Mailing List Archive: mythtv segmentation faults

mythtv segmentation faults - the mystery deepens

Apr 18, 2012, 11:30 AM

Post #1 of 5 (1889 views)

ahhhhrrrg I spoke too soon!

I thought that having removed and reinstalled the 0.25-285 rpms everything
was working fine but now after a couple of days running all the mythtv
suite of programs *segfault as soon as the program is started *again.

No rpms have been installed or changed
the box has just been on and has made a few recordings (plus had a few
recordings watched and deleted).

What is really strange is that rebooting makes no difference (all the
mythtv programs still segfault immediately) but removing and reinstalling
the same rpms (0.25-285) as were already installed (again) fixes this!

So can anyone suggest what sort of issue would:
1) takes several hours (or days?) to occur (on system which has previously
been running stably continuously for over five years)
2) once the issue has occurred it will cause *all* mythtv suite
(mythfrontend/mythbackend etc) programs to consistently (and immediately)
segfault as soon as the program is started
3) the issue will persist over restart(s)
4) the issue is "cleared" by simply removing and reinstalling the same rpms?
5) this issue started occurring at about the 0.24.2-281 went into
atrpms-stable (around 9th of April) - but could well due to an update from
an unrelated package at around this time....

some traces from mythtvpreviewgen:

strace mythpreviewgen
http://pastebin.com/JziKdUWG

ldd /usr/bin/mythpreviewgen
http://pastebin.com/SfDZucGr

strace -eopen mythpreviewgen
http://pastebin.com/a79TTtxM

as previously any help greatly appreciated (the wife is not to happy with
the situation!)

many thanks

Jerome

Re: mythtv segmentation faults - the mystery deepens [ In reply to ]

steve at sjcnet

Apr 19, 2012, 2:10 AM

Post #2 of 5 (1858 views)

Permalink

That is certainly a strange problem, and I'd be tempted to think that a
file is somehow getting corrupted.

Maybe you could do a clean install of the rpms, wait for a day and the
problem to occur and then run something like "find /usr -mtime -1" to
see if any of the installed files have been modified?

I haven't upgraded my CentOS 6.2 x86_64 combined frontend/backend to
0.25 yet - hopefully next week.

Steve.

On 19/04/2012 4:30 AM, Jerome Hettich wrote:
> ahhhhrrrg I spoke too soon!
>
> I thought that having removed and reinstalled the 0.25-285 rpms
> everything was working fine but now after a couple of days running all
> the mythtv suite of programs *segfault as soon as the program is
> started *again.
>
> No rpms have been installed or changed
> the box has just been on and has made a few recordings (plus had a few
> recordings watched and deleted).
>
> What is really strange is that rebooting makes no difference (all the
> mythtv programs still segfault immediately) but removing and
> reinstalling the same rpms (0.25-285) as were already installed
> (again) fixes this!
>
> So can anyone suggest what sort of issue would:
> 1) takes several hours (or days?) to occur (on system which has
> previously been running stably continuously for over five years)
> 2) once the issue has occurred it will cause *all* mythtv suite
> (mythfrontend/mythbackend etc) programs to consistently (and
> immediately) segfault as soon as the program is started
> 3) the issue will persist over restart(s)
> 4) the issue is "cleared" by simply removing and reinstalling the same
> rpms?
> 5) this issue started occurring at about the 0.24.2-281 went into
> atrpms-stable (around 9th of April) - but could well due to an update
> from an unrelated package at around this time....
>
> some traces from mythtvpreviewgen:
>
> strace mythpreviewgen
> http://pastebin.com/JziKdUWG
>
> ldd /usr/bin/mythpreviewgen
> http://pastebin.com/SfDZucGr
>
> strace -eopen mythpreviewgen
> http://pastebin.com/a79TTtxM
>
> as previously any help greatly appreciated (the wife is not to happy
> with the situation!)
>
> many thanks
>
> Jerome
>
>
> _______________________________________________
> atrpms-users mailing list
> atrpms-users@atrpms.net
> http://lists.atrpms.net/mailman/listinfo/atrpms-users

Re: mythtv segmentation faults - the mystery deepens [ In reply to ]

neil at fnxweb

Apr 19, 2012, 2:19 AM

Post #3 of 5 (1857 views)

Permalink

Around about 18/04/12 19:30, Jerome Hettich typed ...
> I thought that having removed and reinstalled the 0.25-285 rpms everything
> was working fine but now after a couple of days running all the mythtv
> suite of programs segfault as soon as the program is started again.
>
> What is really strange is that rebooting makes no difference (all the
> mythtv programs still segfault immediately) but removing and reinstalling
> the same rpms (0.25-285) as were already installed (again) fixes this!

Ah, OK, are you using the newer nvidia gfx driver?

I had a similar thing happen to me on my main Fedora 14 box a few months
ago (I don't recall if it was mythfrontend or not, though): somehow, the
newer nvidia kmdls were causing prelink to cock up. Any running processes
would be unaffected by any overnight run of prelink as they'd continue using
the â€œvirginâ€ versions, and of course an RPM replacement would briefly â€œfixâ€ it.

I had hoped that this was a transient fault, but haven't updated since to
test it. I undid the prelink (prelink -a -u) and then disabled prelink
(PRELINKING=no in /etc/sysconfig/prelink): my box is fast enough that I'm
not worried about that.

â€˜prelink -a -uâ€™ is relatively fast anyway (well, was for me :) ) so could
well be worth trying whatever.

--
[neil@fnx ~]# rm -f .signature
[neil@fnx ~]# ls -l .signature
ls: .signature: No such file or directory
[neil@fnx ~]# exit

_______________________________________________
atrpms-users mailing list
atrpms-users@atrpms.net
http://lists.atrpms.net/mailman/listinfo/atrpms-users

Re: mythtv segmentation faults - the mystery deepens [ In reply to ]

neil at fnxweb

Apr 20, 2012, 12:21 AM

Post #4 of 5 (1867 views)

Permalink

Around about 19/04/12 18:39, Jerome Hettich typed ...
> looks like this is it!
> overnight last night the myth* programs started seg faulting again
> (presumably after prelinking) and removing prelinking fixes this - so I
> think you have hit the nail on the head.

Glad it's sorted; I now have the â€œjoyâ€ of updating my box to 0.25 this
weekend, so crossed-fingers there aren't any nasties ...

> many many thanks for the help Neil (and Steve for your suggestions) - it
> definitely would have taken me a long time (if at all) to figure this out
> without your help.

It was largely down to luck, I think, that I found refs. to the nvidia
drivers and prelink issues; I need to have another look into it to see if
it's going to be fixed. A cursory glance now implies it's not nvidia per
se, but prelink that's stuffing up newer nvidia binaries (and presumably
apps. linked against them).

--
[neil@fnx ~]# rm -f .signature
[neil@fnx ~]# ls -l .signature
ls: .signature: No such file or directory
[neil@fnx ~]# exit

_______________________________________________
atrpms-users mailing list
atrpms-users@atrpms.net
http://lists.atrpms.net/mailman/listinfo/atrpms-users

Re: mythtv segmentation faults - the mystery deepens [ In reply to ]

neil at fnxweb

Apr 20, 2012, 12:43 AM

Post #5 of 5 (1866 views)

Permalink

Around about 20/04/12 08:21, Neil Bird typed ...
> It was largely down to luck, I think, that I found refs. to the nvidia
> drivers and prelink issues; I need to have another look into it to see if
> it's going to be fixed. A cursory glance now implies it's not nvidia per
> se, but prelink that's stuffing up newer nvidia binaries (and presumably
> apps. linked against them).

A previous discussion of this issue (on this list!) seems to have also
found that you can work around the issue by blacklisting the nvidia drivers
for prelink in prelink.conf.

Ref:
http://www.gossamer-threads.com/lists/atrpms/users/1543?do=post_view_threaded

Axel, might it be worthwhile adding

/etc/prelink.conf.d/atrpms-nvidia.conf:
-b /usr/lib/nvidia-graphics-*

?

Apparently, doing this allows you to enable prelink as before, but omit
the nvidia-related crashes. I shall try it this weekend. I suppose it
should be possible to determine the version(s) where this issue started, and
only omit those (e.g., /usr/lib/nvidia-graphics-2*) but it may be safer to
do the lot.

--
[neil@fnx ~]# rm -f .signature
[neil@fnx ~]# ls -l .signature
ls: .signature: No such file or directory
[neil@fnx ~]# exit

_______________________________________________
atrpms-users mailing list
atrpms-users@atrpms.net
http://lists.atrpms.net/mailman/listinfo/atrpms-users