View previous topic :: View next topic |
Author |
Message |
__bjoern n00b
Joined: 12 Oct 2014 Posts: 28
|
Posted: Tue Sep 08, 2015 1:21 pm Post subject: |
|
|
To be fair, there was no kernel panic, because PID 1 didn't crash. But what good does a system do you, if init goes ape sh*t to the point were you can't even login to it anymore? You have to reboot anyway, so it might as well be a kernel panic. |
|
Back to top |
|
|
Anon-E-moose Watchman
Joined: 23 May 2008 Posts: 6098 Location: Dallas area
|
Posted: Tue Sep 08, 2015 1:27 pm Post subject: |
|
|
__bjoern wrote: | But what good does a system do you, if init goes ape sh*t to the point were you can't even login to it anymore? You have to reboot anyway, so it might as well be a kernel panic. |
++ _________________ PRIME x570-pro, 3700x, 6.1 zen kernel
gcc 13, profile 17.0 (custom bare multilib), openrc, wayland |
|
Back to top |
|
|
Naib Watchman
Joined: 21 May 2004 Posts: 6051 Location: Removed by Neddy
|
Posted: Tue Sep 08, 2015 1:28 pm Post subject: |
|
|
it was a pid1 panic and pid1 shouldn't panic because that will bring down the system (hence it being simple) _________________
Quote: | Removed by Chiitoo |
|
|
Back to top |
|
|
gwr Apprentice
Joined: 19 Nov 2014 Posts: 194
|
Posted: Tue Sep 08, 2015 1:45 pm Post subject: |
|
|
steveL wrote: |
Some idiot "developer" doesn't; so what? He doesn't grok shell either.
|
We are not his intended audience. He just has to convince the thousands of Windows admins who are intimidated by the command line and the learning it requires. He's playing into the hands of the people who want glory without effort. |
|
Back to top |
|
|
Sulman n00b
Joined: 15 Feb 2014 Posts: 63
|
Posted: Tue Sep 08, 2015 1:58 pm Post subject: |
|
|
I'm well aware perfection is unattainable in software, but there's a certain attitude from the very top in the systemd camp that has bothered me from the start, and it is no better illustrated by that crash.
People should care about these things, and they do not.
I'm grateful for Torvald's uncompromising attitude to quality. |
|
Back to top |
|
|
depontius Advocate
Joined: 05 May 2004 Posts: 3509
|
Posted: Tue Sep 08, 2015 3:40 pm Post subject: |
|
|
__bjoern wrote: | I tried pressing Ctrl+Alt+Fn (where n is 1,2,3,...) but it didn't work. The only thing i could do at this point was to press the reset button. |
The Magic SysRq keystrokes are your friend here. If there's any life at all left in the kernel, you can get your disks sync'ed and remounted RO. _________________ .sigs waste space and bandwidth |
|
Back to top |
|
|
__bjoern n00b
Joined: 12 Oct 2014 Posts: 28
|
Posted: Tue Sep 08, 2015 3:46 pm Post subject: |
|
|
Yeah i know of the magic sysrq keys but to be honest, i'm to lazy to type all that, pressing the reset button on my PC is faster. Thankfully i hadn't any unsaved data, my journaling filesystem took care of the rest. |
|
Back to top |
|
|
steveL Watchman
Joined: 13 Sep 2006 Posts: 5153 Location: The Peanut Gallery
|
Posted: Tue Sep 08, 2015 6:19 pm Post subject: |
|
|
__bjoern wrote: | So, today is was doing my system updates and noticed that it had installed a dbus update. Now, when i notice that a system service has been updated i usually go ahead and do a "systemctl restart <system-service>". I typed "systemctl restart dbus".
As soon as i pressed Enter, my system was having a seizure.
..The only thing i could do at this point was to press the reset button.
Some of you might think that this is to be expected, because systemd relies on dbus as a transport, but i mean come on. I would have expected that systemd has some sort of mechanism in place to cope with the dbus daemon being restarted. |
Well, yeah, especially if it just restarted it, itself.
As the other guy said: Holger Winkelmann wrote: | I always thought systemd uses dbus messaging, but does not rely or depend on the DBUS daemon? | so as ever, the sizzle we were sold, turns out to be a sausage full of crap.
Quote: | After my computer restarted i decided to look more into this and just killed the dbus daemon. My computer crashed again, as expected.
I thought "In all this years none of the systemd developers tested restarting the freaking dbus daemon?! How shitty is that?". Turned out i was wrong, because a quick google search revealed this (take it from Poettering itself): http://lists.freedesktop.org/archives/systemd-devel/2013-January/008456.html
My takeaway from this E-Mail is: "Well, we know that the D-Bus protocol is utter crap, so we just shove this into the kernel, because when the kernel crashes, you're fscked anyway. But hey, i've never seen dbus crash, so it's not a problem at all". |
Lul; the part I found funny was: Lennart Poettering wrote: | Thankfully, D-Bus is rock solid these days and has been for a while.
The best way to get it to crash is probably by explicitly sending SIGSEGV to it... |
Lol; yeah or to restart it using the "bestest init-manager of all" from the same team.
Quote: | I'm not really sure what he means with "sane getty". I can assure you, i've seen no "sane getty" on my system when it crashed.
This obviously has also nasty security implications. If somebody manages to crash the system dbus daemon, the whole system will fall over. Great. |
Indeed; it's what some of us were "harping on" about a couple of years ago, and since: that dbus was fine in the desktop session, though nowhere near as useful, nor as performant, as dcop, but it's a real mistake to turn such a crappy design into a "system bus", running as root (ffs.)
It just displays a woeful ignorance both of what a kernel does, in a multiprocessing system, and the POSIX APIs tuned for particular uses.
Nothing new there, ofc.
Quote: | I have to say this is pathetic. I'm really disappointed right now. |
The only thing I'd say, is: don't let this Winbloze-type experience put you off the really good work that so many people have done, over the decades, to get us where we are now.
Blame Lennux and deliberate marketing-driven groupthink, not Linux. :-) |
|
Back to top |
|
|
__bjoern n00b
Joined: 12 Oct 2014 Posts: 28
|
Posted: Tue Sep 08, 2015 6:31 pm Post subject: |
|
|
steveL wrote: |
The only thing I'd say, is: don't let this Winbloze-type experience put you off the really good work that so many people have done, over the decades, to get us where we are now.
Blame Lennux and deliberate marketing-driven groupthink, not Linux. |
Don't worry, i will continue to use Linux . I wrote this post this morning in a bit of a rant-y fashion, because it was so mind boggling that such a flaw exists in a software so widely used. |
|
Back to top |
|
|
ArneBab Guru
Joined: 24 Jan 2006 Posts: 429 Location: Graben-Neudorf, Germany
|
Posted: Fri Sep 18, 2015 10:10 am Post subject: |
|
|
__bjoern wrote: | So, today is was doing my system updates and noticed that it had installed a dbus update. Now, when i notice that a system service has been updated i usually go ahead and do a "systemctl restart <system-service>". I typed "systemctl restart dbus".
As soon as i pressed Enter, my system was having a seizure. |
From my experience with KDE using dbus, that fits what I would have expected to happen. That's one of the reasons why I never wanted systemd on my systems. _________________ Being unpolitical means being political without realizing it. - Arne Babenhauserheide ( http://draketo.de )
pkgcore: So fast that it feels unreal - by doing only what is needed. |
|
Back to top |
|
|
orlfman n00b
Joined: 31 Jul 2006 Posts: 68
|
Posted: Tue Nov 10, 2015 9:41 pm Post subject: |
|
|
__bjoern wrote: | Hi Guys,
so i've been following the systemd disscussion on this forum for a rather long time now, but never wrote anything, because i had'nt really anything useful to say. But today i stumbled upon something that is so crap, that i want to bring it up here, for everyone to laugh about.
So, today is was doing my system updates and noticed that it had installed a dbus update. Now, when i notice that a system service has been updated i usually go ahead and do a "systemctl restart <system-service>". I typed "systemctl restart dbus".
As soon as i pressed Enter, my system was having a seizure. The graphical session just outright crashed and i was staring at a black screen. After about a minute my display manager came back, as if nothing happened. But it went away again after a few seconds, without me doing anything.
I tried pressing Ctrl+Alt+Fn (where n is 1,2,3,...) but it didn't work. The only thing i could do at this point was to press the reset button.
Some of you might think that this is to be expected, because systemd relies on dbus as a transport, but i mean come on. I would have expected that systemd has some sort of mechanism in place to cope with the dbus daemon being restarted.
After my computer restarted i decided to look more into this and just killed the dbus daemon. My computer crashed again, as expected.
I thought "In all this years none of the systemd developers tested restarting the freaking dbus daemon?! How shitty is that?". Turned out i was wrong, because a quick google search revealed this (take it from Poettering itself): http://lists.freedesktop.org/archives/systemd-devel/2013-January/008456.html
My takeaway from this E-Mail is: "Well, we know that the D-Bus protocol is utter crap, so we just shove this into the kernel, because when the kernel crashes, you're fscked anyway. But hey, i've never seen dbus crash, so it's not a problem at all".
I'm not really sure what he means with "sane getty". I can assure you, i've seen no "sane getty" on my system when it crashed.
This obviously has also nasty security implications. If somebody manages to crash the system dbus daemon, the whole system will fall over. Great.
I have to say this is pathetic. I'm really disappointed right now. |
pottering has a lot of comments like that.
take for example on a systemd system one cannot have /var as a separate partition and have it umount cleanly. stick it on another drive and it will always fail to umount when restarting. this is because of journald. when brought to his attention he said he knows about it, but isn't that big of a concern and not worth the effort to fix. "most people don't have a separate partition."
why red hat went with systemd for RHEL is beyond me. maybe they think they can "fix it." _________________ Orlfman |
|
Back to top |
|
|
Tony0945 Watchman
Joined: 25 Jul 2006 Posts: 5127 Location: Illinois, USA
|
Posted: Tue Nov 10, 2015 9:50 pm Post subject: |
|
|
steveL wrote: |
Blame Lennux and deliberate marketing-driven groupthink, not Linux. |
"Lennux" I love it! (The term, not the crapware). |
|
Back to top |
|
|
steveL Watchman
Joined: 13 Sep 2006 Posts: 5153 Location: The Peanut Gallery
|
Posted: Wed Nov 11, 2015 7:55 am Post subject: |
|
|
steveL wrote: | Blame Lennux and deliberate marketing-driven groupthink, not Linux. :-) |
Tony0945 wrote: | "Lennux" I love it! (The term, not the crapware). |
Heh me too; but it's not my invention. Someone else on the forums came up with it, that I recall, but buggered if I'm digging through all those threads.. ;)
So sorry to whomever it was, feel free to claim attribution: and thanks for the best term for "GnomeOS/Linux-coreOS"[1] I've heard. :-)
==
[1] which completely misses the point that BSD userlands (which they claim to be trying to emulate) are made up of many separate projects, each of which respects the ecosystem; and their components can be replaced at will.
IOW: they make full use of real modularity, in order to deliver that "tightly-integrated" userland, with any sort of robustness.
As well as to innovate. |
|
Back to top |
|
|
miket Guru
Joined: 28 Apr 2007 Posts: 488 Location: Gainesville, FL, USA
|
Posted: Sat Nov 21, 2015 3:13 am Post subject: |
|
|
Saw this on LWN: Quote: | Lennart Poettering introduces the sd-event API for the implementation of event loops. "sd-event.h, of course, is not the first event loop API around, and it doesn't implement any really novel concepts. When we started working on it we tried to do our homework, and checked the various existing event loop APIs, maybe looking for candidates to adopt instead of doing our own, and to learn about the strengths and weaknesses of the various implementations existing. Ultimately, we found no implementation that could deliver what we needed... |
Haven't I heard this song before somewhere? |
|
Back to top |
|
|
truekaiser l33t
Joined: 05 Mar 2004 Posts: 801
|
Posted: Sat Nov 21, 2015 3:38 am Post subject: |
|
|
miket wrote: | Saw this on LWN: Quote: | Lennart Poettering introduces the sd-event API for the implementation of event loops. "sd-event.h, of course, is not the first event loop API around, and it doesn't implement any really novel concepts. When we started working on it we tried to do our homework, and checked the various existing event loop APIs, maybe looking for candidates to adopt instead of doing our own, and to learn about the strengths and weaknesses of the various implementations existing. Ultimately, we found no implementation that could deliver what we needed... |
Haven't I heard this song before somewhere? |
Clue me in on what a loop api is? |
|
Back to top |
|
|
khayyam Watchman
Joined: 07 Jun 2012 Posts: 6227 Location: Room 101
|
Posted: Sat Nov 21, 2015 9:20 am Post subject: |
|
|
truekaiser wrote: | miket wrote: | Saw this on LWN: Quote: | Lennart Poettering introduces the sd-event API for the implementation of event loops. "sd-event.h, of course, is not the first event loop API around, and it doesn't implement any really novel concepts. When we started working on it we tried to do our homework, and checked the various existing event loop APIs, maybe looking for candidates to adopt instead of doing our own, and to learn about the strengths and weaknesses of the various implementations existing. Ultimately, we found no implementation that could deliver what we needed... |
Haven't I heard this song before somewhere? |
Clue me in on what a loop api is? |
truekaiser ... a loop is an example of control flow, and an API provides the/a method by which this is done. Miket's point, I think, is that "do[ing] our homework" translates to "we ignored pre-exiting methods and came up with our own" ... alluding, I think, and miket can correct me, to kdbus.
best ... khay |
|
Back to top |
|
|
steveL Watchman
Joined: 13 Sep 2006 Posts: 5153 Location: The Peanut Gallery
|
Posted: Sat Nov 21, 2015 1:16 pm Post subject: |
|
|
truekaiser wrote: | Clue me in on what a loop api is? |
An oxymoron; event loops are done using poll (or select), or an OS-specific interface like epoll or kqueue.
Libev and libevent are the two shim layers most people use.
It's an oxymoron since you cannot simply hand over everything to some convenience layer, just as you cannot simply shove everything down one, inappropriate bus.
It seems that, having learnt all the wrong lessons from Winbloze, Poeterring is reaching out for all the wrong lessons of Java.
Though it's the same idiotic approach, overall: pretending we can make everything nice and shiny, even though it means we're setting you up to fail, by not allowing you just to use the standard APIs, nor to mix and match (no, no, no); you know: to do your job as a programmer.
Anyone using ErringWare is setting themselves up for a shedload of maintenance work, which will culminate in needing to rip it apart in a year or two.
Good luck following that brick-road of OS-specific obsolescence.
IMO systemdbust is just a distraction while RedHat partner up with Microsoft; baking GPL evasion into the Linux userland was step one. |
|
Back to top |
|
|
miket Guru
Joined: 28 Apr 2007 Posts: 488 Location: Gainesville, FL, USA
|
Posted: Sun Nov 22, 2015 12:13 am Post subject: |
|
|
khayyam wrote: | translates to "we ignored pre-exiting methods and came up with our own" ... alluding, I think, and miket can correct me, to kdbus. |
Well yes: kdbus, su, logger, device management, authorization, control-groups management, xinitd, and on and on. That song has many verses.
As to the event-loop thing, I'm still scratching my head trying to figure out why you'd have just that kind of thing in a system-level process when there is so much parallelism going on. Normally, the place to have an event loop is within an application so that it can handle events received from some source like a window manager or a socket. Is this API for slurping up event messages for such an event loop? If that's the case, I don't know how you could make a single standardized way to improve on current practice.
Maybe what he has in mind is a standard way to be the relayer/producer of events in event loops. This could turn out to be the next little "gift" he proposes for the Gnome project to assimilate. Egads, if you've come up with other ways to avoid systemd lock-in while using Gnome, this way would be more sure circumvent them. Without systemd, you would not be able to have in Gnome the central component of any window manager: the event dispatcher.
But then, I may be completely off base about this. A comment on the LWN article asked why the Poettering crew did not consider libdispatch. A little searching shows that that libdispatch is "the user space implementation of the Grand Central Dispatch API as seen in Mac OS X version 10.6 Snow Leopard." The best handle I could get on *that* was on the Wikipedia page. Grand Central Dispatch does events, but also threads. From what I can tell, it might be too heavy even for systemd (wouldn't that be a first!)
I'm back to scratching my head. |
|
Back to top |
|
|
steveL Watchman
Joined: 13 Sep 2006 Posts: 5153 Location: The Peanut Gallery
|
Posted: Sun Nov 22, 2015 9:10 am Post subject: |
|
|
No need to scratch your head, miket, as you've nailed it: it's simply YAF mechanism for "vendor" lock-in.
You indeed cannot "make a single standardized way to improve on current practice" in terms of abstracting the OS-specifics of event-handling; libev already does it, with minimal code.
In any case, Poeterring has a track record of forcing OS-dependencies where none are needed, nor welcome; if you're using the system-idiot approach then you're already locked into Lennux, and all the bloat and entanglement it implies.
Only a deranged programmer, or a complete newb, would consider handing off their event-handling to not only an OS-specific layer, but a "portability layer" which ties them into one specific, badly-conceived userland, on one specific niche OS.
Not when the current practice is to use a minimal shim (a real portability layer) which allows them to run on practically every OS out there.
Or an employee who has no choice about the matter (less relevant in FLOSS, which is the whole point of it: we can focus solely on the craft, if we are not blinkered); or a corporate stooge advocating that you drink the kool-aid^W^W^W sign-up to RedHat/Microsoft's patently-false agenda. |
|
Back to top |
|
|
The_Great_Sephiroth Veteran
Joined: 03 Oct 2014 Posts: 1602 Location: Fayetteville, NC, USA
|
Posted: Mon Nov 23, 2015 1:58 pm Post subject: |
|
|
I've been out of this conversation for a while. I suppose the main reason is due to me leaving Debian for Gentoo to avoid systemd and not having to deal with it. I absolutely love Gentoo. It is fast and stable. It reminds me of my early days with Epoch and then Debian 2. I had to learn so much and it just beat the pants off of Windows, but I kept Windows on my powerhouse systems because I was a gamer. Good days.
Now, due to the fuss about systemd, which I despise at this point, I gave it a shot on an old system. Gentoo 64bit, KDE. systemd, and the same apps as my work laptop. I am NOT pleased with it. It crashes or locks up, it is slower than the non-systemd clone, and it does weird things at times. I actually have two Gentoo installs on this system. They're identical, except one is systemd and one is plain old-school style. Now, the results!
- Boot time is almost identical, with less than a second of difference
- The systemd install pulled in more dependencies and was bloated
- The systemd system is slower, despite being at the start of the disk
- The systemd system tends to just "freeze" and be non-responsive, requiring a hard reset
- The systemd system seems to have apps crash, which never happens on the regular install
- I experience what appears to be a kernel panic, without a panic. Maybe PID1 died?
I am about done with this test. I do not like systemd. I am not even mentioning the many security issues with systemd. My experience tells me that it is unstable, slow, and bloated. That's enough to push me back to Windows. I game on Windows 7 Professional 64bit and it has yet to blue-screen or crash on me. Granted, I don't browse the web on my gaming systems and that likely has something to do with it, but at this point, it is more stable than systemd. _________________ Ever picture systemd as what runs "The Borg"? |
|
Back to top |
|
|
Tony0945 Watchman
Joined: 25 Jul 2006 Posts: 5127 Location: Illinois, USA
|
Posted: Mon Nov 23, 2015 3:30 pm Post subject: |
|
|
The_Great_Sephiroth wrote: |
- Boot time is almost identical, with less than a second of difference
- The systemd install pulled in more dependencies and was bloated
- The systemd system is slower, despite being at the start of the disk
- The systemd system tends to just "freeze" and be non-responsive, requiring a hard reset
- The systemd system seems to have apps crash, which never happens on the regular install
- I experience what appears to be a kernel panic, without a panic. Maybe PID1 died?
I am about done with this test. I do not like systemd. I am not even mentioning the many security issues with systemd. My experience tells me that it is unstable, slow, and bloated. That's enough to push me back to Windows. I game on Windows 7 Professional 64bit and it has yet to blue-screen or crash on me. Granted, I don't browse the web on my gaming systems and that likely has something to do with it, but at this point, it is more stable than systemd. |
Thank you for the detailed tests using the same machine. That is Science!
Win7 is not bad, with regard to stability. On my wife's laptop, I have also run systemrescuecd using the option to reside in memory. It takes a long time to read the CD, then it blows the doors off of Windows 7.
Have you tried gaming with Wine? I've run some video editing software with Wine and it actually was slightly faster than XP. I attribute that to either a superior filesystem (ext4) or video libraries optimized for my CPU. |
|
Back to top |
|
|
Navar Guru
Joined: 20 Aug 2012 Posts: 353
|
Posted: Mon Nov 23, 2015 7:23 pm Post subject: |
|
|
When complexity has reached critical mass, you hold a conference. |
|
Back to top |
|
|
|
|
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
|
|