Gentoo On systemd, Torvalds On Bullshit, udev Forked
As a followup to my previous reservations about systemd and the means of its introduction, there is a very informative discussion happening on Gentoo Forums, including design flaws of systemd, the consolekit problem, and the expenses of policykit. I think the Gentoo community and philosophy really shines there:
“…it’s because everything he comes out with wants to take over our machines, with a mess of so-called “integration” requiring changes across the board. Til he finally realises what everyone was on about, and drops the project for his next shiny adventure, leaving everyone else to pick up the pieces.” – steveL
Elsewhere, Linus Torvalds is heard to “call bullshit” (and more!) on Kay Sievers’s excuses and poor maintenance. It’s nice to see Linus is aware of what’s going on.
Further, a udev fork is already underway and gaining popularity, which is good due to the aforementioned problems with udev maintenance and mission creep. Hopefully this will grow into a strong and independent replacement.
At the very least, it’s great to see there is a population of Linux users, admins and devs who are aware of the implications of these changes, and that they’re already exploring alternatives.
Updated reading:
Earlier reading:
28 Comments
Sorry, the comment form is closed at this time.
IgnorantGuru's Blog 
Thanks for catching this! I’ve been busy lately and haven’t had the time to read the Gentoo forums. I think it’s great that the Gentoo community (and Linus) are aware of what’s going on and are meeting it with a sceptical eye. All new ideas deserve to be fully scrutinized before jumping in blindly.
Linus has another choice post
http://article.gmane.org/gmane.linux.kernel/1368617
Where he calls Lennart and Kai, “”Two-faced lying weasel” would be the most polite thing I could say. But it almost certainly will involve a lot of cursing”
It really gives me no confidence in gnome/redhat developers, especially any gnomish “plumbers”.
I recently had a problem with the horror that is udisks and udisks2. I first had it under Arch, if udisks and udisks2 are both installed(which can’t be helped if one wants to test gnome 3 or new xfce and also use KDE) then when I mount a usb drive of anytype I get random intermittent interference on the usb bus, delaying my mouse and keyboard. If I have only one udisks version installed this interference drop significantly, but unfortunately is still there. Now I see the same exact issue on a different machine with the new Slackware 14, because I installed both XFCE and KDE. So I ripped both udisks out, as I actually find I prefer to use pmount, even if udisks didn’t have such horrible bugs.
Thanks – great link. I like how Linus tells it like it is! “FIX UDEV ALREADY, DAMMIT”
Yes, I would say these Red Hat people are tearing Linux apart and turning it into a convoluted and buggy thing. People seem to go along with this garbage in their quest for the latest bells and whistles (even if they don’t work or break other things).
pmount is good simple solution. I also suggest giving udevil a try which is much more flexible and capable overall. udisks should never have been written as a daemon – unnecessary. Having all that complex code running as root isn’t a good idea either (I have seen and worked with the udisks sources and I recommend avoiding it).
I just threw on pmount because I didn’t see a slackbuild for udevil when 14 was first released, though there is one now. I used udevil on my Arch machine and pretty much every userspace mounter, just to see if any of them had the same problem as the newer udisks and udisks2, but none had the same problem. I really thought at first it might be a subtle bug with userspace mounting and my specific hardware. Now with a second machine with newer udisks and the problem and no issues with other mounters, it’s clear the problem lies with the newer udisks and udisks2.
That whole thread is wonderful, I just found some choice words by Al Viro as well
“That, or just adding usr/udev in the kernel git tree and telling the
vertical integrators to go kiss a lamprey…”
“Looks sane. TBH, I’d still prefer to see udev forcibly taken over and put into
usr/udev in kernel tree – I don’t trust that crowd at all and the fewer
critical userland bits they can play leverage games with, the safer we are.
Al, that -><- close to volunteering for maintaining that FPOS kernel-side…"
As Linus said, putting udev in the kernel only seems like a perfect solution because the kernel maintainers actually maintain their stuff. But dumping everything on the kernel devs is a slippery slope. I’m not sure where udev really belongs from a technical standpoint though.
These RH devs should never have been maintaining udev in the first place, given their record, and it should be taken away from them, as Linus suggested, given their poor maintenance of it. Sadly, I suspect nothing like that will happen – for some reason these RH devs seem to have tremendous influence. I don’t know if there’s something analogous to political corruption going on (which frankly I suspect) or whether, being paid, they just have time to produce tantalizing solutions (which they then don’t maintain properly). At any rate, I don’t view it as mere incompetence; I think there’s a broken agenda at work, or an agenda which includes breaking non-RH?
Yeah that thread was great.
Yeah it’s clear Al doesn’t really want to put udev in kernel, but he makes it clear that it’s not just run off the mill bad maintainers, but maintainers who want to use it to leverage changes in both kernel and userspace. Al points to the real problem, which is really less about incompetent maintenance and more about the maintainers creating “problems” to then force their agenda into both kernel and userspace.
Wow I’ll have to read the rest of that – sounds spot-on. That’s very dirty pool and if they allow that kind of thing to continue (as they seem to be) it’s going to have serious consequences for almost every distro (and already is, obviously).
My hunch is that this is all related to locking down Linux with DRM and related technologies that don’t let you control your own system. Linux is just too free for corporate interests (especially media companies, etc, and remember that big money is involved there). This is already seen in use of hal to enforce DRM in Flash on Linux. hal being deprecated I think they’re trying to move deeper into CK and PK related stuff for this, and never mind what they break on the way.
So how much are these RH devs being bribed to pollute Linux in this way?? I doubt we can meet the offer so the solution is to watch them very carefully and lock them out of this influence. And users, admins, and devs should do everything to avoid solutions offered under these terms.
I think it is probably somewhat more gnomish than pure redhatish. There is a very big interest in embedded car systems, embedded plane entertainment systems, I know Lennart works with many of these companies and that they seem really really interested in systemd, and my guess is these industries want systems that can be pretty tightly nailed down, as well as other things. Maybe this is why gnome is so focused on touch interfaces.
Redhat desktop team itself really wants plug and pray style multiseat, and desktop virtualization with full mutimedia, that works well enough so enterprise wants it, breaks just enough and is convoluted enough when it breaks that they will buy support for desktop. Redhat desktop team is not very profitable, so they want to be important too.
You’ll notice Mauro Carvalho Chehab who is a kernel driver programmer for Redhat and doesn’t really agree with Kay, so I think some aspects are less Redhat itself and more Gnome/Redhat desktop.
I’ll just throw this in here too – Linus knows his stuff.
I think not breaking interfaces is paramount. That and not using closed binary formats etc.
One step forward, two steps back. The “we know better” mentality seems to be pervasive in all projects that get big enough, and it isn’t just the developers fault but also the users who expect the developers to make all the decisions for them. Unfortunately I think that it is a lot because of the success of distributions that cater to a less tech-savvy audience. It is good to have more users, but the idea of complaining about something that should be known as entirely subjective and therefore trivial like defaults is sad. Not having a specified linux base for proprietary software to cater to means that you _can_ make your own choices from the ground up and not just follow what someone else already decided for you. The post by Miguel de Icaza is especially frightening. The notion of a unified linux stack is an unfortunate one that I hear from newbies, but to hear it from someone as seasoned as him is concerning. Even as I say positive things about systemd in an environment mostly hostile to it, I don’t hope for it to be a forced standard, but perhaps a de-facto standard due to people liking it more than the alternatives. Choosing something else might be less easy, but should never be thought of as less _right_. That’s a powerful idea, and something we need more of right now.
On the udev maintenance mess: If something you do breaks something, unless what you see is a bug not just an inconvenience, you take responsibility for it. Heck, even if it isn’t your bug then it’s something that you should still deal with yourself because the priority should be your stuff working properly until the real fix is out.
Well said. I agree. It seems that Miguel and others are looking to create a unified, monolithic beast that’s just like Windows or OS X, but with GNU/Linux under the hood. That’s a big problem in the free software world, because as more applications require systemd, GNOME, udev, or whatever else that gets added into the monolith, GNU/Linux as a platform will become more and more homogenous. That’s great for business, but bad for the software ecosystem.
IMO, business and freedom are at odds with each other and aren’t compatible. The decisions and ideas being presented by these recent “celebrities” are in the interest of business, and little more.
Also on the topic of breaking interfaces, the gtk3 port is nearing completion. As far as I can see from my testing, it only needs 3 things at this point:
1. It only builds with –disable-desktop-integration, that needs to be finished
2. There is an odd bug where the first right click causes the icons to shift temporarily. Subsequent clicks don’t do it.
3. It needs a configure switch. Currently the only way to build it with gtk3 support is to manually edit the configure script.
I’ll work on 1 & 2, but I don’t know how to do a configure switch properly.
Code is here:
https://github.com/BwackNinja/spacefm/tree/next-gtk3
> on the topic of breaking interfaces, the gtk3 port is nearing completion
lol
I’m excited – looks like you’ve made some great progress. I look forward to seeing it in gtk3. I can add the configure switch as I merge it in, but I won’t get to this for a bit. Still finalizing 0.8.1.
My experience is a little bit special and gives me maybe slightly unusual viewpoint. I used to be a Unix programmer, then abandoned this for almost 20 years. And now I’m back to Linux, however it is more a hobby to me at this point. And what I saw upon my return is somewhat a mess. Not totally, of course, but in so many places. Somwhere I’d call it a terrible mess, somewhat it is quite bearable. I tried out “intergrated solutions”/DE and got rid of them asap. I felt as if I am running M$ Win – no vaguest idea what’s happening inside. And unfortunately, it’s a clearly visible tendency.
I bought my notebook with Gnome-based SUSE Enterprise pre-installed and was shocked. So slooowww, so inconvenient. Small detail: Being switched on, it connected to my Linux home server via Samba immediately. Then I spend nearly an hour to get it connected to another Linux box via NFS!!! HP’s support on Linux ends up with statement “WE RECOMMEND WINDOWS 7”. Full stop. I felt as if SUSE was saying “Sorry I’m not Windows, but I’m trying so hard”! That’s a corporate point of view on Linux. No need to say I wiped it off the same day.
After half a year on ArchBang I switched to Calculate/Gentoo getting tired to thrill every time I do the system upgrade. Now I’m quite satisfied with the distro. The only disappointing thing to me is that the famous statement of Gentoo to be well-documented appeared to be a myth. I could not find any sane description of so many things! Not only concerning a Gentoo, but on common Linux stuff, too. I spend few days with google trying to understand something about consolekit/policykit stuff but found only a hundred copies of the same short and non-descriptive text. I tried to understand if I really need this stuff and what it does exactly but gave it up finally. Now I have both on my machine but still don’t know why and dreaming one beautiful day to throw it out. And my feeling is it’s done on purpose. Yes, developers too often underestimate the value of documentation; yes, they may be just lazy or prefere to do something with the code instead of writing docs, but the feeling is still there. Gates are forced to close. No, Gates is forcing to close te gates. That’s it.
Well said and you describe the corporate end of Linux well. I too first started with SUSE and eventually dropped it (and KDE) as it was becoming far to ‘Windows’ for me. At the other end of the spectrum there are a lot of cool light ‘n fast Linux tools. The problem are these increasing core components that attempt to drag in de facto DE components.
It is possible to dump CK/PK/udisks with udevil+spacefm. Though this is one of the few file managers you’ll be able to run, it does open the possibility for a lean and fast system which is more sane to configure (or will keep you more sane ;)
I’ve been looking over Sabayon Gentoo – I may give it a try. I’ve been using Aptosid for quite awhile now with good results, but it is debian so while well maintained it’s not exactly light and flexible without hacking.
I tried Sabayon before I run onto Calculate and found it surprisingly slow. Arch isn’t any faster than Ubuntu (phoronix compared, not me. I never used Ubuntu :), but Sabayon is even slower (I’ve AMD E-450 APU therefore easily notice how fast it is running). I was surprised how the guys managed to make so slow Gentoo derivative. I’d suggest to give Calculate a try. Fast, sane distro, precompiled, geared toward a corporate use but still sane and providing an environmental choice (including xfce or even scratch distros I actually used (bare X + openbox)). Intersting add-on is template-based configs that allows to easily maintain lots of similar workplaces. I personally don’t need this feature, so I used Calculate more as a pre-compiled Gentoo and almost migrated to a pure Gentoo by now. It is much more solid than bleeding-edge Arch and is blazingly fast. And USE flags and slots are something!
And is there any guidelines available how to dump CK/PK? As I said I couldn’t get any vision how it works and what really does. Will be enough to just cut them off the system and properly configure udevil, ot there is somthing I’m not aware of?
Thanks for the info! Hadn’t seen Calculate. I actually didn’t mind bare gentoo – had X, Openbox, and Firefox running in a day. Overall I liked it – sort of like Arch. Where I ran into trouble was with my Brother printer driver. But I may give it another go. I run a pretty simple system in general so its fun setting it up – just time consuming.
> And is there any guidelines available how to dump CK/PK?
This probably varies with distro, and I haven’t gotten into the details yet due to keeping udisks for debugging for now (it’s not getting installed on my new gentoo setup). Generally udisks (not udev) requires CK/PK. So without udisks they should simply be uninstalled I think. Some other apps (like nm-applet) use CK too though, so software selection is critical. For apps/daemons that don’t depend on CK/PK, I don’t observe any impact. I know a number of spacefm+udevil users have successfully dumped the lot (udisks, CK, PK, devicekit, etc).
Also, even if you can’t uninstall CK/PK due to other apps that use them, freeing your file manager and user mount solution from the headaches of CK/PK is worth it. This simply requires using tools that don’t use them, even if they sit on the system for other apps to use.
Dumping CK and PK requires you to look at everything that depends on them. Most of gnome-control-center depends on it, but you can patch those parts out. To shutdown/reboot as a user, you need to make sure that shutdown, poweroff, halt, reboot, etc are set to SUID root, and regular graphical methods of shutting down won’t work, so you it is easiest to have shortcuts to those on a panel or menu. Gvfs depends on libgdu, which in turn depends on PK. Without gvfs, you can’t mount drives from file-open dialogs. Bluetooth support (bluez) is usually built with PK support, but works without it (though it needs a rebuild) and I believe it also needs a udev rule for /dev/rfkill to be mode at least 664 or 666 to enable and disable bluetooth. That’s about all I can think of off the top of my head.
Thanks! This is enough for me to give it a try in a meanwhile. I don’t have much Gnome stuff installed (gksu and probably few other things, and none of KDE) — I’m trying to keep my system as simple as possible (things are easily getting complicated by themselves:) And I consider NM too buggy and unpredictable to be used – so far I’m stuck with wicd. I don’t even keep it running all the time – only to connect to Wi-Fi when I’m outside.
The initial step in Gentoo is as easy as to put minus sign in front of corresponding USE flags, then wait few hours until everything related is recompiled… This will remove all dependencies on CK/PK that can be removed. If they can’t, I’ll have to probably trade one package for another… Okay, we’ll see…
Don’t try Sabayon with existing (important) data. When I did it, it detected my ext4 partitions and locked them in an LVM VolumeGroup without telling me that it had done it and the docs didn’t outline how to reverse it. Thankfully I had a backup, but that should really be documented by the Sabayon guys.
Aside from that, it worked. But for some reason, the installation process wasn’t as well covered as Gentoo’s is, which ultimately led me to Gentoo.
See this message:
From: Richard Yao <ryao gentoo.org>
Subject: With regard to udev stabilization
Newsgroups: gmane.linux.gentoo.project
Date: 2012-11-13 02:40:53 GMT
Dear Everyone,
It is no secret that many of us are unhappy with the direction that udev
has taken under the leadership of the systemd developers. That includes
Linus Torvalds, who is ‘leery of the fact that the udev maintenance
seems to have gone into some “crazy mode” where they have made changes
that were known to be problematic, and are pure and utter stupidity.’
https://lkml.org/lkml/2012/10/2/505
After speaking with several other Gentoo developers that share Linus’
concerns, I have decided to form a team to fork udev. Our plan is to
eliminate the separate /usr requirement from our fork, among other
things. We will announce the project later this week.
I understand that the council is scheduled to vote on a topic related to
udev stabilization. Would it be possible to delay the vote for another
month so that we have time to get organized?
Yours truly,
Richard Yao
Full conversation here:
http://thread.gmane.org/gmane.linux.gentoo.project/2262
Excellent to see! Let’s hope Linus has the wisdom to support this and take the kernel+udev in a new direction, AWAY from the grubby hands of Red Hat. Good to see the support the idea is getting there. But these political forces have a lot of say in Linux, somehow.
They should call it “udev” – steal the name just like Red Hat continuously does. Since when did Red Hat come to control most of Linux?
So after my jump to Ubuntu for several years i am back on gentoo at last (have now a quicker system to reduce compiling time). But… I choose to install gnome 3.8…. perhaps to quick?
systemd cancer taking over my magic sysrq configuration:
I enable all magic sysrq keys in my /etc/sysctl.conf, yet systemd
overrides this with /usr/lib/sysctl.d/50-default.conf and disables
some!
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=725422
This is finally being actioned in the above bug, took them 4 months to agree and ‘fix’ it, and as far as I can tell its not actually released – looks like its waiting for the next packaging to happen??
Venting here as I’ve just successfully fought a shockingly incompetent bug where sysctl.conf was suddently being completely ignored (https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=732920), and even after the OP posted the actual problem it took 2 months to fix, and its still not in Testing!!
My opinion of Debian is starting to fall now with the shit I’ve been fighting recently. I’m supposed to trust it as a competent hardcore OS.