Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <60f78eaa-ace7-c27d-8e45-4777ecf3faa2@alu.unizg.hr>
Date:   Tue, 6 Jun 2023 15:57:35 +0200
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.11.0
Subject: Re: POSSIBLE BUG: selftests/net/fcnal-test.sh: [FAIL] in vrf "bind -
 ns-B IPv6 LLA" test
To:     Guillaume Nault <gnault@redhat.com>
Cc:     netdev@vger.kernel.org, "David S. Miller" <davem@davemloft.net>,
        Eric Dumazet <edumazet@google.com>,
        Jakub Kicinski <kuba@kernel.org>,
        Paolo Abeni <pabeni@redhat.com>, Shuah Khan <shuah@kernel.org>,
        linux-kernel@vger.kernel.org, linux-kselftest@vger.kernel.org
References: <b6191f90-ffca-dbca-7d06-88a9788def9c@alu.unizg.hr>
 <ZHeN3bg28pGFFjJN@debian> <a379796a-5cd6-caa7-d11d-5ffa7419b90e@alu.unizg.hr>
 <ZH84zGEODT97TEXG@debian>
Content-Language: en-US, hr
From:   Mirsad Todorovac <mirsad.todorovac@alu.unizg.hr>
In-Reply-To: <ZH84zGEODT97TEXG@debian>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk

On 6/6/23 15:46, Guillaume Nault wrote:
> On Tue, Jun 06, 2023 at 08:24:54AM +0200, Mirsad Goran Todorovac wrote:
>> On 5/31/23 20:11, Guillaume Nault wrote:
>>> I believe this condition should be relaxed to allow the case where
>>> ->sk_bound_dev_if is oif's master device (and maybe there are other
>>> VRF cases to also consider).
>>
>> I've tried something like this, but something makes the kernel stuck
>> here:
>>
>> TEST: ping out, blocked by route - ns-B loopback IPv6                         [ OK ]
>> TEST: ping out, device bind, blocked by route - ns-B loopback IPv6            [ OK ]
>> TEST: ping in, blocked by route - ns-A loopback IPv6                          [ OK ]
>> TEST: ping out, unreachable route - ns-B loopback IPv6                        [ OK ]
>> TEST: ping out, device bind, unreachable route - ns-B loopback IPv6           [ OK ]
>>
>> #################################################################
>> With VRF
>>
>> [hanged process and kernel won't shutdown]
>>
>> The code is:
>>
>> ---
>>   net/ipv6/ping.c | 12 +++++++++++-
>>   1 file changed, 11 insertions(+), 1 deletion(-)
>>
>> diff --git a/net/ipv6/ping.c b/net/ipv6/ping.c
>> index c4835dbdfcff..81293e902293 100644
>> --- a/net/ipv6/ping.c
>> +++ b/net/ipv6/ping.c
>> @@ -73,6 +73,9 @@ static int ping_v6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
>>          struct rt6_info *rt;
>>          struct pingfakehdr pfh;
>>          struct ipcm6_cookie ipc6;
>> +       struct net *net = sock_net(sk);
>> +       struct net_device *dev = NULL;
>> +       struct net_device *mdev = NULL;
>>          err = ping_common_sendmsg(AF_INET6, msg, len, &user_icmph,
>>                                    sizeof(user_icmph));
>> @@ -111,10 +114,17 @@ static int ping_v6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
>>          else if (!oif)
>>                  oif = np->ucast_oif;
>> +       if (oif) {
>> +               dev = dev_get_by_index(net, oif);
>> +               mdev = netdev_master_upper_dev_get(dev);
>> +       }
>> +
>>          addr_type = ipv6_addr_type(daddr);
>>          if ((__ipv6_addr_needs_scope_id(addr_type) && !oif) ||
>>              (addr_type & IPV6_ADDR_MAPPED) ||
>> -           (oif && sk->sk_bound_dev_if && oif != sk->sk_bound_dev_if))
>> +           (oif && sk->sk_bound_dev_if && oif != sk->sk_bound_dev_if &&
>> +                   !(mdev && sk->sk_bound_dev_if &&
>> +                             mdev != dev_get_by_index(net, sk->sk_bound_dev_if))))
>>                  return -EINVAL;
>>          ipcm6_init_sk(&ipc6, np);
>>
>> I am obviously doing something very stupid.
> 
> The problem is that dev_get_by_index() holds a reference on 'dev' which
> your code never releases. Also netdev_master_upper_dev_get() needs rtnl
> protection. These should have generated some kernel oops.
> 
> You can try this instead:
> 
> -------- >8 --------
> 
> diff --git a/net/ipv6/ping.c b/net/ipv6/ping.c
> index c4835dbdfcff..f804c11e2146 100644
> --- a/net/ipv6/ping.c
> +++ b/net/ipv6/ping.c
> @@ -114,7 +114,8 @@ static int ping_v6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
>   	addr_type = ipv6_addr_type(daddr);
>   	if ((__ipv6_addr_needs_scope_id(addr_type) && !oif) ||
>   	    (addr_type & IPV6_ADDR_MAPPED) ||
> -	    (oif && sk->sk_bound_dev_if && oif != sk->sk_bound_dev_if))
> +	    (oif && sk->sk_bound_dev_if && oif != sk->sk_bound_dev_if &&
> +	     l3mdev_master_ifindex_by_index(sock_net(sk), oif) != sk->sk_bound_dev_if))
>   		return -EINVAL;
>   
>   	ipcm6_init_sk(&ipc6, np);

Yes, I stupidly forgot that.

I came across a fixed version:

------
diff --git a/net/ipv6/ping.c b/net/ipv6/ping.c
index c4835dbdfcff..c1d81c49b775 100644
--- a/net/ipv6/ping.c
+++ b/net/ipv6/ping.c
@@ -73,6 +73,10 @@ static int ping_v6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
         struct rt6_info *rt;
         struct pingfakehdr pfh;
         struct ipcm6_cookie ipc6;
+       struct net *net = sock_net(sk);
+       struct net_device *dev = NULL;
+       struct net_device *mdev = NULL;
+       struct net_device *bdev = NULL;

         err = ping_common_sendmsg(AF_INET6, msg, len, &user_icmph,
                                   sizeof(user_icmph));
@@ -111,10 +115,26 @@ static int ping_v6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
         else if (!oif)
                 oif = np->ucast_oif;

+       if (oif) {
+               rcu_read_lock();
+               dev = dev_get_by_index_rcu(net, oif);
+               rcu_read_unlock();
+               rtnl_lock();
+               mdev = netdev_master_upper_dev_get(dev);
+               rtnl_unlock();
+       }
+
+       if (sk->sk_bound_dev_if) {
+               rcu_read_lock();
+               bdev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
+               rcu_read_unlock();
+       }
+
         addr_type = ipv6_addr_type(daddr);
         if ((__ipv6_addr_needs_scope_id(addr_type) && !oif) ||
             (addr_type & IPV6_ADDR_MAPPED) ||
-           (oif && sk->sk_bound_dev_if && oif != sk->sk_bound_dev_if))
+           (oif && sk->sk_bound_dev_if && oif != sk->sk_bound_dev_if &&
+                   !(mdev && sk->sk_bound_dev_if && bdev && mdev == bdev)))
                 return -EINVAL;

         ipcm6_init_sk(&ipc6, np);

However, this works by the test (888 passed) but your two liner is obviously
better :-)

Best regards,
Mirsad

-- 
Mirsad Goran Todorovac
Sistem inženjer
Grafički fakultet | Akademija likovnih umjetnosti
Sveučilište u Zagrebu

System engineer
Faculty of Graphic Arts | Academy of Fine Arts
University of Zagreb, Republic of Croatia