+__tlsdesc_dynamic:
+	stp x1,x2,[sp,#-32]!
+	stp x3,x4,[sp,#16]
+	mrs x1,tpidr_el0      // tp
+	ldr x0,[x0,#8]        // p
+	ldr x2,[x0]           // p->modidx
+	add x3,x1,#8
+	ldr x3,[x3]           // dtv
+	ldr x4,[x3]           // dtv[0]
+	cmp x2,x4
+	b.hi 1f
+	ldr x2,[x3,x2,lsl #3] // dtv[p->modidx]
+	ldr x0,[x0,#8]        // p->off
+	add x0,x0,x2
+2:	sub x0,x0,x1
+	ldp x3,x4,[sp,#16]
+	ldp x1,x2,[sp],#32
+	ret
+
+	// save all registers __tls_get_addr may clobber
+	// ugly because addr offset must be in [-512,509]
+1:	stp x29,x30,[sp,#-160]!
+	stp x5,x6,[sp,#16]
+	stp x7,x8,[sp,#32]
+	stp x9,x10,[sp,#48]
+	stp x11,x12,[sp,#64]
+	stp x13,x14,[sp,#80]
+	stp x15,x16,[sp,#96]
+	stp x17,x18,[sp,#112]
+	stp q0,q1,[sp,#128]
+	stp q2,q3,[sp,#-480]!
+	stp q4,q5,[sp,#32]
+	stp q6,q7,[sp,#64]
+	stp q8,q9,[sp,#96]
+	stp q10,q11,[sp,#128]
+	stp q12,q13,[sp,#160]
+	stp q14,q15,[sp,#192]
+	stp q16,q17,[sp,#224]
+	stp q18,q19,[sp,#256]
+	stp q20,q21,[sp,#288]
+	stp q22,q23,[sp,#320]
+	stp q24,q25,[sp,#352]
+	stp q26,q27,[sp,#384]
+	stp q28,q29,[sp,#416]
+	stp q30,q31,[sp,#448]
+	bl __tls_get_addr
+	mrs x1,tpidr_el0
+	ldp q4,q5,[sp,#32]
+	ldp q6,q7,[sp,#64]
+	ldp q8,q9,[sp,#96]
+	ldp q10,q11,[sp,#128]
+	ldp q12,q13,[sp,#160]
+	ldp q14,q15,[sp,#192]
+	ldp q16,q17,[sp,#224]
+	ldp q18,q19,[sp,#256]
+	ldp q20,q21,[sp,#288]
+	ldp q22,q23,[sp,#320]
+	ldp q24,q25,[sp,#352]
+	ldp q26,q27,[sp,#384]
+	ldp q28,q29,[sp,#416]
+	ldp q30,q31,[sp,#448]
+	ldp q2,q3,[sp],#480
+	ldp x5,x6,[sp,#16]
+	ldp x7,x8,[sp,#32]
+	ldp x9,x10,[sp,#48]
+	ldp x11,x12,[sp,#64]
+	ldp x13,x14,[sp,#80]
+	ldp x15,x16,[sp,#96]
+	ldp x17,x18,[sp,#112]
+	ldp q0,q1,[sp,#128]
+	ldp x29,x30,[sp],#160
+	b 2b