RDNSS: support more ipv6 addresses #193

WenChao1Hou · 2022-10-30T13:30:28Z

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 · 2022-10-30T15:34:19Z

send.c

@@ -594,6 +596,12 @@ static struct safe_buffer_list *add_ra_options_rdnss(struct safe_buffer_list *sb

 		memset(&rdnssinfo, 0, sizeof(rdnssinfo));

+		size_t const bytes = sizeof(rdnssinfo) + sizeof(struct in6_addr) * rdnss->AdvRDNSSNumber;
+		if (bytes > (256 * 8)) {


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)

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 · 2022-10-30T15:37:18Z

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

robbat2 · 2022-10-30T15:37:53Z

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 · 2022-10-30T16:07:23Z

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

More testing, expected to fail until PR#193 is merged. Reference: radvd-project#193 Signed-off-by: Robin H. Johnson <[email protected]>

stappersg · 2022-10-30T18:07:29Z

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 · 2022-10-31T01:59:45Z

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

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 Signed-off-by: WenChao1Hou <[email protected]>

the-j0k3r · 2022-11-05T13:07:27Z

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 · 2022-11-07T16:55:51Z

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 · 2022-11-07T17:03:29Z

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 · 2022-11-07T18:49:50Z

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.

Neustradamus · 2024-12-06T06:03:50Z

Dear all,

All people are welcome to comment the current situation, we are in December 2024, a perfect patch is needed.

It is really important to respect the RFC8106: IPv6 Router Advertisement Options for DNS Configuration:

https://datatracker.ietf.org/doc/html/rfc8106

@agowa, @AdSchellevis, @fichtner, @marjohn56, @bmeirellesRJ, @mhanor, @stu-gott, @wcbonner, @the-j0k3r, @WenChao1Hou, @BrainSlayer, @Mile-Lile, @bnutzer, @stu-gott, @MarioVerbelen, @ruben-herold, @ipilcher, @pyk1998, @romanrm, @corvus1, @ihipop

If you know some IPv6 specialists, please come here, there are several tickets, thanks in advance.

Neustradamus · 2024-12-08T05:46:18Z

@robbat2, @reubenhwk, @stappersg, @RICCIARDI-Adrien, @y-kzm, @fichtner, @melroy89, @xdbob, @audreylace, @XWwalker, @initramfs, @thesamesam, @jetomit, @mpontillo, @kepstin, @oskar456, @zajdee, @yas-nyan, @gaoxingwang, @juhamatk, @manonfgoo, @andrew-sayers, @jpds, @steglicd, @samboyles1, @nomis, @haegar, @ezeeyahoo, @Testsr, @landgraf, @hxx-fetch, @prometheanfire, @tlsalmin, @ffontaine, @BrainSlayer, @Chocobo1, @sshambar, @bluca, @dhallas, @florianl, @fooker, @trofi, @l29ah, @thediveo, @smlng, @candrews, @alexaring, @steglicd, @pavlix, @Erkan-Yilmaz, @abrodkin, all others: What do you think?

All people are welcome to comment, help...

If you know other IPv6 experts, specialists, it will be nice too.

This PR needs to be followed by this @robbat2 PR:

test: add more RDNSS testing for PR#193 #195

Linked to the next 2.20 release build:

time to tag/release 2.20? #185

Some other tickets can be solved too:

time to tag/release 2.20? #185 (comment)

Thanks in advance, we need to have a good 2.20 since 2.19 (2020-09-23).

Neustradamus · 2024-12-08T05:47:04Z

@crrodriguez, @neocturne, @danrl, @vapier, @GheRivero, @segoon, @moonlinux, @sean797, @ferdinandoterada, @fgont, @nachtgeist, all others: What do you think?

danrl · 2024-12-08T14:59:55Z

As a compromise that should work for most reasonable cases I can imagine right now radvd could behave the following way:

Exit at startup time when a config asks for more RDNSS than would fit inside a RA limited by minimum MTU 1280 bytes. This should be at config time for static setups. Strictly speaking the RFC would require additional RAs to be send to fit an excessive number of RDNSS but all RA daemons I have seen put limits in, even my own ratools suite which was written for academic purpose of following the RFC absurdly strictly.
Continue operating in other cases of excess, but log an error periodically, maybe in RA announcement intervals.

One could argue that randomizing the list of excessive RDNSS each time a RA is sent and then picking as many from the top as fit within 1280 bytes would eventually allow all RDNSS to he announced at least once. However, this makes RADVD unpredictable in behavior which is bad. Predictable is important for reliable testing and debugging. It is more important than following the letter of the RFC. Well tested, reasonable behavior is better than covering all use cases.

My suggestion for moving forward: Limit RDNSS to a reasonable number. 2048 is too many. We can barely fit 70 in there even if we could use the entire packet minus header for 16 bytes addresses. Now how many RDNSS can we fit in for a reasonable scenario? Maybe one where RDNSS is the only option and we don’t even have a Prefix option. That is the most somewhat reasonable scenario I can think of. The limit should be at max this number. The math shall be provided in the commit message for future archeology.

HTH

melroy89 · 2024-12-08T19:01:20Z

I would expect a test in: https://github.com/radvd-project/radvd/tree/master/test (which is normally part of the same delivery)

Reference: radvd-project#193 Signed-off-by: Robin H. Johnson <[email protected]>

This is based on PR#193, with significant cleanups & safety checks added. - RFC 6106 section 5.3.1 recommended that the number of RDNSS addresses should be limited to three, however that was for clients, not servers. - RFC 8106 section 5.3.1 clarified the recommendation to at least 3 from multiple sources. Based on this, set the limit for a single RDNSS section to 127 servers in radvd, based on the size of the length field. Note that it may cause fragmentation as the single option exceeds usable RA option space (MTU less headers); and server configurations should generally use fewer. See also the RFC6106 errata & RFC6980 section 2 for behaviors of very large RA options. Reference: https://www.rfc-editor.org/errata/rfc6106 Reference: https://datatracker.ietf.org/doc/html/rfc6980#section-2 Reference: https://datatracker.ietf.org/doc/html/rfc6106#section-5.3.1 Reference: https://datatracker.ietf.org/doc/html/rfc8106#section-5.3.1 Based-on: radvd-project#193 Based-on-work-by: WenChao1Hou <[email protected]> Signed-off-by: Robin H. Johnson <[email protected]>

robbat2 · 2024-12-28T21:23:27Z

Replaced by #245

This is based on PR#193, with significant cleanups & safety checks added. - RFC 6106 section 5.3.1 recommended that the number of RDNSS addresses should be limited to three, however that was for clients, not servers. - RFC 8106 section 5.3.1 clarified the recommendation to at least 3 from multiple sources. Based on this, set the limit for a single RDNSS section to 127 servers in radvd, based on the size of the length field. Note that it may cause fragmentation as the single option exceeds usable RA option space (MTU less headers); and server configurations should generally use fewer. See also the RFC6106 errata & RFC6980 section 2 for behaviors of very large RA options. Reference: https://www.rfc-editor.org/errata/rfc6106 Reference: https://datatracker.ietf.org/doc/html/rfc6980#section-2 Reference: https://datatracker.ietf.org/doc/html/rfc6106#section-5.3.1 Reference: https://datatracker.ietf.org/doc/html/rfc8106#section-5.3.1 Based-on: radvd-project#193 Based-on-work-by: WenChao1Hou <[email protected]> Signed-off-by: Robin H. Johnson <[email protected]>

WenChao1Hou mentioned this pull request Oct 30, 2022

radvd exits on too many addresses in RDNSS section #182

Closed

robbat2 reviewed Oct 30, 2022

View reviewed changes

robbat2 approved these changes Oct 30, 2022

View reviewed changes

robbat2 changed the title ~~RDNSS section support more ipv6 aaddrss~~ RDNSS: support more ipv6 addresses Oct 30, 2022

robbat2 added the rdnss label Oct 30, 2022

robbat2 added a commit to robbat2/radvd that referenced this pull request Oct 30, 2022

test: add more RDNSS testing for PR#193

91dc733

More testing, expected to fail until PR#193 is merged. Reference: radvd-project#193 Signed-off-by: Robin H. Johnson <[email protected]>

robbat2 mentioned this pull request Oct 30, 2022

test: add more RDNSS testing for PR#193 #195

Merged

WenChao1Hou force-pushed the master branch from 746ec1e to 6ac66c1 Compare October 31, 2022 18:11

WenChao1Hou requested a review from robbat2 November 1, 2022 01:24

Merge branch 'radvd-project:master' into master

723908f

Neustradamus assigned robbat2 May 3, 2024

Neustradamus mentioned this pull request Nov 22, 2024

time to tag/release 2.20? #185

Closed

This was referenced Dec 7, 2024

Change the Prefix field of the Route Information Option to be variable-length. #234

Merged

implement setup_allrouters_membership() et al for FreeBSD #145 #241

Merged

stappersg mentioned this pull request Dec 9, 2024

remove unused 'total_seen_options' variable #242

Merged

robbat2 added a commit to robbat2/radvd that referenced this pull request Dec 28, 2024

test: add more RDNSS testing for PR#193

7b4da96

Reference: radvd-project#193 Signed-off-by: Robin H. Johnson <[email protected]>

robbat2 mentioned this pull request Dec 28, 2024

feat: Support more addresses in RDNSS section #245

Merged

robbat2 closed this Dec 28, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

RDNSS: support more ipv6 addresses #193

RDNSS: support more ipv6 addresses #193

WenChao1Hou commented Oct 30, 2022 •

edited

Loading

robbat2 Oct 30, 2022

WenChao1Hou Oct 31, 2022

robbat2 left a comment

robbat2 commented Oct 30, 2022

robbat2 commented Oct 30, 2022

robbat2 commented Oct 30, 2022

stappersg commented Oct 30, 2022

WenChao1Hou commented Oct 31, 2022

the-j0k3r commented Nov 5, 2022 •

edited

Loading

reubenhwk commented Nov 7, 2022

wcbonner commented Nov 7, 2022

reubenhwk commented Nov 7, 2022

Neustradamus commented Dec 6, 2024

Neustradamus commented Dec 8, 2024 •

edited

Loading

Neustradamus commented Dec 8, 2024

danrl commented Dec 8, 2024

melroy89 commented Dec 8, 2024 •

edited

Loading

robbat2 commented Dec 28, 2024

RDNSS: support more ipv6 addresses #193

RDNSS: support more ipv6 addresses #193

Conversation

WenChao1Hou commented Oct 30, 2022 • edited Loading

robbat2 Oct 30, 2022

Choose a reason for hiding this comment

WenChao1Hou Oct 31, 2022

Choose a reason for hiding this comment

robbat2 left a comment

Choose a reason for hiding this comment

robbat2 commented Oct 30, 2022

robbat2 commented Oct 30, 2022

robbat2 commented Oct 30, 2022

stappersg commented Oct 30, 2022

WenChao1Hou commented Oct 31, 2022

the-j0k3r commented Nov 5, 2022 • edited Loading

reubenhwk commented Nov 7, 2022

wcbonner commented Nov 7, 2022

reubenhwk commented Nov 7, 2022

Neustradamus commented Dec 6, 2024

Neustradamus commented Dec 8, 2024 • edited Loading

Neustradamus commented Dec 8, 2024

danrl commented Dec 8, 2024

melroy89 commented Dec 8, 2024 • edited Loading

robbat2 commented Dec 28, 2024

WenChao1Hou commented Oct 30, 2022 •

edited

Loading

the-j0k3r commented Nov 5, 2022 •

edited

Loading

Neustradamus commented Dec 8, 2024 •

edited

Loading

melroy89 commented Dec 8, 2024 •

edited

Loading