net/ipv6: Fix gfp_flags arg to addrconf_prefix_route

Eric noticed that __ipv6_ifa_notify is called under rcu_read_lock, so
the gfp argument to addrconf_prefix_route can not be GFP_KERNEL.

While scrubbing other calls I noticed addrconf_addr_gen has one
place with GFP_ATOMIC that can be GFP_KERNEL.

Fixes: acb54e3c ("net/ipv6: Add gfp_flags to route add functions")
Reported-by: syzbot+2add39b05179b31f912f@syzkaller.appspotmail.com
Reported-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid Ahern <dsahern@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

net/ipv6/addrconf.c

@@ -3245,7 +3245,7 @@ static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
 			addrconf_add_linklocal(idev, &addr, 0);
 		else if (prefix_route)
 			addrconf_prefix_route(&addr, 64, idev->dev,
-					      0, 0, GFP_ATOMIC);
+					      0, 0, GFP_KERNEL);
 		break;
 	case IN6_ADDR_GEN_MODE_NONE:
 	default:
@@ -5620,7 +5620,7 @@ static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
 		if (!ipv6_addr_any(&ifp->peer_addr))
 			addrconf_prefix_route(&ifp->peer_addr, 128,
 					      ifp->idev->dev, 0, 0,
-					      GFP_KERNEL);
+					      GFP_ATOMIC);
 		break;
 	case RTM_DELADDR:
 		if (ifp->idev->cnf.forwarding)