bpf: cpumap fix potential lost wake-up problem
authorJesper Dangaard Brouer <brouer@redhat.com>
Mon, 23 Oct 2017 17:39:28 +0000 (19:39 +0200)
committerDavid S. Miller <davem@davemloft.net>
Tue, 24 Oct 2017 09:40:22 +0000 (18:40 +0900)
commit31749468c3f9d77927ed3144259dc208e6625ede
tree33d083824b454cd344121d6ef8ed2ac1d37afeb4
parent7f6b437e9b82a6d702a7f8f75c83ffdec6e03c54
bpf: cpumap fix potential lost wake-up problem

As pointed out by Michael, commit 1c601d829ab0 ("bpf: cpumap xdp_buff
to skb conversion and allocation") contains a classical example of the
potential lost wake-up problem.

We need to recheck the condition __ptr_ring_empty() after changing
current->state to TASK_INTERRUPTIBLE, this avoids a race between
wake_up_process() and schedule(). After this, a race with
wake_up_process() will simply change the state to TASK_RUNNING, and
the schedule() call not really put us to sleep.

Fixes: 1c601d829ab0 ("bpf: cpumap xdp_buff to skb conversion and allocation")
Reported-by: "Michael S. Tsirkin" <mst@redhat.com>
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
kernel/bpf/cpumap.c