openrc support thread

[slackline]

The update wiped out my /etc/conf.d/net too. I tried to fix everything as well as I could, but I'm a little afraid of rebooting and seeing that I made a typo somewhere.

Don't worry too much, if you've made a typo its not too hard to bring the interface up manually. If its just a plain wired eth0 then something like...

Code: Select all

# ifconfig eth0 up
# dhcpcd eth0

..will get it up and running should the init-script fail. Make sure you've a symlink for /ec/init.d/net.eth0 and that its added to your default runlevel...

Code: Select all

# ln -s /etc/init.d/net.lo /etc/init.d/net.eth0
# rc-update add net.eth0 default

If you're worried about the /etc/conf.d/net syntax then read /usr/share/doc/openrc/net.example to ensure that you've got the configuration and syntax correct first. Most of it seems fairly similar to my eyes.

[risa2000]

After reading the migration guide, I wonder what is the reason behind decision to encode names and versions into parameter names. E.g:

Code: Select all

# For 2.6.23-gentoo-r5, pass video_br=2 to cx88_dvb
module_cx88_dvb_args_2_6_23_gentoo_r5="video_br=2"
# For 2.6.x series kernels, always pass vendor and product
module_usbserial_args_2_6="vendor=0x1410 product=0x2110"
# Always pass debug to ieee1394
module_ieee1394_args="debug" 

I see module name mixed with uname (where some characters like dots and hyphen are converted to underscore).
Why not use some hierarchical structure (like INI file) instead?

All files in /etc/conf.d/ have to be shell parseable. The .INI style is not.

Hmm, I thought that the point was moving from shell scripts to something else. Anyway, INI is not parsable by shell, but it is parsable by Python, would it help? On gentoo I guess Python is system requirement.

Also I believe the module name and its (optional) parameters should not be divided into two lexically separated information. Either I give the module (and eve2ntually its parameters) or I do not. Or is it foreseen that it will be possible to provide parameters for modules, without explicitly specifying the modules?2

It's easiest to do it like it is, otherwise we would have to use something like bash arrays, which is now a big no no for OpenRC scripts which should work on any POSIX shell.

I got the objection about INI or shell requirement. I hope however the configuration file could be (and will be) changed later (or rather sooner). With the syntax you propose I have following problems: readability (experienced the same stuff already in /etc/conf.d/net), lost of the context (i.e. modules defined in one place, parameters for them in other) and finally the fact that mapping from uname to variable name is not one to one (i.e. you cannot get uname from variable name, unless you try to guess.).

[UberLord]

During shutdown I get 'start-stop-daemon: fopen `/var/run/gdm.pid': No such file or directory'

That's an issue with gdm - it's always behaved like that.
start-stop-daemon is just being a little more verbose, that's all.

[tallica]

That's an issue with gdm - it's always behaved like that.
start-stop-daemon is just being a little more verbose, that's all.

OK, I knew it's not critical thanks.

[Januszzz]

I would file a bug asking for that

Well, it is reported since mid 2007:

https://bugs.gentoo.org/show_bug.cgi?id=182094

Its not hard to use ebuild from bugzilla, like I did sth about January. But I had some issues with that (I found on gentoo-embedded a man who tweaked the same issues) so I guess putting it in ~arch would be nice move, as more people would have chance to fix it (the issue was about useflags not being properly translated to configure options).

[trad511]

Have trolled the forums for the following but can't say I've seen a solution to the problem. After installing baselayout-2 and openrc I get everything to come up except for my dmcrypt filesystems. The problem seems related to / being mounted read-only at the time dmcrypt systems are being mounted. I also have an interesting LVM error. This looks to me like an initialization order problem, but I'm at a loss right now. Any advice would be helpful.

This is not a genkernel. This system was booting ok prior to upgrade. The following snips are in order of boot. Seems to me that dmcrypt is coming in slightly early in the boot process.


LVM error:

Locking type 1 initialisation failed.

First dmcrypt problem:

* dm-crypt map crypt_swap ...
* cryptsetup will be called with : -c aes -h sha1 -d /dev/urandom create crypt_swap /dev/sda2
[ ok ]
* Running pre_mount commands for crypt_swap ...
[ ok ]
mkdir: cannot create directory `/mnt/remdev.2185': Read-only file system

Second dmcrypt problem:

My home directory is encrypted and this doesn't even come up as an option any more. The only thing I see is the following error:

* Remounting root filesystem read/write ...
[ ok ]
* Updating /etc/mtab ...
[ ok ]
* Mounting local filesystems ...
mount: special device /dev/mapper/crypt_home does not exist
* Some local filesystem failed to mount
[ !! ]

here's my rc-status boot (which looks good to me):

rc-status boot
Runlevel: boot
hwclock [ started ]
modules [ started ]
lvm [ started ]
device-mapper [ started ]
dmcrypt [ started ]
fsck [ started ]
root [ started ]
mtab [ started ]
localmount [ started ]
hostname [ started ]
sysctl [ started ]
bootmisc [ started ]
termencoding [ started ]
consolefont [ started ]
keymaps [ started ]
urandom [ started ]
net.lo [ started ]
alsasound [ started ]
procfs [ started ]
swap [ started ]

Currently I have rc_hotplug="NO" and rc_coldplug="YES", but have tried rc_hotplug="YES" to no effect.

Thanks Uberlord! The code is beautiful and the speed is blazing!

[UberLord]

You need to get dmcrypt to run after root. Try putting this in /etc/conf.d/dmcrypt

rc_need="root"

Notice that rc-status sorts the services in the order they would be started, which means fixing it easier

[Frustie]

and while updating my production system to update world

Ain't you brave. Your running an ~arch Production Server? You running emerge <anything> without checking what is going to happen first? Don't sound like a production system at all, more like a "non-production system".

well yes, i do check , hence the -av i use, but i didnt notice/expect it to be so soon anyway, tonight ill see what happens and schedule some maintenance time

[energyman76b]

yes, I am moody. Wouldn't you moody too, if:
-boot blows up in your face
-you loose some config files that contained some important stuff
-there is a lack of documentation
and
all this trouble for no benefit?

As I pointed out above, your running unstable if your using these (i.e. ARCH="~x86" or similar). Why are you therefore surprised that things aren't quite perfect? If thats what your after don't run unstable, its very simple.

All you've done (with one expcetion) is post complaints, rather than providing feedback and details of the problems/errors your encountering which would help resolve them.

Your current situation is mainly a consequence of your choice of keywords, you've no one to "blame" but yourself for that.

what about 'net was nuked without asking, that is wrong' is not feedback?
When my box does not boot because of bugs well known since October, and I even post the corresponding bugzilla links, I am trolling? That is not 'feedback' or 'problem reporting'?
Complaining is feedback. Ok? I am complaining about problems. That is feedback. I could also post 'hey, everything is superdandy' but it is not, it would be a lie. That would not be feedback - that would be brown nosing.

If boot does go wrong because fsck says 'Could not check some filesystems' what do you else want?

Could you stop your fanboy-ism?

And about speed:
rc_parallel="YES"
does not boot faster than
rc_paralled="yes"
does not boot faster than the old:
RC_PARALLEL_STARTUP="yes"

[bssteph]

petition for help, maybe similar to the dmcrypt answer above me: openrc and EVMS: http://forums.gentoo.org/viewtopic-t-688383.html

[UberLord]

I know nothing about EVMS, sorry.

[slackline]

well yes, i do check , hence the -av i use, but i didnt notice/expect it to be so soon anyway, tonight ill see what happens and schedule some maintenance time

Code: Select all

EMERGE_DEFAULT_OPTS="--ask --verbose"

Just to save three key-strokes

[12345moon12345]

currently I've got several mingetty root autologins in my inittab.
Then I have this as .bash_profile:
if [ `tty` = "/dev/tty1" ]; then
startx
fi
It's already fast but I suspect it could even be faster if there would be an initscript that directly starts X, without using bash and maybe without using mingetty?

Off course there is XDM but I don't want to login this way (I'm the only user of this PC in my own house.

Any thoughts?
Could this be added default to openrc?

bye,

Victor

[UberLord]

Hmm, I thought that the point was moving from shell scripts to something else. Anyway, INI is not parsable by shell, but it is parsable by Python, would it help? On gentoo I guess Python is system requirement.

OpenRC works just fine on base NetBSD and FreeBSD systems also. Python isn't there
Neither is perl nor ruby. Also, python exists in /usr and OpenRC requires nearly everything it really needs in /

I got the objection about INI or shell requirement. I hope however the configuration file could be (and will be) changed later (or rather sooner). With the syntax you propose I have following problems: readability (experienced the same stuff already in /etc/conf.d/net), lost of the context (i.e. modules defined in one place, parameters for them in other) and finally the fact that mapping from uname to variable name is not one to one (i.e. you cannot get uname from variable name, unless you try to guess.).

Come up with something better that is shell parseable then

[UberLord]

If boot does go wrong because fsck says 'Could not check some filesystems' what do you else want?

I would expect you to have a look and the conf file for fsck.
You can set fsck_args to -a and your reiser4 problem is "solved."

Could you stop your fanboy-ism?

And about speed:
rc_parallel="YES"
does not boot faster than
rc_paralled="yes"
does not boot faster than the old:
RC_PARALLEL_STARTUP="yes"

Depends on what baselayout version your comparing to. Old baselayout-1.13's and 2.0_rc's had a fair chunk of C code in them to work out the dependencies, which was the real slowness.
Maybe you're just unlucky, but as I said earlier I really don't care.

[UberLord]

currently I've got several mingetty root autologins in my inittab.
Then I have this as .bash_profile:
if [ `tty` = "/dev/tty1" ]; then
startx
fi
It's already fast but I suspect it could even be faster if there would be an initscript that directly starts X, without using bash and maybe without using mingetty?

Off course there is XDM but I don't want to login this way (I'm the only user of this PC in my own house.

Any thoughts?

You could try making a custom init script to start X on your chosen VT early enough.

Could this be added default to openrc?

No, as it's a bad idea for a whole load of reasons
But if it works for you then more power to you

[slackline]

currently I've got several mingetty root autologins in my inittab.
Any thoughts?

Well its generally a bad idea to login as root, far better practice to do so as user then su to root (or use sudo).

[golding]

Very pretty app! Also very well coded and there for the taking, without any really bad problems.

A few small points, though.
1) The migration page does not spend any time explaining what it means about rc and rc.conf.
2) There is a major bug with resolvconf. (What is 'uniqify'?)
3) There is a change in the way conf.d/net is written.

[jaBo]

My linux is broken, cos of my own stupidity. I have emerged baselayout-2 and all my scripts were lost.
Some how also my wired network is broken, I cant get the correct ip, tryed downgrading dhcpcd and tryed other version of wpa_supplicant.
I only got more confussed then I was already.

So I tried to keep baselayout-2 and then add up openrc following the online guide.I am forced to download openrc from windows and than emerge it. emerging openrc does not start, its complaining about that I dont have baselayout-2 which is a dependency of openrc.

I will try to download fresh portage. Maybe this will somehow persuade openrc that I have baselayout-2 already emerged.

I dont expect any fast solutions from you guys, but I have to ask you if there is a way, to fix my mess ?

no network --> no intenet --> no sync
And also mounting fails ... ... .... I have read-only status. Can not edit confing files.

Somehow I wish not to do everything from begining. I lost lots of time just to make my sound intel_hda work. (was like 3 months without sound)

But lesson learned.

[Biloute]

It's already fast but I suspect it could even be faster if there would be an initscript that directly starts X, without using bash and maybe without using mingetty?

Off course there is XDM but I don't want to login this way (I'm the only user of this PC in my own house.

Any thoughts?

The solution is in gentoo-wiki : http://gentoo-wiki.com/TIP_Passwordless_Login

[easy target]

Hi,

I can't run noip on system start. 'Starting noip' isn't shown during boot and here is what I do after boot:

Code: Select all

localhost ~ # pidof noip2

localhost ~ # rc-update | grep noip
                 noip |      default                 
localhost ~ # rc-status -a | grep noip
localhost ~ # rc-update add noip default
 * rc-update: noip already installed in runlevel `default'; skipping
localhost ~ # rc --service noip start
 * Starting noip ...                                                      [ ok ]
localhost ~ # pidof noip2 
3551
localhost ~ # rc --service noip start
 * WARNING: noip has already been started
localhost ~ # rc-status -a | grep noip
localhost ~ #

As You can see, noip is in rc-update in runlevel default, but it doesn't start.

[Martux]

I have an issue with xinit not starting kde (3.5.9).
I have set the following:

Code: Select all

ls -l /etc/runlevels/default/
xdm -> /etc/init.d/xdm

/etc/env.d/90xsession

Code: Select all

XSESSION="kde-3.5"

I also have automatic login for my user enabled in KDE and used to have an /root/.xinitrc with exec startkde in it, but it also does not work with startkde or even without any xinitrc...
The xdm service does not even start, local service is the last thing to start up, after that i am directly on the login prompt.
That's all information i'm aware of, any help is appreciated.
Marcus

[UberLord]

Very pretty app! Also very well coded and there for the taking, without any really bad problems.

A few small points, though.
1) The migration page does not spend any time explaining what it means about rc and rc.conf.
2) There is a major bug with resolvconf. (What is 'uniqify'?)
3) There is a change in the way conf.d/net is written.

resolvconf works fine. resolvconf-gentoo is dead, openresolv is the way forward.
Not sure what you mean about the other points

[UberLord]

I will try to download fresh portage. Maybe this will somehow persuade openrc that I have baselayout-2 already emerged.

You need baselayout-2.0.0 - any of the rc's just won't cut it.

[UberLord]

I can't run noip on system start. 'Starting noip' isn't shown during boot and here is what I do after boot:

Post a bug at bugs.gentoo.org and hopefully a plucky Gentoo dev can fix it for you.