[PATCH net] selftests/wireguard: make sure rp_filter disabled on vethc

[PATCH net] selftests/wireguard: make sure rp_filter disabled on vethc

[Hangbin Liu]

Some distros may enable strict rp_filter by default, which will previent
vethc receive the packets with unrouteable reverse path address.

Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
---
 tools/testing/selftests/wireguard/netns.sh | 1 +
 1 file changed, 1 insertion(+)

diff --git a/tools/testing/selftests/wireguard/netns.sh b/tools/testing/selftests/wireguard/netns.sh
index 7ed7cd95e58f..37b12f642254 100755
--- a/tools/testing/selftests/wireguard/netns.sh
+++ b/tools/testing/selftests/wireguard/netns.sh
@@ -363,6 +363,7 @@ ip1 -6 rule add table main suppress_prefixlength 0
 ip1 -4 route add default dev wg0 table 51820
 ip1 -4 rule add not fwmark 51820 table 51820
 ip1 -4 rule add table main suppress_prefixlength 0
+n1 sysctl -w net.ipv4.conf.vethc.rp_filter=0
 # Flood the pings instead of sending just one, to trigger routing table reference counting bugs.
 n1 ping -W 1 -c 100 -f 192.168.99.7
 n1 ping -W 1 -c 100 -f abab::1111
-- 
2.26.3

Re: [PATCH net] selftests/wireguard: make sure rp_filter disabled on vethc

[Jason A. Donenfeld]

Hi Hangbin,

Thanks. I've queued this up in my tree and will send it out on the
next wireguard push.

Jason

Re: [PATCH net] selftests/wireguard: make sure rp_filter disabled on vethc

[Hangbin Liu]

On Tue, May 25, 2021 at 8:19 PM Jason A. Donenfeld <Jason@zx2c4.com> wrote:
>
> Hi Hangbin,
>
> Thanks. I've queued this up in my tree and will send it out on the
> next wireguard push.

Hi Jason,

I have a question about the wg-quick(8)-style policy routing test.

ip1 -6 addr add fc00::9/96 dev vethc
ip1 -6 route add default via fc00::1
                                             ^^ Here why we add a
default route via none exist address?
ip2 -4 addr add 192.168.99.7/32 dev wg0
ip2 -6 addr add abab::1111/128 dev wg0

Thanks
Hangbin

Re: [PATCH net] selftests/wireguard: make sure rp_filter disabled on vethc

[Jason A. Donenfeld]

Hi Hangbin,

fc00::1 lives inside of fc00::9/96.

Thanks,
Jason

Re: [PATCH net] selftests/wireguard: make sure rp_filter disabled on vethc

[Hangbin Liu]

On Wed, May 26, 2021 at 04:38:09PM +0200, Jason A. Donenfeld wrote:
> Hi Hangbin,
> 
> fc00::1 lives inside of fc00::9/96.

Yes, I know fc00::1 belongs to fc00::9/96 subnet. I just don't understand
why we need to add a default v6 route for vethc, while the default route
address fc00::1 is not configured on any device.

This step looks useless for me.

Thanks
Hangbin

Re: [PATCH net] selftests/wireguard: make sure rp_filter disabled on vethc

[Jason A. Donenfeld]

On Tue, May 25, 2021 at 2:15 PM Hangbin Liu <liuhangbin@gmail.com> wrote:
>
> Some distros may enable strict rp_filter by default, which will previent
> vethc receive the packets with unrouteable reverse path address.
>
> Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
> ---
>  tools/testing/selftests/wireguard/netns.sh | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/tools/testing/selftests/wireguard/netns.sh b/tools/testing/selftests/wireguard/netns.sh
> index 7ed7cd95e58f..37b12f642254 100755
> --- a/tools/testing/selftests/wireguard/netns.sh
> +++ b/tools/testing/selftests/wireguard/netns.sh
> @@ -363,6 +363,7 @@ ip1 -6 rule add table main suppress_prefixlength 0
>  ip1 -4 route add default dev wg0 table 51820
>  ip1 -4 rule add not fwmark 51820 table 51820
>  ip1 -4 rule add table main suppress_prefixlength 0
> +n1 sysctl -w net.ipv4.conf.vethc.rp_filter=0

The VM does not ship with sysctl, and you'll notice that other changes
go through /proc directly. Since it's a trivial change, I'll rewrite
your commit.

Jason

Re: [PATCH net] selftests/wireguard: make sure rp_filter disabled on vethc

[Hangbin Liu]

On Mon, May 31, 2021 at 8:39 PM Jason A. Donenfeld <Jason@zx2c4.com> wrote:
> > +n1 sysctl -w net.ipv4.conf.vethc.rp_filter=0
>
> The VM does not ship with sysctl, and you'll notice that other changes
> go through /proc directly. Since it's a trivial change, I'll rewrite
> your commit.

Sorry to make this trouble. I saw other selftests also using sysctl, so I didn't
realize some VM may not have sysctl installed.

Thanks for your fix.

Hangbin