From: Anoob Soman
Date: Wed, 15 Feb 2017 20:25:39 +
> Commit 6664498280cf ("packet: call fanout_release, while UNREGISTERING a
> netdev"), unfortunately, introduced the following issues.
...
> To fix the above problems, remove the call to fanout_release() under
> rcu_read_lock(). Instead, ca
On Wed, 2017-02-15 at 21:30 +, Anoob Soman wrote:
> If statement looks like this.
> if (atomic_dec_and_test(sk_ref))
> list_del(f->list);
> else
> f = NULL;
>
> there are no multiple lines under if.
Yes, I just noticed that.
Thanks.
On Wed, 2017-02-15 at 13:02 -0800, Eric Dumazet wrote:
> On Wed, 2017-02-15 at 20:25 +, Anoob Soman wrote:
>
> > +static struct packet_fanout *fanout_release(struct sock *sk)
> > {
> > struct packet_sock *po = pkt_sk(sk);
> > struct packet_fanout *f;
> > @@ -1728,17 +1736,17 @@ static
On 15/02/17 21:02, Eric Dumazet wrote:
On Wed, 2017-02-15 at 20:25 +, Anoob Soman wrote:
+static struct packet_fanout *fanout_release(struct sock *sk)
{
struct packet_sock *po = pkt_sk(sk);
struct packet_fanout *f;
@@ -1728,17 +1736,17 @@ static void fanout_release(struct
On Wed, 2017-02-15 at 20:25 +, Anoob Soman wrote:
> +static struct packet_fanout *fanout_release(struct sock *sk)
> {
> struct packet_sock *po = pkt_sk(sk);
> struct packet_fanout *f;
> @@ -1728,17 +1736,17 @@ static void fanout_release(struct sock *sk)
> if (f) {
>
Commit 6664498280cf ("packet: call fanout_release, while UNREGISTERING a
netdev"), unfortunately, introduced the following issues.
1. calling mutex_lock(&fanout_mutex) (fanout_release()) from inside
rcu_read-side critical section. rcu_read_lock disables preemption, most often,
which prohibits call
