Bug#766092: systemd: Boot hangs indefinitely

[Martin Steigerwald]

Am Donnerstag, 23. Oktober 2014, 00:15:48 schrieb Marcelo Laia:
> On 22/10/14 at 07:15pm, Michael Biebl wrote:
> > Am 22.10.2014 um 19:08 schrieb Michael Biebl:
> > > If you even had more files in /tmp and maybe you have an HDD, it's very
> > > well possible, that systemd-tmpfiles will need several minutes.
> > > That's why I wanted to know, how long you let the systemd-tmpfiles job 
> > > run.> 
> > Case in point: A user at [1] reported, that he had about 2 million files
> > 
> > in /tmp and it took several hours to clean that up:
> > >  * systemd-tmpfiles-setup.service hung like forever during boot
> > >  * booting from a live system showed that 'ls ./tmp' hangs
> > >  * 'ls -U ./tmp | wc -l' showed nearly 2 million files in tmp
> > >  * 'find ./tmp -type l -delete' took several hours to delete the links
> 
> I have tried:
> 
> # rm -rf *
> 
> and got "the list was to big" or something else.
> 
> My system only work after reboot.

Well thats a pitfall on Unix shells. The shell completes the wildcard, not the 
application (unlike on AmigaDOS, well both approaches have advantages and 
disadvantages). Unless you use a tool that can use wildcards by itself.

I think find -delete in the right directory (!!!) could work for you.

Or some find -exec rm \; which calls rm for each file, but would be inefficient 
due to that, or some find | xargs -n1000 rm like combination.

-- 
Martin 'Helios' Steigerwald - http://www.Lichtvoll.de
GPG: 03B0 0D6C 0040 0710 4AFA  B82F 991B EAAC A599 84C7