RDNSS: support more ipv6 addresses

[WenChao1Hou]

RFC 6106 recommended that the number of RDNSS addresses thatshould be learned and maintained through the RDNSS RA option should be limited to three. RFC 8106 removes that recommendation; thus, the number of RDNSS addresses to maintain is determined by an implementer's local policy.The number of RDNSS addresses to maintain is determined by the user's own profile

Fix:#182

[robbat2]

Why did you pick 2048?

• 1280 bytes is the min IPv6 MTU; what would happen in that case?
• can radvd safely send multiple RA packets with different RDNSS entries? (other framework code for example splits large options into their own packets)

[WenChao1Hou]

in my opinion
1、The IPv6 MTU value may change the processing of the kernel protocol stack.
2、RDNSS 2048 is consistent with DNSSL packet maximum in radvd

Do you have any good suggestions for me?

[robbat2]

looks mostly good, some thoughts added; will wait for CI validation.

[robbat2]

Please amend the commit to include a Signed-Off-By trailer per DCO.

[robbat2]

Compile failure, wonder if you missed something?

send.c:620:31: error: use of undeclared identifier 'rdnssbuf'
                safe_buffer_append(sbl->sb, rdnssbuf->buffer, rdnssbuf->used);
                                            ^
send.c:620:49: error: use of undeclared identifier 'rdnssbuf'
                safe_buffer_append(sbl->sb, rdnssbuf->buffer, rdnssbuf->used);
                                                              ^
send.c:621:20: error: use of undeclared identifier 'rdnssbuf'
                safe_buffer_free(rdnssbuf);
                                 ^

[robbat2]

I've got some tests kicking around to add as well after you have it fixed.

[stappersg]

Please amend the commit to include a Signed-Off-By trailer per DCO.

@WenChao1Hou that is up-to-you. So when it is day again in your time zone, please add Signed-Off-By.

[WenChao1Hou]

Compile failure, wonder if you missed something?

send.c:620:31: error: use of undeclared identifier 'rdnssbuf'
                safe_buffer_append(sbl->sb, rdnssbuf->buffer, rdnssbuf->used);
                                            ^
send.c:620:49: error: use of undeclared identifier 'rdnssbuf'
                safe_buffer_append(sbl->sb, rdnssbuf->buffer, rdnssbuf->used);
                                                              ^
send.c:621:20: error: use of undeclared identifier 'rdnssbuf'
                safe_buffer_free(rdnssbuf);
                                 ^

Compile failure, wonder if you missed something?

send.c:620:31: error: use of undeclared identifier 'rdnssbuf'
                safe_buffer_append(sbl->sb, rdnssbuf->buffer, rdnssbuf->used);
                                            ^
send.c:620:49: error: use of undeclared identifier 'rdnssbuf'
                safe_buffer_append(sbl->sb, rdnssbuf->buffer, rdnssbuf->used);
                                                              ^
send.c:621:20: error: use of undeclared identifier 'rdnssbuf'
                safe_buffer_free(rdnssbuf);
                                 ^

it was an oversight on my part，The variable name was not updated in time when submitted, I will update MR as soon as possible

[the-j0k3r]

We had merged this patch downstream at this code state https://svn.dd-wrt.com/changeset/50745, I did not test it myself but clearly it caused some instability and was reverted

One such report https://forum.dd-wrt.com/phpBB2/viewtopic.php?p=1275031#1275031

Same thing happened to me. If you clean those two ipv6 dns lines (leave them empty) radvd deamon will not crash. Your/my /tmp/radvd.conf look/ed like:

root@tp:/tmp# cat radvd.conf
interface br0
{
IgnoreIfMissing on;
AdvSendAdvert on;
MinRtrAdvInterval 3;
MaxRtrAdvInterval 10;
AdvHomeAgentFlag off;
AdvManagedFlag off;
AdvOtherConfigFlag on;
AdvLinkMTU 1452;
prefix ::/64
{
AdvOnLink on;
AdvAutonomous on;
AdvValidLifetime 30;
AdvPreferredLifetime 20;
};
RDNSS 2606:4700:4700::1111 2606:4700:4700::1001{};
};

On 50786 once you leave those boxes empty you'll end up whith the same conf sans that RDNSS line and radvd won´t die.

Syslog entries of radvd deamon dying:

Nov 3 07:07:12 tp daemon.info radvd[9224]: version 2.19 started
Nov 3 07:07:12 tp daemon.err radvd[9226]: Exiting, privsep_read_loop had readn return 0 bytes
Nov 3 07:07:12 tp daemon.err radvd[9226]: Exiting, privsep_read_loop is complete.

Previous firm with that line working: 50485.

@wcbonner @BrainSlayer

[reubenhwk]

Just catching up here....why are more than 3 RDNS servers needed? I saw that more than three doesn't work, but I did NOT see any explanation as to why more than three are needed.

[wcbonner]

Just catching up here....why are more than 3 RDNS servers needed? I saw that more than three doesn't work, but I did NOT see any explanation as to why more than three are needed.

There isn't really a huge need for more than three, but it would be nice if the program didn't crash or error out if more are listed in the config file.

The problem I ran into was in a downstream project when I wanted to specify two dns servers and the program was already adding two servers it had received from another network broadcast or dhcp interaction.

radvd was silently exiting in the background and my local network wasn't getting advertisements.

[reubenhwk]

Ack. It's been a long time since I've work on this code... Nonetheless I have concerns that not having a hard limit would eventually lead to MTU issues. At the same time I'm not sure that exiting the program is also the right thing to do. Not sure. Maybe best effort is good enough, maybe not.