Changing elogind IPCRemove default.

Mark Hindley mark at hindley.org.uk
Sat Jan 25 11:26:02 GMT 2020


On Fri, Jan 24, 2020 at 03:22:06PM +0100, Thorsten Glaser wrote:
> On Fri, 24 Jan 2020, Mark Hindley wrote:
> 
> > This is essential for his usage.
> 
> Not just that, it is also essential to not break programs using
> shared memory as well, which I mentioned.

I fully understand that.

That is not really the question here. I am really asking if changing the default
might have any downsides.
 
> > In bug #949698, the reporter is demanding that we change the default for
> > IPCRemove from true to false so that /dev/shm is not cleared on user logout.
> 
> This is just a workaround anyway. The *proper* fix is to only remove
> those files from /dev/shm/ which elogind itself (or something on its
> behalf) created and leave everything else alone, of course.

I am not sure that is correct. 

For good or bad, it appears the behaviour of RemoveIPC originates in
systemd. See

 https://github.com/systemd/systemd/issues/2039

elogind is not trying to clear up after itself (AFAIK it creates no files in
/dev/shm), it is actively limiting the use of resources after a user logs out.

The NEWS file says

 'Traditionally, SysV and POSIX IPC had no life-cycle limits. With this
 functionality, that is corrected. This may be turned off by using the
 RemoveIPC= switch of logind.conf'

I am explicitly *not* arguing that this view is correct. I just want us to fully
understand the implications of changing the default when it is easy for the
sysadmin to configure the behaviour as they want.

Thanks

Mark




More information about the Debian-init-diversity mailing list