[arch-general] at times when booting system-logind fails and system hangs

[Figue]

On 09/05/15 00:03, Steven Noonan wrote:
> On Fri, May 1, 2015 at 6:10 PM, Javier Vasquez <j.e.vasquez.v at gmail.com> wrote:
>>> On Fri, May 1, 2015 at 4:11 AM, Mark Lee <mark at markelee.com> wrote:
>>> ...
>>>
>>> This is unrelated, but does Arch Linux respect rootwait?
>> Linux does, :-)  I don't need on this laptop, but I have Arch also on
>> a USB disk, and for that one, at times linux doesn't find root, so I
>> found it just needed to wait for it to become available...
>>
>>> I don't see any issues in there that'd cause your systemd issues. Did you try
>>> reinstalling systemd?
>> No I haven't.  Though I wouldn't suppose such re-install (if no new
>> version implied) would help.  I'll upgrade Arch to see.  The other
>> option is to tried the systemd patches, or the patched systemd offered
>> by Christian.  I haven't done that either...
> For what it's worth I've been having these same issues. systemd-logind
> fails to start with "connection timed out" and then everything is
> useless until the machine is reset. On most of my machines it happens
> in maybe 1 out of every 10 boots, but on one of my machines it's
> *every* boot. I can't figure out what's causing it, even with
> systemd.log_level=debug.
>
> I'm noticing that right before the systemd-logind timeout messages it
> tries to start "dbus-daemon --system --address=systemd: --nofork
> --nopidfile --systemd-activation" -- it seems like this should have
> happened much earlier though, when dbus.socket started. Or am I
> misunderstanding the relationship between dbus.socket and
> dbus.service?
>
> - Steven
I made some research some time ago and I found this: 
http://cgit.freedesktop.org/systemd/systemd/commit/?id=13790add4bf64 
which seems that is the commit that make a regression in journald. That 
causes boot failures due to logind. This Debian bug report has more 
information: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=778970

So applying this patch in systemd package should avoid the bug: 
http://hastebin.com/idaquqarud.coffee

Just add to prepare() in the PKGBUILD. Something like this: 
http://hastebin.com/roqabepilo.vhdl

I've been running this patched version for a few months and it works fine.

Joan Figueras