NM fails with 'config-failed' despite successful MM connection

[Bjørn Mork]

Bjørn Mork <bjorn at mork.no> writes:

> I don't know where it got '28' from.  Maybe cached from an earlier
> instance of the same device?  This is a laptop which I often suspend,
> and the modem goes completely away and is rediscovered as a completely
> new USB device on every such event.

Yes, definitely a cache issue.  I don't understand why I have started to
hit it only recently. I am mostly tracking Debian sid on this laptop,
which means that NM is at v1.4.4.

Looking at the code, I got the impression that the ifindex cache should
be updated by udev events in this case, since the wwan0 interface is
removed and then re-added.

Running with 'trace' shows that the new device with ifindex 38 is added
on resume:

Jan  5 20:46:31 miraculix NetworkManager[8315]: <trace> [1483645591.1354] platform-linux: udev-add[wwan0,38]: device added

But attempting to connect after this still fails, due to using the
non-existing 37 ifindex:

Jan  5 20:47:00 miraculix NetworkManager[8315]: <debug> [1483645620.1126] device[0x563f669f5600] (cdc-wdm0): bringing up device
Jan  5 20:47:00 miraculix NetworkManager[8315]: <debug> [1483645620.1127] platform: link: setting up '(null)' (37)
Jan  5 20:47:00 miraculix NetworkManager[8315]: <debug> [1483645620.1127] platform-linux: link: change 37: flags: set 0x1/0x1 ([up] / [up])
Jan  5 20:47:00 miraculix NetworkManager[8315]: <debug> [1483645620.1128] platform-linux: do-request-link: 37 
Jan  5 20:47:00 miraculix NetworkManager[8315]: <debug> [1483645620.1129] platform-linux: netlink: recvmsg: error message from kernel: No such device (19) for request 2900
Jan  5 20:47:00 miraculix NetworkManager[8315]: <debug> [1483645620.1130] platform-linux: netlink: recvmsg: error message from kernel: No such device (19) for request 2901
Jan  5 20:47:00 miraculix NetworkManager[8315]: <debug> [1483645620.1130] platform-linux: do-change-link[37]: failure changing link: failure 19 (No such device)
Jan  5 20:47:00 miraculix NetworkManager[8315]: <info>  [1483645620.1131] device (cdc-wdm0): state change: config -> failed (reason 'config-failed') [50 120 4]


Any ideas?  To me it looks like

 cache_update_link_udev (platform, ifindex, udev_device);

should be called immediately after the 'udev-add' line is logged. Why
didn't it update the cache?


Bjørn