From 81e18eb3cd61f7b68a8f99b3e6c728f2a4d79221 Mon Sep 17 00:00:00 2001
From: Rich Felker <dalias@aerifal.cx>
Date: Tue, 14 Apr 2015 23:45:08 -0400
Subject: use hidden __tls_get_new for tls/tlsdesc lookup fallback cases

previously, the dynamic tlsdesc lookup functions and the i386
special-ABI ___tls_get_addr (3 underscores) function called
__tls_get_addr when the slot they wanted was not already setup;
__tls_get_addr would then in turn also see that it's not setup and
call __tls_get_new.

calling __tls_get_new directly is both more efficient and avoids the
issue of calling a non-hidden (public API/ABI) function from asm.

for the special i386 function, a weak reference to __tls_get_new is
used since this function is not defined when static linking (the code
path that needs it is unreachable in static-linked programs).
---
 src/ldso/aarch64/tlsdesc.s | 6 ++++--
 src/ldso/i386/tlsdesc.s    | 4 +++-
 src/ldso/x86_64/tlsdesc.s  | 4 +++-
 src/thread/i386/tls.s      | 4 +++-
 4 files changed, 13 insertions(+), 5 deletions(-)

diff --git a/src/ldso/aarch64/tlsdesc.s b/src/ldso/aarch64/tlsdesc.s
index 32064bd7..05ddbab4 100644
--- a/src/ldso/aarch64/tlsdesc.s
+++ b/src/ldso/aarch64/tlsdesc.s
@@ -8,6 +8,8 @@ __tlsdesc_static:
 	ldr x0,[x0,#8]
 	ret
 
+.hidden __tls_get_new
+
 // long __tlsdesc_dynamic(long *a)
 // {
 // 	struct {size_t modidx,off;} *p = (void*)a[1];
@@ -37,7 +39,7 @@ __tlsdesc_dynamic:
 	ldp x1,x2,[sp],#32
 	ret
 
-	// save all registers __tls_get_addr may clobber
+	// save all registers __tls_get_new may clobber
 	// ugly because addr offset must be in [-512,509]
 1:	stp x29,x30,[sp,#-160]!
 	stp x5,x6,[sp,#16]
@@ -63,7 +65,7 @@ __tlsdesc_dynamic:
 	stp q26,q27,[sp,#384]
 	stp q28,q29,[sp,#416]
 	stp q30,q31,[sp,#448]
-	bl __tls_get_addr
+	bl __tls_get_new
 	mrs x1,tpidr_el0
 	ldp q4,q5,[sp,#32]
 	ldp q6,q7,[sp,#64]
diff --git a/src/ldso/i386/tlsdesc.s b/src/ldso/i386/tlsdesc.s
index 3ac61297..00a9fc09 100644
--- a/src/ldso/i386/tlsdesc.s
+++ b/src/ldso/i386/tlsdesc.s
@@ -5,6 +5,8 @@ __tlsdesc_static:
 	mov 4(%eax),%eax
 	ret
 
+.hidden __tls_get_new
+
 .global __tlsdesc_dynamic
 .type __tlsdesc_dynamic,@function
 __tlsdesc_dynamic:
@@ -22,6 +24,6 @@ __tlsdesc_dynamic:
 	pop %edx
 	ret
 1:	push %eax
-	call __tls_get_addr
+	call __tls_get_new
 	pop %ecx
 	jmp 2b
diff --git a/src/ldso/x86_64/tlsdesc.s b/src/ldso/x86_64/tlsdesc.s
index 57b78e0b..eb21b9de 100644
--- a/src/ldso/x86_64/tlsdesc.s
+++ b/src/ldso/x86_64/tlsdesc.s
@@ -5,6 +5,8 @@ __tlsdesc_static:
 	mov 8(%rax),%rax
 	ret
 
+.hidden __tls_get_new
+
 .global __tlsdesc_dynamic
 .type __tlsdesc_dynamic,@function
 __tlsdesc_dynamic:
@@ -29,7 +31,7 @@ __tlsdesc_dynamic:
 	push %r10
 	push %r11
 	mov %rax,%rdi
-	call __tls_get_addr
+	call __tls_get_new
 	pop %r11
 	pop %r10
 	pop %r9
diff --git a/src/thread/i386/tls.s b/src/thread/i386/tls.s
index 6e01adf2..76d5d462 100644
--- a/src/thread/i386/tls.s
+++ b/src/thread/i386/tls.s
@@ -10,6 +10,8 @@ ___tls_get_addr:
 	add (%edx,%ecx,4),%eax
 	ret
 1:	push %eax
-	call __tls_get_addr
+.weak __tls_get_new
+.hidden __tls_get_new
+	call __tls_get_new
 	pop %edx
 	ret
-- 
cgit v1.2.1