384.7 IPV6 Stateful Generates Logged Errors, Stateless Does Not

[GHammer]

Fixed with https://github.com/RMerl/asuswrt-merlin.ng/commit/ea82390ca534138fd2fb00015ab9d0a17a286ac5

Thank you.
Hopefully @john9527 or @RMerlin can compile a test load for this.

 

[RMerlin]

Have to wonder how many similar issues exist throughout the code on the aarch86 architecture... httpd is another service that seem prone to crash on the RT-AC86U.

 

[GHammer]

Have to wonder how many similar issues exist throughout the code on the aarch86 architecture... httpd is another service that seem prone to crash on the RT-AC86U.

I've only seen that once and restarting httpd from the console cured it.

 

[RMerlin]

I've only seen that once and restarting httpd from the console cured it.

Httpd is also covered by the watchdog, so it will restart within a minute.

 

[Sean B.]

It was a pointer! I knew it! Nice find @theMIROn .

 

[john9527]

Thank you.
Hopefully @john9527 or @RMerlin can compile a test load for this.

@theMIROn is also a great wizard!

I did a test build from master it you want to try it
RT-AC86U_384.7_test1-gea82390ca_cferom_ubi.w
https://1drv.ms/f/s!Ainhp1nBLzMJkEFTnFp4Sh2wvQva

 

[GHammer]

@theMIROn is also a great wizard!

I did a test build from master it you want to try it
RT-AC86U_384.7_test1-gea82390ca_cferom_ubi.w
https://1drv.ms/f/s!Ainhp1nBLzMJkEFTnFp4Sh2wvQva

Indeed he is!
Back in a few with the results.

 

[GHammer]

Applied the test load.
Didn't get the kernel fault! A big plus.
Looks like a good address and size:

dhcp6_client: bound address 2001:558:6017:1a4:34a7:8c78:d117:ee12/128, prefix 2601:19b:4800:2121::/64

Just one anomaly:

Aug 21 10:37:49 kernel: ^[[0;33;41m[ERROR pktrunner] runnerUcast_inet6addr_event,202: Could not rdpa_ucast_ipv6_host_address_table_find^[[0m

And the ever present

Aug 21 10:37:52 dnsmasq[1763]: warning: no upstream servers configured

 

[themiron]

Aug 21 10:37:52 dnsmasq[1763]: warning: no upstream servers configured

that's false positive, safe to ignore

 

[GHammer]

that's false positive, safe to ignore

Good to know and makes sense as it is present regardless of state.

 

[Grisu]

Httpd is also covered by the watchdog, so it will restart within a minute.

Had it today again on RT-AC68U that GUI crashed, waited some minutes, just nothing, SSH restart_httpd and GUI in same second refreshed the page.
Happened when I changed 2G from auto to legacy mode (sorry dont know if I have been connected via 2G or 5G when it happened).
It should count in a new widow from 0-100% to finish, but this has not been shown in case of crash.

 

[GHammer]

@theMIROn @john9527 @RMerlin

Thank you for the troubleshooting and quick fix!
No errors (or any entry at all) in the log since applying the test build.
All devices up and have DHCP or static addresses.
And, what I was trying to do is working.

 

[GHammer]

How were you able to assign static IPv6 addresses? My attempts were ultimately unsuccessful.

Not the thread for that. Create a different one.

 

[RMerlin]

Note that the dnsmasq author decided to tackle the issue from a different angle. We'll probably have a different patch to test in the near future.

 

[GHammer]

Note that the dnsmasq author decided to tackle the issue from a different angle. We'll probably have a different patch to test in the near future.

Happy to help in any way I can.

 

[HowIFix]

Latest Commits from dnsmasq: http://thekelleys.org.uk/gitweb/?p=dnsmasq.git;a=summary

• Properly deal with unaligned addresses in DHCPv6 packets. [master]
• Fix broken DNSSEC records in previous.
• Don't return NXDOMAIN to empty non-terminals.
• Add --dhcp-name-match
• Handle case of --auth-zone but no --auth-server.

 

[RMerlin]

Latest Commits from dnsmasq: http://thekelleys.org.uk/gitweb/?p=dnsmasq.git;a=summary

• Properly deal with unaligned addresses in DHCPv6 packets. [master]

That's only part of the story, we've been involved in email exchanges about it...

 

[john9527]

@GHammer

I don't even know if they are finished yet, but since I've been doing test builds I did a 'test2' build with the patch from the dnsmasq author replacing @theMIROn 's patch. Same download location.

7b885328c (HEAD -> test) dnsmasq: (upstream) properly deal with unaligned addresses in DHCPv6 packets
c3617d4d3 dnsmasq: (upstream) fix broken DNSSEC records in ef267f514
ef267f514 dnsmasq: (upstream) don't return NXDOMAIN to empty non-terminals
7223b63e5 Revert "dnsmasq: dhcpv6: fix unaligned access crash on aarch64"

 

[GHammer]

@GHammer

I don't even know if they are finished yet, but since I've been doing test builds I did a 'test2' build with the patch from the dnsmasq author replacing @theMIROn 's patch. Same download location.

7b885328c (HEAD -> test) dnsmasq: (upstream) properly deal with unaligned addresses in DHCPv6 packets
c3617d4d3 dnsmasq: (upstream) fix broken DNSSEC records in ef267f514
ef267f514 dnsmasq: (upstream) don't return NXDOMAIN to empty non-terminals
7223b63e5 Revert "dnsmasq: dhcpv6: fix unaligned access crash on aarch64"

I'll load this and be back in a few.

 

[GHammer]

@john9527 I loaded test2 and rebooted. The router shutdown/powered off at that point.
When I cycled power and had access again, I don't see any errors related to this topic.
I do see these, but no noticeable problems with any devices or inbound access.

May  5 01:05:10 lldpd[925]: cannot get ethtool link information with GLINKSETTINGS (requires 4.9+): Operation not permitted
May  5 01:05:11 WAN_Connection: ISP's DHCP did not function properly.
May  5 01:05:12 rc_service: udhcpc 1007:notify_rc start_firewall
May  5 01:05:12 wan: finish adding multi routes
May  5 01:05:12 odhcp6c[1009]: Failed to send DHCPV6 message to ff02::1:2 (Cannot assign requested address)