Fix incorrect double-checked locking related to _res_hconf.initialized.

_res_hconf.initialized was not suitable for use in a multi-threaded
environment due to the lack of atomics and memory barriers.  Use of it was
also unnecessary because _res_hconf_init did the right thing by using
__libc_once.  This patch fixes the glibc-internal uses by just calling
_res_hconf_init unconditionally, and switches to a release MO atomic store
for _res_hconf.initialized to fix the glibc side of the synchronization
problem (which will maintain backward compatibility, but cannot fix the
lack of acquire MO on any glibc-external loads).

	[BZ #20477]
	* resolv/res_hconf.c (do_init): Use atomic access.
	* resolv/res_hconf.h: Add comments.
	* nscd/aicache.c (addhstaiX): Call _res_hconf_init unconditionally.
	* nss/getXXbyYY_r.c (REENTRANT_NAME): Likewise.
	* sysdeps/posix/getaddrinfo.c (gaih_inet): Likewise.

1 files changed, 9 insertions, 0 deletions

diff --git a/resolv/res_hconf.h b/resolv/res_hconf.h
index b97734d..a3d23f3 100644
--- a/resolv/res_hconf.h
+++ b/resolv/res_hconf.h
@@ -25,6 +25,15 @@
 
 struct hconf
 {
+  /* We keep the INITIALIZED member only for backwards compatibility.  New
+     code should just call _res_hconf_init unconditionally.  For this field
+     to be used safely, users must ensure that either (1) a call to
+     _res_hconf_init happens-before any load from INITIALIZED, or (2) an
+     assignment of zero to INITIALIZED happens-before any load from it, and
+     these loads use acquire MO if the intent is to skip calling
+     _res_hconf_init if the load returns a nonzero value.  Such acquire MO
+     loads will then synchronize with the release MO store to INITIALIZED
+     in do_init in res_hconf.c; see pthread_once for more detail.  */
   int initialized;
   int unused1;
   int unused2[4];