[v2,2/8] linux: Use __libc_single_threaded on fork
Commit Message
Changes from v1:
* Fix __libc_single_threaded backwards usage.
---
Checked on x86_64-linux-gnu.
---
sysdeps/nptl/fork.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
Comments
* Adhemerval Zanella via Libc-alpha:
> Changes from v1:
> * Fix __libc_single_threaded backwards usage.
> ---
> Checked on x86_64-linux-gnu.
> ---
> sysdeps/nptl/fork.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/sysdeps/nptl/fork.c b/sysdeps/nptl/fork.c
> index 4fb4c2d254..f78267b68c 100644
> --- a/sysdeps/nptl/fork.c
> +++ b/sysdeps/nptl/fork.c
> @@ -34,6 +34,7 @@
> #include <malloc/malloc-internal.h>
> #include <nss/nss_database.h>
> #include <unwind-link.h>
> +#include <sys/single_threaded.h>
>
> static void
> fresetlockfiles (void)
> @@ -55,7 +56,7 @@ __libc_fork (void)
> handlers in the single-thread case, to make fork safer to use in
> signal handlers. POSIX requires that fork is async-signal-safe,
> but our current fork implementation is not. */
> - bool multiple_threads = THREAD_GETMEM (THREAD_SELF, header.multiple_threads);
> + bool multiple_threads = __libc_single_threaded == 0;
>
> __run_fork_handlers (atfork_run_prepare, multiple_threads);
Okay, this code isn't performance-sensitive, given that fork itself is
so expensive, so it doesn't matter which single-thread check is used.
Thanks,
Florian
@@ -34,6 +34,7 @@
#include <malloc/malloc-internal.h>
#include <nss/nss_database.h>
#include <unwind-link.h>
+#include <sys/single_threaded.h>
static void
fresetlockfiles (void)
@@ -55,7 +56,7 @@ __libc_fork (void)
handlers in the single-thread case, to make fork safer to use in
signal handlers. POSIX requires that fork is async-signal-safe,
but our current fork implementation is not. */
- bool multiple_threads = THREAD_GETMEM (THREAD_SELF, header.multiple_threads);
+ bool multiple_threads = __libc_single_threaded == 0;
__run_fork_handlers (atfork_run_prepare, multiple_threads);