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.

12 files changed, 79 insertions, 17 deletions

diff --git a/arch/arm/pthread_arch.h b/arch/arm/pthread_arch.h
index 5f96f2b0..43a1c012 100644
--- a/arch/arm/pthread_arch.h
+++ b/arch/arm/pthread_arch.h
@@ -1,6 +1,9 @@
-typedef pthread_t (*__pthread_self_func_t)(void) __attribute__((const));
+typedef char *(*__ptr_func_t)(void) __attribute__((const));
 
-#define __pthread_self ((__pthread_self_func_t)0xffff0fe0)
+#define __pthread_self() \
+	((pthread_t)(((__ptr_func_t)0xffff0fe0)()+8-sizeof(struct pthread)))
+
+#define TLS_ABOVE_TP
+#define TP_ADJ(p) ((char *)(p) + sizeof(struct pthread) - 8)
 
-#define CANCEL_REG_SP 16
 #define CANCEL_REG_IP 18
diff --git a/arch/arm/reloc.h b/arch/arm/reloc.h
index c37af07b..b41314de 100644
--- a/arch/arm/reloc.h
+++ b/arch/arm/reloc.h
@@ -34,8 +34,8 @@ static inline void do_single_reloc(
 		break;
 	case R_ARM_TLS_TPOFF32:
 		*reloc_addr += def.sym
-			? def.sym->st_value - def.dso->tls_offset
-			: 0 - self->tls_offset;
+			? def.sym->st_value + def.dso->tls_offset + 8
+			: self->tls_offset + 8;
 		break;
 	}
 }
diff --git a/arch/i386/pthread_arch.h b/arch/i386/pthread_arch.h
index 0ea0aaca..1c06c764 100644
--- a/arch/i386/pthread_arch.h
+++ b/arch/i386/pthread_arch.h
@@ -5,5 +5,6 @@ static inline struct pthread *__pthread_self()
 	return self;
 }
 
-#define CANCEL_REG_SP 7
+#define TP_ADJ(p) (p)
+
 #define CANCEL_REG_IP 14
diff --git a/arch/microblaze/pthread_arch.h b/arch/microblaze/pthread_arch.h
index 6c0ab5ad..259d3d61 100644
--- a/arch/microblaze/pthread_arch.h
+++ b/arch/microblaze/pthread_arch.h
@@ -5,4 +5,6 @@ static inline struct pthread *__pthread_self()
 	return self;
 }
 
+#define TP_ADJ(p) (p)
+
 #define CANCEL_REG_IP 32
diff --git a/arch/mips/pthread_arch.h b/arch/mips/pthread_arch.h
index 77b7330d..f8e35ae4 100644
--- a/arch/mips/pthread_arch.h
+++ b/arch/mips/pthread_arch.h
@@ -1,12 +1,16 @@
 static inline struct pthread *__pthread_self()
 {
-	struct pthread *self;
 #ifdef __clang__
-	__asm__ __volatile__ (".word 0x7c03e83b ; move %0, $3" : "=r" (self) : : "$3" );
+	char *tp;
+	__asm__ __volatile__ (".word 0x7c03e83b ; move %0, $3" : "=r" (tp) : : "$3" );
 #else
-	__asm__ __volatile__ (".word 0x7c03e83b" : "=v" (self) );
+	register char *tp __asm__("$3");
+	__asm__ __volatile__ (".word 0x7c03e83b" : "=r" (tp) );
 #endif
-	return self;
+	return (pthread_t)(tp - 0x7000 - sizeof(struct pthread));
 }
 
+#define TLS_ABOVE_TP
+#define TP_ADJ(p) ((char *)(p) + sizeof(struct pthread) + 0x7000)
+
 #define CANCEL_REG_IP (3-(union {int __i; char __b;}){1}.__b)
diff --git a/arch/mips/reloc.h b/arch/mips/reloc.h
index 785489ed..f5e9c77b 100644
--- a/arch/mips/reloc.h
+++ b/arch/mips/reloc.h
@@ -23,6 +23,17 @@ static inline void do_single_reloc(
 	case R_MIPS_COPY:
 		memcpy(reloc_addr, (void *)sym_val, sym_size);
 		break;
+	case R_MIPS_TLS_DTPMOD32:
+		*reloc_addr = def.dso ? def.dso->tls_id : self->tls_id;
+		break;
+	case R_MIPS_TLS_DTPREL32:
+		*reloc_addr += def.sym->st_value;
+		break;
+	case R_MIPS_TLS_TPREL32:
+		*reloc_addr += def.sym
+			? def.sym->st_value + def.dso->tls_offset - 0x7000
+			: self->tls_offset - 0x7000;
+		break;
 	}
 }
 
diff --git a/arch/x86_64/pthread_arch.h b/arch/x86_64/pthread_arch.h
index 836187f5..77c8387c 100644
--- a/arch/x86_64/pthread_arch.h
+++ b/arch/x86_64/pthread_arch.h
@@ -5,5 +5,6 @@ static inline struct pthread *__pthread_self()
 	return self;
 }
 
-#define CANCEL_REG_SP 15
+#define TP_ADJ(p) (p)
+
 #define CANCEL_REG_IP 16
diff --git a/src/env/__init_tls.c b/src/env/__init_tls.c
index e70025d7..ab5992ce 100644
--- a/src/env/__init_tls.c
+++ b/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);
diff --git a/src/ldso/dynlink.c b/src/ldso/dynlink.c
index 5353979e..2ec8fb8c 100644
--- a/src/ldso/dynlink.c
+++ b/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;
diff --git a/src/thread/arm/tls.s b/src/thread/arm/tls.s
new file mode 100644
index 00000000..59736ac2
--- /dev/null
+++ b/src/thread/arm/tls.s
@@ -0,0 +1,4 @@
+.global __aeabi_read_tp
+.type __aeabi_read_tp,%function
+__aeabi_read_tp:
+	ldr pc,=0xffff0fe0
diff --git a/src/thread/pthread_create.c b/src/thread/pthread_create.c
index 627ff5fb..e67616e7 100644
--- a/src/thread/pthread_create.c
+++ b/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();
 
diff --git a/src/thread/pthread_self.c b/src/thread/pthread_self.c
index a93d8129..23dbaa5e 100644
--- a/src/thread/pthread_self.c
+++ b/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();