about summary refs log tree commit diff
path: root/rt/tst-aio2.c
diff options
context:
space:
mode:
authorStefan Liebler <stli@linux.ibm.com>2019-02-18 16:12:01 +0100
committerStefan Liebler <stli@linux.ibm.com>2019-02-18 16:12:01 +0100
commit6c29942cbf059aca47fd4bbd852ea42c9d46b71f (patch)
tree0f51ed545083ac82cb9ed7d115d2c2c6a48d6891 /rt/tst-aio2.c
parentad18dadec945329f95b53b63ed69f91878a329df (diff)
downloadglibc-6c29942cbf059aca47fd4bbd852ea42c9d46b71f.tar.gz
glibc-6c29942cbf059aca47fd4bbd852ea42c9d46b71f.tar.xz
glibc-6c29942cbf059aca47fd4bbd852ea42c9d46b71f.zip
misc/tst-clone3: Fix waiting for exited thread.
From time to time the test misc/tst-clone3 fails with a timeout.
Then futex_wait is blocking.  Usually ctid should be set to zero
due to CLONE_CHILD_CLEARTID and the futex should be waken up.
But the fail occures if the thread has already exited before
ctid is set to the return value of clone().  Then futex_wait() will
block as there will be nobody who wakes the futex up again.

This patch initializes ctid to a known value before calling clone
and the kernel is the only one who updates the value to zero after clone.
If futex_wait is called then it is either waked up due to the exited thread
or the futex syscall fails as *ctid_ptr is already zero instead of the
specified value 1.

ChangeLog:

	* sysdeps/unix/sysv/linux/tst-clone3.c (do_test):
	Initialize ctid with a known value and remove update of ctid
	after clone.
	(wait_tid): Adjust arguments and call futex_wait with ctid_val
	as assumed current value of ctid_ptr.
Diffstat (limited to 'rt/tst-aio2.c')
0 files changed, 0 insertions, 0 deletions