[ofa-general] Re: IPoIB kernel Oops -- possible race condition identified.

[Roland Dreier]

 > -	priv->mcast_mtu = IPOIB_UD_MTU(ib_mtu_enum_to_int(priv->broadcast->mcmember.mtu));
 > +	spin_lock_irq(&priv->lock);
 > +	if (priv->broadcast)
 > +		priv->mcast_mtu = IPOIB_UD_MTU(ib_mtu_enum_to_int(priv->broadcast->mcmember.mtu));
 > +	else
 > +		priv->mcast_mtu = priv->admin_mtu;

Looks reasonable, except what about the code shortly before that in
ipoib_mcast_join_task():

	if (!test_bit(IPOIB_MCAST_FLAG_ATTACHED, &priv->broadcast->flags)) {
		if (!test_bit(IPOIB_MCAST_FLAG_BUSY, &priv->broadcast->flags))
			ipoib_mcast_join(dev, priv->broadcast, 0);
		return;
	}

is there any reason why that is safe without locking around the using
priv->broadcast?  (Fixing that place looks harder, since
ipoib_mcast_join() currently may sleep)

 - R.