netdev - Re: VRF NS for lladdr sent on the wrong interface

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Tue, 1 Dec 2020 18:06:53 -0700
From:   David Ahern <dsahern@...il.com>
To:     Stephen Suryaputra <ssuryaextr@...il.com>
Cc:     netdev@...r.kernel.org
Subject: Re: VRF NS for lladdr sent on the wrong interface

On 12/1/20 12:00 PM, Stephen Suryaputra wrote:
> On Mon, Nov 30, 2020 at 06:15:06PM -0700, David Ahern wrote:
>> On 11/23/20 5:23 PM, Stephen Suryaputra wrote:
>>> Hi,
>>>
>>> I'm running into a problem with lladdr pinging all-host mcast all nodes
>>> addr. The ping intially works but after cycling the interface that
>>> receives the ping, the echo request packet causes a neigh solicitation
>>> being sent on a different interface.
>>>
>>> To repro, I included the attached namespace scripts. This is the
>>> topology and an output of my test.
>>>
>>> # +-------+     +----------+   +-------+
>>> # | h0    |     |    r0    |   |    h1 |
>>> # |    v00+-----+v00    v01+---+v10    |
>>> # |       |     |          |   |       |
>>> # +-------+     +----------+   +-------+
>>>
>>
>>
>>
>> after setup,
>>
>> ip netns exec h0 ping -c 1 ff02::1%h0_v00
>>
>> works, but
>>
>>  ip netns exec h1 ping -c 1 ff02::1%h1_v10
>>
>> does not. No surprise then that cycling v00 in r0 causes the reverse.
>> The problem is the route order changes:
>>
>> root@...ntu-c-2-4gib-sfo3-01:~# diff -U3 /tmp/1 /tmp/2
>> --- /tmp/1	2020-12-01 01:07:39.795361392 +0000
>> +++ /tmp/2	2020-12-01 01:07:51.991808848 +0000
>> @@ -1,6 +1,6 @@
>>  local fe80::8466:b3ff:fecc:3a4f dev r0_v01 table 10 proto kernel metric
>> 0 pref medium
>>  local fe80::b4ec:a8ff:fec3:33d9 dev r0_v00 table 10 proto kernel metric
>> 0 pref medium
>> -fe80::/64 dev r0_v00 table 10 proto kernel metric 256 pref medium
>>  fe80::/64 dev r0_v01 table 10 proto kernel metric 256 pref medium
>> -ff00::/8 dev r0_v00 table 10 metric 256 pref medium
>> +fe80::/64 dev r0_v00 table 10 proto kernel metric 256 pref medium
>>  ff00::/8 dev r0_v01 table 10 metric 256 pref medium
>> +ff00::/8 dev r0_v00 table 10 metric 256 pref medium
>>
>> With your patch does ping from both hosts work?
> 
> Yes, it does.
> 
>> What about all of the tests in
>> tools/testing/selftests/net/fcnal-test.sh? specifically curious about
>> the 'LLA to GUA' tests (link local to global). Perhaps those tests need
>> a second interface (e.g., a dummy) that is brought up first to cause the
>> ordering to be different.
> 
> The script needs nettest to be in the path...
> 

nettest is in the same directory. Build it and then run the script -
with your patch applied. We need to see if it affects existing tests.