Re: [uclibc-ng-devel] [PATCH] linuxthreads.old: Fix segfault when pthread_mutex_unlock() called against unlocked NORMAL mutex.

Hi Leonid, Leonid Lisovskiy wrote,

Althought, it is undefined behavior, there is no reason for segfault.

Signed-off-by: Leonid Lisovskiy lly.dev@gmail.com

libpthread/linuxthreads/mutex.c | 3 ++- test/pthread/ex8-mtx-odd.c | 56 +++++++++++++++++++++++++++++++++++++++++ 2 files changed, 58 insertions(+), 1 deletion(-) create mode 100644 test/pthread/ex8-mtx-odd.c

diff --git a/libpthread/linuxthreads/mutex.c b/libpthread/linuxthreads/mutex.c index 8908414..8cb779b 100644 --- a/libpthread/linuxthreads/mutex.c +++ b/libpthread/linuxthreads/mutex.c @@ -171,7 +171,8 @@ int attribute_hidden __pthread_mutex_unlock(pthread_mutex_t * mutex) { switch (mutex->__m_kind) { case PTHREAD_MUTEX_ADAPTIVE_NP:

• __pthread_unlock(&mutex->__m_lock);

• if (mutex->__m_lock.__status != 0)

•  __pthread_unlock(&mutex->__m_lock);
  

  return 0; case PTHREAD_MUTEX_RECURSIVE_NP: if (mutex->__m_owner != thread_self())

Unfortunately this seems to break crisv32 support. (qemu-system-cris) The ex1 test is failing (never ends).

You can either verify this directly with OpenADK or embedded-test.sh from embedded-test.org: mksh embedded-test.sh --arch=crisv32 --debug --libc=uclibc-ng --test=libc --libc-source=uclibc-ng/

uclibc-ng directory is latest git with your linuxthreads.old patches.

Any idea? On which arch you have tested?

best regards Waldemar