FS#2848 - multicast ff02::2 not responding since 19.07

[openwrt-bot]

guifipedro:

cannot do multicast in nanostation m5 xw (and maybe applies for other devices with eth0.1 and eth0.2 simulating eth0 and eth1)

multicast is a core feature for running some routing protocols as bmx6 per cable)

iface="my ethernet"
ping6 ff02::2%$iface

did tests with several platforms

• 18.06.4 ar71xx custom firmware -> works
• 18.06.7 ar71xx openwrt.org -> works
• 19.07.1 ar71xx custom firmware -> does not work
• 19.07.1 ar71xx openwrt.org -> does not work
• 19.07.1 ath79 openwrt.org -> does not work
• snapshot ar71xx openwrt.org -> does not work
  • 0a21d2eb80787083852d92f1519f268df4873c07873ea6e4c1ba17af791fd801 3904.0 KB Mon Jun 17 14:08:48 2019
• snapshot ath79 openwrt.org -> does not work
  • e6831b128017c83082110191a4cbdf37994beb7e2dae651b5b36b854b4674805 4096.3 KB Thu Feb 20 09:43:23 2020
  • OpenWrt SNAPSHOT, r12252-060b58fd6e

when I was figuring out what was happening, I did bmx6 routing (multicast) through br-lan in 3 devices, and I looked results with tcpdump -i br-lan. From the other devices I see that the multicast traffic from the affected device arrives (multicast packets go out), but when I see tcpdump -i br-lan in the affected device I only see his own multicast traffic (it's not entering anything else)

hence, looks like 19.07.1 introduced something that it does not allow this

[openwrt-bot]

guifipedro:

I was pending to try ping ff02::2%eth0 in different devices:

in nexx wt3020 8MB with OpenWrt 19.07.2 r10947-65030d81f3 -> works

in TP-Link TL-WDR4300 v1 with OpenWrt 19.07.1 r10911-c155900f66 -> works

(this two devices have eth0.1 and eth0.2 switches)

Hence, the problem is only focused in nanostation-m-xw

Tried new versions:

• nanostation-m-xw:
  • with ar71xx and 19.07.2 -> does not work
  • with ath79 and 19.07.2 -> does not work
  • with ar71xx and 18.06.8 -> works

help

[openwrt-bot]

rogerpueyo:

Hi,

So I made some tests and narrowed down the range of commits where the one introducing the bug is:

Last known working commit:

• r9528 [[https://git.openwrt.org/?p=openwrt/openwrt.git;a=commit;h=1cfbf95393decf047900e1b66b076a030704db20|1cfbf95393decf047900e1b66b076a030704db20 - kernel: bump 4.14 to 4.14.104]]

First known not working commit:

• r9891 [[https://git.openwrt.org/?p=openwrt/openwrt.git;a=commit;h=da5bd73d702e7cb9f7da174977b606ca12ffd96b|da5bd73d702e7cb9f7da174977b606ca12ffd96b - kernel: bump 4.14 to 4.14.112]]

[openwrt-bot]

rogerpueyo:

Hi,

I could find the precise commit where IPv6 multicast stops working.

First, there is commit [[https://git.openwrt.org/?p=openwrt/openwrt.git;a=commit;h=d3506d1bc160e4ce2833dff357f2a4b219772c77|d3506d1bc160e4ce2833dff357f2a4b219772c77 - ar71xx: ag71xx: fix compile error when enabling debug]] that broke the whole network for the device. It did not actually affect multicast, though.

Then came a series of commits related to AR8216. In particular, with commit [[https://git.openwrt.org/?p=openwrt/openwrt.git;a=commit;h=d6366ce3665f010a7ae7061a557689643073040a|r9699 d6366ce - generic: ar8216: mib_work_func: read all port mibs everytime]] IPv6 multicast would still work, but the network is broken. Therefore:

• [[https://git.openwrt.org/?p=openwrt/openwrt.git;a=commit;h=d6366ce3665f010a7ae7061a557689643073040a|d6366ce366]] + revert [[https://git.openwrt.org/?p=openwrt/openwrt.git;a=commit;h=d3506d1bc160e4ce2833dff357f2a4b219772c77|d3506d1bc1]]: still works (network + IPv6 multicast)

Next commit is [[https://git.openwrt.org/?p=openwrt/openwrt.git;a=commit;h=c8c2ef1d495dd3fd3096ac508e91a02f9c583ea8|r9700 c8c2ef1 - generic: ar8216: do a software reset for switch during hw_init]] and this one makes IPv6 multicast stop working. Plus, network is still broken from the debug-related commit. So:

• [[https://git.openwrt.org/?p=openwrt/openwrt.git;a=commit;h=c8c2ef1d495dd3fd3096ac508e91a02f9c583ea8|c8c2ef1d49]] + revert [[https://git.openwrt.org/?p=openwrt/openwrt.git;a=commit;h=d3506d1bc160e4ce2833dff357f2a4b219772c77|d3506d1bc1]]: IPv6 multicast does not work anymore

Finally, there's [[https://git.openwrt.org/?p=openwrt/openwrt.git;a=commit;h=31ea08a64bd977e0dc0a65804e4103cb52d81cea|31ea08a64bd977e0dc0a65804e4103cb52d81cea - ar71xx: ag71xx: Fix broken networking on some devices (FS#2177)]] restoring the broken network from d3506d1, but IPv6 multicast will not work anymore.

Summary

So, it seems that [[https://git.openwrt.org/?p=openwrt/openwrt.git;a=commit;h=c8c2ef1d495dd3fd3096ac508e91a02f9c583ea8|c8c2ef1d495dd3fd3096ac508e91a02f9c583ea8 - generic: ar8216: do a software reset for switch during hw_init]] introduced this IPv6 multicast bug, and it happened after [[https://git.openwrt.org/?p=openwrt/openwrt.git;a=commit;h=d3506d1bc160e4ce2833dff357f2a4b219772c77|d3506d1bc160e4ce2833dff357f2a4b219772c77 - ar71xx: ag71xx: fix compile error when enabling debug]] broke the device's network, but unrelated.

[openwrt-bot]

rogerpueyo:

More tests:

• ar71xx: current master + revert c8c2ef1: IPv6 multicast works
• ath79: current master + revert c8c2ef1: IPv6 multicast works