Re: xdm failing to start on FBSD 10.0 r2340030 erratically

From: O. Hartmann <ohartman_at_zedat.fu-berlin.de>
Date: Mon, 09 Apr 2012 11:59:49 +0200
Am 04/09/12 11:15, schrieb Anton Shterenlikht:
> On Mon, Apr 09, 2012 at 10:54:07AM +0200, O. Hartmann wrote:
>> Am 04/08/12 17:29, schrieb David Wolfskill:
>>> On Sun, Apr 08, 2012 at 01:29:36PM +0200, O. Hartmann wrote:
>>>> I loose hair ...
>>>> Since yesterady's "make world" (last make world: the day before
>>>> yesterday), getting FreeBSD 10.0-CURRENT/amd64 to r234000 or so, the X11
>>>> system on all of our FreeBSD 10.0-CUR/amd64 boxes start rejecting the
>>>> start of xdm display manager. xdm is started from /etc/ttys on ttyv7.
>>>> This worked before flawless.
>>>>
>>>> At this very moment, I do have X11 started via xdm - but this is a
>>>> erratic and non-reproduceable process!
>>>>
>>>> This morning, I update world and kernel to r234030. I recompiled many
>>>> ports via "portmaster -f xorg xdm", hoping the new kernel/world could
>>>> affect the ports, but this isn't.
>>>>
>>>> Starting Xorg X11 server works fine. xdm fails. The log in
>>>> /var/log/xdm.log looks like:
>>>>
>>>> Build Date: 07 April 2012  04:51:08PM
>>>> ...
>>>>
>>>>
>>>> I feel a bit helpless around here since I can not get close to what is
>>>> happening. The erratic behaviour of starting xdm is frightening.
>>>> Starting xdm via /etc/ttys doesn't work at all, but sometimes, with a
>>>> bit luck, xdm starts when started from the console.
>>>> ....
>>>
>>> I am not having trouble starting xdm via /etc/ttys; my environment is
>>> known to differ from yours in the following ways:
>>>
>>> * My ports (save for x11/nvidia-driver and misc/compat8x) are built
>>>   under stable/8.  (I track stable/8, stable/9, and head on a daily
>>>   basis, and have a common /usr/local among them.  I also update
>>>   any installed ports that have updates available daily.)
>>>
>>> * I am running FreeBSD/i386, vs. FreeBSD/amd64.
>>>
>>> * My last 2 updates were r233994 (yesterday) and r234031 (today).
>>>
>>> * I have a mildly hacked-up startup script for xdm.  I doubt this is an
>>>   issue -- I've been doing this since 2006/03/05 19:04:03 (according to
>>>   the RCS log), though I have modified the script a few times
>>>   since its inception.  But the point here is that I'm not directly
>>>   invoking the xdm executable from /etc/ttys.
>>
>> When I start xdm via console, it starts up, but only with a "Xservers"
>> file that contains nothing but comments. If there are the lines
>>
>> localhost
>> 127.0.0.1
>> 192.168.0.128
>> !*
>>
>> it will will probably start not. Since xdm does not show this behaviour
>> in a reproducible manner I susepct a bug.
>>
>>>
>>> * I don't know that this is different, but it may well be:  my xorg.conf
>>>   includes a stanza:
>>>
>>>   Section "ServerFlags"
>>>       Option         "AutoAddDevices" "False"
>>>   EndSection
>>
>> I should go with this and try. But as far as I know, since I have USB
>> devices (mouse, keyboard), unpluggin and pluggin them is then, without
>> hal and dbus, not recognized anymore, isn't it?
>>
>> There was a discussion once going one for this subject.
>>
>>
>> Either way, at this very moment I do not know whether the OS or the X11
>> is faulty and I'd like to report this as a bug.
>>
>>>
>>>   because my experiences with hald & dbus were so unpleasant.  I don't
>>>   use them; I don't even try to start them.
>>>
>>> Peace,
>>> david
>>
>> Regards,
>> Oliver
>>
> 
> I see the same (or similar) behavior on r231158M amd64
> Compaq 6715s laptop.
> 
> I've:
> 
> BUZI> pkg info xdm
> xdm-1.1.11: X.Org X display manager
> BUZI> 

pkohartmann_at_thor: [~] g_info |grep xdm
xdm-1.1.11          X.Org X display manager

> 
> Starting xdm from /etc/ttys:
> 
> BUZI> grep bin/xdm /etc/ttys
> ttyv8   "/usr/local/bin/xdm -nodaemon"  xterm   on  secure
> BUZI>

ohartmann_at_thor: [~] grep bin/xdm /etc/ttys
ttyv7   "/usr/local/bin/xdm -nodaemon"  xterm   on  insecure

> 
> causes the screen to flicker a few, maybe 3-5 times,
> between the graphical and the text console, and then
> I'm back to the text console.

Sometimes I also recognize a "flicker" and I see the xdm login requester
popping up, but only for a fraction of a second - and then vanishing.

> 
> However, starting xdm manually seems to work fine
> every time.

Not every time for me. Manipulating Xservers or Xaccess, even opening,
changing a char in the commented out portions and saving them, can
change the luck of starting or not - as I said, I can not reproduce this
behaviour in a predicted manner.

Starting xdm manually via console also does not work any time. If you
would kill xdm, removing the /var/run/xdm.pid file and try again - I my
case, this ending up in the error reported by xdm's /var/log/xdm.log:

Build Date: 07 April 2012  04:51:08PM

Current version of pixman: 0.24.2
        Before reporting problems, check http://wiki.x.org
        to make sure that you have the latest version.
Markers: (--) probed, (**) from config file, (==) default setting,
        (++) from command line, (!!) notice, (II) informational,
        (WW) warning, (EE) error, (NI) not implemented, (??) unknown.
(==) Log file: "/var/log/Xorg.0.log", Time: Sat Apr  7 18:38:24 2012
(==) Using config file: "/etc/X11/xorg.conf"
xdm info (pid 2055): sourcing /usr/local/share/X11/xdm/Xsetup_0
xdm error (pid 2050): Unknown session exit code 2560 from process 2055
xdm info (pid 2050): Exiting

> 
> The log file is no help:
> 
> BUZI> cat /var/log/xdm.log 
> xdm error (pid 1464): Can't lock pid file /var/run/xdm.pid, another xdm is running (pid 939)
> BUZI> 
> 
> 

Try to restart xdm via /etc/ttys and then watch the content of your
/var/log/xdm.log.

By the way:

My graphics card is a nVidia GPU GTX560Ti, running with the lates
nvidia-driver:
ohartmann_at_thor: [~] pkg_info | grep nvidia
nvidia-driver-295.33 NVidia graphics card binary drivers for hardware
OpenGL rendering

My world and kernel are at:

FreeBSD 10.0-CURRENT #5 r234046: Mon Apr  9 01:22:35 CEST 2012


As I reported before, this behaviour was introduced around r234000 and
came out of the blue, since I didn't change anything of the
configuration nor did I recompile anything else but the OS and kernel
that day. A reboot left me floating as described.

The only change I saw that day in the source updates was something with
init.c.
This could be a hint, but I'm a noob in terms of OS techniques, so this
is only an observation, not a conclusion.

Regards,
Oliver




Received on Mon Apr 09 2012 - 07:59:57 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:40:25 UTC