Bug#748651: Bug#745550: May be related to systemd

[Felipe Sateler]

On Tue, May 27, 2014 at 12:13 PM, Eric Valette <Eric.Valette at free.fr> wrote:
> On 05/27/2014 05:23 PM, Felipe Sateler wrote:
>
>> Hi Eric,
>
>> So, let me summarize what you did:
>
>> 1. After some upgrades, pulseaudio didn't see the analog output, just
>> the HDMI one.
>
> I just *first* noticed I had no more sound. As in both case I use analog
> output and not HDMI (because monitors I use have no speakers), the cause was
> probably the same (no more analog audio device) on the two machines although
> after fixing it on one computer I only verified that reinstalling
> sysinit-core on the second and rebooting also fixed the problem on the
> second machine.
>
> 2. Around the same time, pulseaudio started hanging on startup,
> returning EAGAIN from the system dbus socket
>
> While trying to debug the "no sound", I noticed first kde audio/multimedia
> config was hanging on my first machine and later on my other machine at work
> that indeed the pulseaudio was stuck for nearly 2 mins (this probably
> causing the kde multimedia config panel to hang waiting for pulse) and that
> anyway, at the end the analog audio device was not there.
>
> 3. Installing sysvinit-core fixes both issues
>
> Yes on both machines.
>
> Is this understanding correct?
>
> Yes.
>
>
> If so, I'm still not sure these are both the same bug: (3) could be a
> race in the boot scripts only exposed by systemd, unrelated to the
> dbus issue.

I'm going to treat them as separate for now. The originally reported
bug does not look like this one.

>
> I dunno if the bugs are indeed exactly the same or two expressions of a same
> root cause but they are both fixed when going back to sysv init. The race in
> the boot happens on very different type machine (an old core i5 laptop
> single core/hyperthreaded with classical disk 4GB RAM and a high end core I7
> quad core hyperthreaded with SSD and 16 GByte memory).
>
>
> > Are you willing to try to reproduce the issues again using systemd?
> > This can be done by passing init=/bin/systemd in the grub screen (no
> > need to reinstall systemd-sysv). Although I fear we will need help
> > from the dbus and systemd folks to debug this.
>
> If you do not manage to reproduce the bug I can spend a limited amount of
> time at it. The fact that the two machines are very different may help
> reproducing it. Maybe you could try to get info from original reporter to
> see if his problem is also fixed by reinstalling  sysvinit-core as an extra
> hint.

Thanks, I indeed cannot reproduce it (and I'm using systemd 208 as well).

Could you please boot with systemd but with full logging? Append to
the boot line:

init=/bin/systemd systemd.log_level=debug

Also please attach a full log of pulseaudio. See
https://wiki.ubuntu.com/PulseAudio/Log for instructions on how to get
one.
And then attach the log produced by `journalctl -b > system.log`
(please get the full log after running pulseaudio).

>
> NB: both machines were using systemd 208 from experimental. I dunno if
> problem is related to this version of systemd because I'm nearly 100% sure I
> have been using systemd208 with sound for a while before it broke the same
> day after upgrading.
>
> And to be fair about my setup that is *unusual*:  both machines have a
> distinct / and /usr (I know this i now considered to be bad for systemd but
> the 7.0 installer stills allows that without even emitting a warning...) ,
> do not use initramfs and self tuned recent kernels dedicated to the machines
> with all drivers build-in (except nvidia one ;-)).

This configuration is not really supported, though. Not sure if it is related.

-- 

Saludos,
Felipe Sateler