If global logging is set to a level higher than 1 (see example below), segmentation faults will follow while serving requests, and the requests will not get served. Configuration example of smb.conf: ... [global] log level = 2 ... Example of one such error being logged: smb | 2021-05-26T21:44:10.089721974Z =============================================================== smb | 2021-05-26T21:44:10.089765622Z INTERNAL ERROR: Signal 11: Segmentation fault in pid 2766 (4.13.8) smb | 2021-05-26T21:44:10.089785955Z If you are running a recent Samba version, and if you think this problem is not yet fixed in the latest versions, please consider reporting this bug, see https://wiki.samba.org/index.php/Bug_Reporting smb | 2021-05-26T21:44:10.089806547Z =============================================================== smb | 2021-05-26T21:44:10.089825991Z PANIC (pid 2766): Signal 11: Segmentation fault in 4.13.8 smb | 2021-05-26T21:44:10.089844639Z unable to produce a stack trace on this platform smb | 2021-05-26T21:44:10.089864194Z dumping core in /var/log/samba/cores/smbd smb | 2021-05-26T21:44:10.115081870Z Registered MSG_REQ_POOL_USAGE smb | 2021-05-26T21:44:10.122442859Z check_ntlm_password: Authentication for user [MY_USERNAME] -> [MY_USERNAME] FAILED with error NT_STATUS_NO_SUCH_USER, authoritative=1 MY_USERNAME in the above excerpt is the username of the Windows user account from which I sent requests to the SMB server. Such errors occur only when the log level is higher than 1, but in this case it's enough to open Windows's built-in file manager (assuming a share is connected from the SMB server), because the error is triggered by every request that needs to be responded. When such errors happen, no file shares are served. notes: - selected the file services component, because I only use that part, but can't confirm it only affects that component. - hardware is 32 bit ARM (Raspberry Pi 4 with 2 GB RAM)
Did you ever get a GDB backtrace of this?
Haven't forgot about the issue, actually I see this every few days because the issue's browser tab is near my email account's tab, but it seems it might take some more time to find out how to make the backtrace.
Does this still happen? We really need more information to know what is going on.
It does not anymore. But I should also note that possibly I was using a very outdated version even when I reported the issue, as I was using a relatively common docker image which has been unmaintained for years: https://github.com/dperson/samba It may be off topic, but I think you should consider releasing some kind of announcement about this situation, considering that (in my understanding) those the use that Docker image are running highly exploitable software with high privileges. The dockerhub page does not give too much statistics, but the (all time) 50 million download suggests to me that it is very popular: https://hub.docker.com/r/dperson/samba/tags
(In reply to mpeter.68m0y from comment #4) > It may be off topic, but I think you should consider releasing some kind of announcement about this situation, considering that (in my understanding) those the use that Docker image are running highly exploitable software with high privileges. Yes, this problem is generally outside our awareness and control. My reading of https://github.com/dperson/samba/blob/master/Dockerfile RUN apk --no-cache --no-progress upgrade && \ apk --no-cache --no-progress add bash samba shadow tini tzdata && \ is that you'll end up with the version of Samba that Alpine Linux has at the time the Dockerfile is run, so if you build from there it is maybe OK. But the Dockerhub files have not kept up. Commenters on https://github.com/dperson/samba/issues/459 suggests these alternatives: https://github.com/dockur/samba https://github.com/ServerContainers/samba I am removing the restriction to Samba developers, because it is not useful for embargo knowledge that there was once a crash on 4.13 in a container on 32 bit arm.