add support for TLS variant I, presently needed for arm and mips

despite documentation that makes it sound a lot different, the only
ABI-constraint difference between TLS variants II and I seems to be
that variant II stores the initial TLS segment immediately below the
thread pointer (i.e. the thread pointer points to the end of it) and
variant I stores the initial TLS segment above the thread pointer,
requiring the thread descriptor to be stored below. the actual value
stored in the thread pointer register also tends to have per-arch
random offsets applied to it for silly micro-optimization purposes.

with these changes applied, TLS should be basically working on all
supported archs except microblaze. I'm still working on getting the
necessary information and a working toolchain that can build TLS
binaries for microblaze, but in theory, static-linked programs with
TLS and dynamic-linked programs where only the main executable uses
TLS should already work on microblaze.

alignment constraints have not yet been heavily tested, so it's
possible that this code does not always align TLS segments correctly
on archs that need TLS variant I.

src/env/__init_tls.c

@@ -15,11 +15,18 @@ void *__copy_tls(unsigned char *mem)
 	if (!image) return mem;
 	void **dtv = (void *)mem;
 	dtv[0] = (void *)1;
+#ifdef TLS_ABOVE_TP
+	mem += sizeof(void *) * 2;
+	mem += -((uintptr_t)mem + sizeof(struct pthread)) & (align-1);
+	td = (pthread_t)mem;
+	mem += sizeof(struct pthread);
+#else
 	mem += __libc.tls_size - sizeof(struct pthread);
 	mem -= (uintptr_t)mem & (align-1);
 	td = (pthread_t)mem;
 	td->dtv = dtv;
 	mem -= size;
+#endif
 	dtv[1] = mem;
 	memcpy(mem, image, len);
 	return td;
@@ -33,7 +40,7 @@ void *__tls_get_addr(size_t *v)
 static void *simple(void *p)
 {
 	*(void **)p = p;
-	return __set_thread_area(p) ? 0 : p;
+	return __set_thread_area(TP_ADJ(p)) ? 0 : p;
 }
 
 weak_alias(simple, __install_initial_tls);

src/ldso/dynlink.c

@@ -539,10 +539,16 @@ static struct dso *load_library(const char *name)
 		}
 		p->tls_id = ++tls_cnt;
 		tls_align = MAXP2(tls_align, p->tls_align);
+#ifdef TLS_ABOVE_TP
+		p->tls_offset = tls_offset + ( (tls_align-1) &
+			-(tls_offset + (uintptr_t)p->tls_image) );
+		tls_offset += p->tls_size;
+#else
 		tls_offset += p->tls_size + p->tls_align - 1;
 		tls_offset -= (tls_offset + (uintptr_t)p->tls_image)
 			& (p->tls_align-1);
 		p->tls_offset = tls_offset;
+#endif
 		p->new_dtv = (void *)(-sizeof(size_t) &
 			(uintptr_t)(p->name+strlen(p->name)+sizeof(size_t)));
 		p->new_tls = (void *)(p->new_dtv + n_th*(tls_cnt+1));
@@ -697,6 +703,18 @@ void *__copy_tls(unsigned char *mem)
 	void **dtv = (void *)mem;
 	dtv[0] = (void *)tls_cnt;
 
+#ifdef TLS_ABOVE_TP
+	mem += sizeof(void *) * (tls_cnt+1);
+	mem += -((uintptr_t)mem + sizeof(struct pthread)) & (tls_align-1);
+	td = (pthread_t)mem;
+	mem += sizeof(struct pthread);
+
+	for (p=head; p; p=p->next) {
+		if (!p->tls_id) continue;
+		dtv[p->tls_id] = mem + p->tls_offset;
+		memcpy(dtv[p->tls_id], p->tls_image, p->tls_len);
+	}
+#else
 	mem += libc.tls_size - sizeof(struct pthread);
 	mem -= (uintptr_t)mem & (tls_align-1);
 	td = (pthread_t)mem;
@@ -706,6 +724,7 @@ void *__copy_tls(unsigned char *mem)
 		dtv[p->tls_id] = mem - p->tls_offset;
 		memcpy(dtv[p->tls_id], p->tls_image, p->tls_len);
 	}
+#endif
 	td->dtv = dtv;
 	return td;
 }
@@ -753,9 +772,12 @@ void *__tls_get_addr(size_t *v)
 
 static void update_tls_size()
 {
-	size_t below_tp = (1+tls_cnt) * sizeof(void *) + tls_offset;
-	size_t above_tp = sizeof(struct pthread) + tls_align;
-	libc.tls_size = ALIGN(below_tp + above_tp, tls_align);
+	libc.tls_size = ALIGN(
+		(1+tls_cnt) * sizeof(void *) +
+		tls_offset +
+		sizeof(struct pthread) +
+		tls_align * 2,
+	tls_align);
 }
 
 void *__dynlink(int argc, char **argv)
@@ -866,9 +888,16 @@ void *__dynlink(int argc, char **argv)
 	}
 	if (app->tls_size) {
 		app->tls_id = tls_cnt = 1;
+#ifdef TLS_ABOVE_TP
+		app->tls_offset = 0;
+		tls_offset = app->tls_size
+			+ ( -((uintptr_t)app->tls_image + app->tls_size)
+			& (app->tls_align-1) );
+#else
 		tls_offset = app->tls_offset = app->tls_size
 			+ ( -((uintptr_t)app->tls_image + app->tls_size)
 			& (app->tls_align-1) );
+#endif
 		tls_align = MAXP2(tls_align, app->tls_align);
 	}
 	app->global = 1;

src/thread/arm/tls.s

@@ -0,0 +1,4 @@
+.global __aeabi_read_tp
+.type __aeabi_read_tp,%function
+__aeabi_read_tp:
+	ldr pc,=0xffff0fe0

src/thread/pthread_create.c

@@ -147,7 +147,7 @@ int pthread_create(pthread_t *restrict res, const pthread_attr_t *restrict attr,
 	new->canary = self->canary;
 
 	a_inc(&libc.threads_minus_1);
-	ret = __clone(start, stack, flags, new, &new->tid, new, &new->tid);
+	ret = __clone(start, stack, flags, new, &new->tid, TP_ADJ(new), &new->tid);
 
 	__release_ptc();
 

src/thread/pthread_self.c

@@ -10,7 +10,7 @@ static int init_main_thread()
 {
 	__syscall(SYS_rt_sigprocmask, SIG_UNBLOCK,
 		SIGPT_SET, 0, __SYSCALL_SSLEN);
-	if (__set_thread_area(main_thread) < 0) return -1;
+	if (__set_thread_area(TP_ADJ(main_thread)) < 0) return -1;
 	main_thread->canceldisable = libc.canceldisable;
 	main_thread->tsd = (void **)__pthread_tsd_main;
 	main_thread->errno_ptr = __errno_location();