If a client disconnected all its interfaces and reconnects when the come back, it will likely start from any ip address returned dns, which means it can try to connect to a different ctdb node. The old node may not noticed the disconnect and still holds the client_guid based smbd. Up to know the new node returned NT_STATUS_NOT_SUPPORTED to the SMB2 Negotiate request, as messaging_send_iov[_from]() will return -1/ENOSYS if a file descriptor os passed to a process on a different node.
This bug was referenced in samba master: 3fd18a0d5b77a9f78c595852c342d4c8c33fac61 0efcfaa49c3d61f2c8116ebafd55b72d3277d0d8 21ef01e7b8368caa050ed82b9d787d1679220b2b 8591d9424371e173b079d5c8a267ea4c2cb266ad
Created attachment 17509 [details] Patches for v4-17-test
Created attachment 17510 [details] Patches for v4-16-test
Created attachment 17511 [details] Patches for v4-15-test
Reassigning to Jule for inclusion in 4.15, 4.16 and 4.17.
Pushed to autobuild-v4-{17,16,15}-test.
This bug was referenced in samba v4-17-test: fc52fe99d7900d5483c94ad49a0095ddd283354f 64daf27dc730cd432dc902de5b93c3097da335f2 095ee4ce189f12e13cb9d70bd99a670623aa08ae 71c94a076bad166098e1c24e45f866abdd6c8f92
This bug was referenced in samba v4-17-stable (Release samba-4.17.0rc5): fc52fe99d7900d5483c94ad49a0095ddd283354f 64daf27dc730cd432dc902de5b93c3097da335f2 095ee4ce189f12e13cb9d70bd99a670623aa08ae 71c94a076bad166098e1c24e45f866abdd6c8f92
Autobuild failed several times for task samba-ctdb during test.
(In reply to Jule Anger from comment #9) Ah, we can't backport the tests as they rely on the new smbstatus --json
Created attachment 17548 [details] Patches for v4-16-test Backport to 4.16 without the tests
Reassiging to Jule for inclusion in 4.16.
Reassigning to metze due a possible regression: https://lists.samba.org/archive/samba/2022-September/241887.html
Metze, I'll re-review once the possible regression is investigated. Sorry, COVID still slowing me down at the moment :-(.
Comment on attachment 17548 [details] Patches for v4-16-test The additional problem is handled in https://bugzilla.samba.org/show_bug.cgi?id=15200, so it's not a regression...
Reassigning again to Jule for inclusion in 4.16.
Pushed to autobuild-v4-16-test
This bug was referenced in samba v4-16-test: ada5ef9d847f17fd00fd1704300258de97c69e6f 52dd57d4b30218b93231cf1a990c8829b0c06033
Closing out bug report. Thanks!
This bug was referenced in samba v4-16-stable (Release samba-4.16.8): ada5ef9d847f17fd00fd1704300258de97c69e6f 52dd57d4b30218b93231cf1a990c8829b0c06033