[systemd-devel] Systemd stops

Tom Hirst tom.hirst at ipe-systems.co.uk
Tue Jun 24 07:56:27 PDT 2014


> 
> > I think I came across this issue the other day (also booting on an arm
> > board, stuck at kernel 3.1), I think that it may be the change to the
> > xattr support in 214 means systemd can no longer be booted if you
> > don't have cgroups xattr support. 213 seems fine, but I haven't tried
> > to bisect it or revert the xattr changes.
> 
> We don't rely on xattr support in the kernel at all, and not in cgroupfs. xattr
> support is fully optional. (that said, we sometimes break these builds
> accidentally, since its not what we test things on)
>

Just done a bit more digging into this and it probably isn't the same issue
the original poster had, but is probably worth reporting if this is supposed
to be a supported kernel version.

 Setup is an arm i.MX6 device (Compulab Utilite) running 3.0.35, booting 
into a Gentoo rootfs over nfs, not using an initrd. 

With 213 compiled with --disable-xattr it boots fine, using 214 or 213 with
--enable-xattr it hangs after "Freeing init memory", with no output at all 
from systemd (booted with systemd.log_level=debug).

I've just compiled up 214, but removed the MNT_FATAL flag from line 97 in
mount-setup.c and that boots and appears to be fully functional. If I understand
the code correctly it should now be trying to mount cgroup with xattr first 
(not fatal if it  fails) then retrying without xattr (but fatal on failure this time). 
I don't know if this will have any other side effects or is the correct fix.

Tom

> Lennart
> 
> --
> Lennart Poettering, Red Hat


More information about the systemd-devel mailing list