View previous topic :: View next topic |
Author |
Message |
Princess Nell l33t
Joined: 15 Apr 2005 Posts: 916
|
Posted: Tue Nov 22, 2016 12:55 am Post subject: Debugging boot and shutdown |
|
|
I have three questions here.
I recently noticed a warning during boot to the effect that binfmt-misc should be added to /etc/conf.d/modules. Why is that? It does get loaded eventually as verified after boot.
This is one I find strange - system uses openrc and no initrd. Once init starts, after "Caching service dependencies", I get this:
Code: | File descriptor 3 (pipe:[1218]) leaked on lvm invocation. Parent PID 397: /bin/sh
Failed to find sysfs mount point
random: lvm: uninitialized urandom read (4 bytes read, 94 bits of entropy available)
/etc/portage/make.conf: line53: /var/lib/layman/make.conf: No such file or directory
Error: lvmetad is the name of a real and virtual service
|
/var is on a separate filesystem and this is very early in the boot process, before local filesystems are mounted. How can I get rid of this? I'm not using lvm and had the lvm service starting in the boot runlevel; removing it doesn't change the above significantly, I think only line starting with "random" is omitted. The only reference to lvmetad is in rc.conf, as per https://forums.gentoo.org/viewtopic-t-1010228.html.
Lastly, this is actually on shutdown. There is some message from openrc about an iptables related task, but I can't remember the correct wording, and it doesn't seem to be issued on every reboot. |
|
Back to top |
|
|
Princess Nell l33t
Joined: 15 Apr 2005 Posts: 916
|
Posted: Tue Nov 22, 2016 8:33 am Post subject: |
|
|
Another thing that may or may not be related: an empty directory /(null) gets created during boot. Looks like something going seriously wrong. |
|
Back to top |
|
|
SylaSebas n00b
Joined: 07 Aug 2013 Posts: 48
|
Posted: Sun Dec 04, 2016 2:44 pm Post subject: |
|
|
hey there
i got partly the same 'error' during booting
Code: |
File descriptor 3 (pipe:[1051]) leaked on lvm invocation. Parent PID 397: /bin/sh
Failed to find sysfs mount point
|
did u find anything related to this yet?
also i don't think I get the /(null) folder, do you just see it when u look at / directory?
thanks for any information |
|
Back to top |
|
|
Princess Nell l33t
Joined: 15 Apr 2005 Posts: 916
|
Posted: Wed Dec 07, 2016 10:41 pm Post subject: |
|
|
Code: |
host / # ll -d /*n*
drwxr-xr-x 2 root root 4096 Dec 7 21:01 /bin/
drwx------ 2 root root 16384 Oct 1 2014 /lost+found/
drwxr-xr-x 7 root root 4096 Apr 10 2015 /mnt/
drwx------ 2 root root 4096 Nov 22 00:32 '/(null)'/
drwxr-xr-x 23 root root 780 Dec 7 22:26 /run/
drwxr-xr-x 2 root root 4096 Dec 5 03:30 /sbin/
host / # ll -a '/(null)'
total 8
drwx------ 2 root root 4096 Nov 22 00:32 ./
drwxr-xr-x 23 root root 4096 Nov 22 00:09 ../
host / #
|
I think this is all related to the latest openrc update to 0.22.4. Another machine still on 0.21.7 doesn't show this behaviour. Need to check the bug tracker. |
|
Back to top |
|
|
charles17 Advocate
Joined: 02 Mar 2008 Posts: 3664
|
Posted: Thu Dec 08, 2016 5:32 am Post subject: Re: Debugging boot and shutdown |
|
|
Princess Nell wrote: | Code: | ...
...
...
/etc/portage/make.conf: line53: /var/lib/layman/make.conf: No such file or directory
...
|
|
Are you using layman at all? You might consider removing that line from /etc/portage/make.conf and using the repos.conf method (default). |
|
Back to top |
|
|
Princess Nell l33t
Joined: 15 Apr 2005 Posts: 916
|
Posted: Thu Dec 08, 2016 11:26 am Post subject: |
|
|
This output must have been created before local filesystems were mounted. |
|
Back to top |
|
|
SylaSebas n00b
Joined: 07 Aug 2013 Posts: 48
|
Posted: Sun Dec 11, 2016 4:43 pm Post subject: |
|
|
i agree on that ...
still didn't figure out where the message is coming from tbh. anyone any idea at all? |
|
Back to top |
|
|
rogerx Tux's lil' helper
Joined: 06 Apr 2004 Posts: 118
|
Posted: Sat Dec 17, 2016 1:01 am Post subject: |
|
|
There's a discussion about this lvm service within the following thread URL:
https://forums.gentoo.org/viewtopic-t-1040538.html
Guessing adding rc_need="!lvmetad" or rc_need="lvmetad" discussed within the last two posts are probably heading in the right direction. But somebody made mention why the LVM error when LVM is NOT required on a system sounds very odd! I agree with the later statement here, although experimenting with the rc_need. _________________ Roger
http://rogerx.freeshell.org/ |
|
Back to top |
|
|
SylaSebas n00b
Joined: 07 Aug 2013 Posts: 48
|
Posted: Sat Dec 17, 2016 4:40 pm Post subject: |
|
|
RC_NEED="lvmetad" is already in /etc/conf.d/lvm but the error still appears
also i removed lvm and lvmetad from runlevels as I dont need them atm but the error is still there
I guess i could remove lvm support from the kernel completely but that wouldnt really solve the problem |
|
Back to top |
|
|
rogerx Tux's lil' helper
Joined: 06 Apr 2004 Posts: 118
|
Posted: Mon Dec 19, 2016 2:04 am Post subject: |
|
|
Ditto.
I was getting two duplicates of the previously mentioned error. Solved one of them, but still one printed warning/error remaining.
lvm is only needed as it's marked depend of crypt-setup _________________ Roger
http://rogerx.freeshell.org/ |
|
Back to top |
|
|
SylaSebas n00b
Joined: 07 Aug 2013 Posts: 48
|
Posted: Tue Dec 20, 2016 6:09 pm Post subject: |
|
|
well I added lvm to package.keywords and did an update - problem seems gone
not the most elegant solution I guess but it works |
|
Back to top |
|
|
rogerx Tux's lil' helper
Joined: 06 Apr 2004 Posts: 118
|
Posted: Thu Dec 22, 2016 1:00 pm Post subject: |
|
|
I can also confirm moving from =sys-fs/lvm2-2.02.116-r4 to marked unstable =sys-fs/lvm2-2.02.166 version cleans-up all of my lvm leaked file descriptors and sysfs errors during boot here as well. (Likely looking at the sys-fs/lvm2 ChangeLog between or after these version releases might add some light upon what happened or didn't happen. Dunno where the ChangeLog files are, as it looks like somebody hid them.)
In my case, it does not even look as if I have added the init.d lvm and lvmetad services to the boot or default runlevels. (eg. "rc-update add lvm boot")
Was somebody running SystemD when they tested the currently marked lvm2 stable ebuild? _________________ Roger
http://rogerx.freeshell.org/ |
|
Back to top |
|
|
Hu Moderator
Joined: 06 Mar 2007 Posts: 21633
|
Posted: Fri Dec 23, 2016 2:28 am Post subject: |
|
|
From the Portage tree ChangeLog: /usr/portage/sys-fs/lvm2/ChangeLog: | *lvm2-2.02.166 (21 Oct 2016)
21 Oct 2016; Robin H. Johnson <robbat2@gentoo.org> files/clvmd.rc-2.02.39,
files/device-mapper.rc-2.02.105-r2, files/dmeventd.initd-2.02.67-r1,
files/lvm-monitoring.initd-2.02.105-r2, files/lvm.rc-2.02.105-r2,
files/lvm.rc-2.02.116-r4, files/lvm.rc-2.02.116-r6,
files/lvmetad.initd-2.02.105-r2, files/lvmetad.initd-2.02.116-r3,
lvm2-2.02.116.ebuild, lvm2-2.02.116-r1.ebuild, lvm2-2.02.116-r2.ebuild,
lvm2-2.02.136.ebuild, +lvm2-2.02.166.ebuild:
bump. | If you seek enlightenment, look not to the Portage tree ChangeLog files, for they describe only the actions of the Gentoo maintainers, and then only very briefly. Upstream's changelog might tell you about changes made by the upstream developers, but lvm2 either does not have one or does not install it. |
|
Back to top |
|
|
rogerx Tux's lil' helper
Joined: 06 Apr 2004 Posts: 118
|
Posted: Fri Dec 23, 2016 5:58 pm Post subject: |
|
|
I noticed those posted lvm2 ChangeLog files (as well as the ChangeLog files available on the Internet) are extremely vacant of any explanations when compared to the following changes within the EBuild's themselves. Think we've been around this issue many times, trying to encourage people whom make the EBuild changes to also make quick and brief meaningful entries within the Gentoo Portage ChangeLog files. After awhile, seems a waste of time. But then again, it makes my coding/scripting look even more exceptionally and excellent.
With the following, I fared a little better.
$ diff -urN /usr/portage/sys-fs/lvm2/lvm2-2.02.116.ebuild /usr/portage/sys-fs/lvm2/lvm2-2.02.166.ebuild
(Although I was simply, sort of, joking concerning SystemD; there is a systemd change at the bottom. Whether it's related, I don't know.)
Aside from a SystemD change to omit installing systemd related stuff, there are several additional patches added to the EBuild. This fix could have been also a source code package change, as already previously mentioned. More often from where I sit, I usually always find the cause of these bugs to be with larger confusing EBuild files, versus smaller and simpler EBuild script files.
Funny, off-topic, but all my ChangeLog files are missing. I was assuming this was a change to Portage, doing something against the norm as usual.
$ find /usr/portage/* -iname "*changelog*" _________________ Roger
http://rogerx.freeshell.org/ |
|
Back to top |
|
|
Princess Nell l33t
Joined: 15 Apr 2005 Posts: 916
|
|
Back to top |
|
|
robertsurma n00b
Joined: 17 Jan 2008 Posts: 56 Location: Tychy
|
Posted: Tue Dec 27, 2016 8:38 am Post subject: |
|
|
The same problem.
I set in /etc/rc.conf:
Code: | rc_lvm_after="sysfs"
rc_lvm_need="sysfs"
rc_sysfs_before="lvm" |
in /etc/conf.d/lvm:
in /etc/lvm/lvm.conf
and of course:
Code: | # rc-update del lvm boot
# rc-update del lvmetad boot |
Not working for me. _________________ Free as a bird... |
|
Back to top |
|
|
rogerx Tux's lil' helper
Joined: 06 Apr 2004 Posts: 118
|
Posted: Tue Dec 27, 2016 3:10 pm Post subject: |
|
|
I'm assuming you're not wanting to use LVM by the last "rc-update del" statements.
If so, revert everything to default and set "=sys-fs/lvm2-2.02.166" within /etc/portage/package.keywords. (Unmask/set keywords for version 2.02.166.)
# echo "=sys-fs/lvm2-2.02*" >> /etc/portage/package.keywords
# emerge -uq sys-fs/lvm2 && etc-update _________________ Roger
http://rogerx.freeshell.org/ |
|
Back to top |
|
|
robertsurma n00b
Joined: 17 Jan 2008 Posts: 56 Location: Tychy
|
Posted: Tue Dec 27, 2016 5:43 pm Post subject: |
|
|
Success! Updating to version lvm2-2.02.166 solved the problem.
As you suggested, additionally I returned to the default settings and now everything is OK. Thanks! _________________ Free as a bird... |
|
Back to top |
|
|
rogerx Tux's lil' helper
Joined: 06 Apr 2004 Posts: 118
|
Posted: Tue Dec 27, 2016 8:54 pm Post subject: |
|
|
Glad you followed-up after achieving success! _________________ Roger
http://rogerx.freeshell.org/ |
|
Back to top |
|
|
Princess Nell l33t
Joined: 15 Apr 2005 Posts: 916
|
Posted: Sun Feb 26, 2017 1:30 pm Post subject: |
|
|
Same here - updating to lvm2-2.02.166 solved that particular problem.
Still have the other problem, though,
Code: | /etc/portage/make.conf: line53: /var/lib/layman/make.conf: No such file or directory |
not sure how to address. This line is issued after "* Caching service dependencies", but I don't know which runlevel service is responsible here. |
|
Back to top |
|
|
rogerx Tux's lil' helper
Joined: 06 Apr 2004 Posts: 118
|
Posted: Sun Feb 26, 2017 5:06 pm Post subject: |
|
|
Somewhere within your /etc/portage/make.conf file a line is listed as "/var/lib/layman/make.conf", and likely is "source /var/lib/layman/make.conf" as contained within my /etc/portage/make.conf file.
I believe this "/var/lib/layman/make.conf" file is really apart of or used by the app-portage/layman tool, which syncs and downloads other Portage overlay trees and files. (eg. The more popular Sunrise overlay)
I would likely just do the following, as you're just likely seeing a warning.
# touch /var/lib/layman/make.conf
Looking at my /var/lib/layman/make.conf file here, the file only contains the names of the overlays I have subscribed to using the layman tool. If you do not use any overlays, just touch/create a blank file as described above, as likely the layman package installed the source line into your make.conf. Or, add an overlay using the layman tool and then delete the overlay, having a similar result to just touching or creating a blank file. _________________ Roger
http://rogerx.freeshell.org/ |
|
Back to top |
|
|
Princess Nell l33t
Joined: 15 Apr 2005 Posts: 916
|
Posted: Sun Feb 26, 2017 10:03 pm Post subject: |
|
|
The problem is that /var isn't mounted yet when this as of yet unidentified code runs. |
|
Back to top |
|
|
|