[systemd-devel] [PATCHv4] core: notify triggered by socket of a service

Umut Tezduyar umut at tezduyar.com
Wed Oct 2 04:59:24 PDT 2013


Hi,

I haven't seen a delay on my boot but I was thinking maybe you can try
to comment out following section to see if it will make any
difference.

+        if (se->state == SERVICE_RUNNING)
+                socket_set_state(s, SOCKET_RUNNING);



On Wed, Oct 2, 2013 at 12:50 PM, Colin Guthrie <gmane at colin.guthr.ie> wrote:
> 'Twas brillig, and Colin Guthrie at 21/09/13 11:57 did gyre and gimble:
>> 'Twas brillig, and Colin Guthrie at 19/09/13 00:54 did gyre and gimble:
>>> Hi,
>>>
>>> While I'm not 100% sure, (not done quite enough testing), but this patch
>>> seems to have introduced some kind of race that results in me seeing a
>>> 1min delay in boot. This delay creeps in before
>>> systemd-udevd-control.socket starts so my dmesg looks like:
>>>
>>>  [    1.549666] RPC: Registered tcp NFSv4.1 backchannel transport module.
>>>  [   61.592613] systemd[1]: ConditionPathIsReadWrite=/sys succeeded
>>> for systemd-udevd-control.socket.
>>>
>>>
>>>
>>> And it was the mention socket related stuff that reminded me of this
>>> thread and patch.
>>>
>>> So I reverted the commit (otherwise running git master) and have
>>> performed three successful boots so far without issue. This is obviously
>>> not conclusive, but sadly I also have a problem shutting down quickly...
>>> seems my httpd.service file now also stalls until it times out. When the
>>> ControlGroup becomes empty, it seems this isn't properly detected and
>>> systemd keeps the unit in the "deactivating" state until it times out.
>>> I'll post a separate mail about this issue.
>>>
>>> I'll keep testing and reporting if I still get the delayed boot.
>>
>> Since applying Olivier Brunel's type=notify patch it seems my httpd
>> timeout on shutdown problem is fixed and I have done several reboots.
>>
>> Seems that after reverting this patch, I've not had any more startup
>> delays, so I would say there is definitely something up with it.
>
>
> Not seen anyone else commenting on this one.... do others not see this
> deadlock I mentioned?
>
> I'll do some more tests with 208 soon.
>
> Col
>
>
> --
>
> Colin Guthrie
> gmane(at)colin.guthr.ie
> http://colin.guthr.ie/
>
> Day Job:
>   Tribalogic Limited http://www.tribalogic.net/
> Open Source:
>   Mageia Contributor http://www.mageia.org/
>   PulseAudio Hacker http://www.pulseaudio.org/
>   Trac Hacker http://trac.edgewall.org/


More information about the systemd-devel mailing list