Gentoo Forums
Gentoo Forums
Gentoo Forums
Quick Search: in
kdbus in the kernel
View unanswered posts
View posts from last 24 hours

Goto page Previous  1, 2, 3 ... 12, 13, 14 ... 25, 26, 27  Next  
Reply to topic    Gentoo Forums Forum Index Kernel & Hardware
View previous topic :: View next topic  
Author Message
steveL
Watchman
Watchman


Joined: 13 Sep 2006
Posts: 5153
Location: The Peanut Gallery

PostPosted: Sat Apr 18, 2015 4:30 pm    Post subject: Reply with quote

What technical merits? All he's doing is the usual "repeat some of what you've already heard to make it sound like I've said something."

Utter nonsense as usual.
Back to top
View user's profile Send private message
krinn
Watchman
Watchman


Joined: 02 May 2003
Posts: 7197

PostPosted: Sat Apr 18, 2015 5:55 pm    Post subject: Reply with quote

ulenrich wrote:
[edit] to not provoke the Gentoo user forum community as Neddy requested I remove my trolling in the last line of this post

If you post that trolling message because you think i was trolling, i wasn't, but i really don't think that many bmw ppt exist with benchmark related to lib/k/dbus ; and GHK might have done himself the same error as me and refer really to that doc.
At least, please take time to reread it, you'll see i have put some warning that i'm of course unsure it is really the doc he is speaking about.
Back to top
View user's profile Send private message
NeddySeagoon
Administrator
Administrator


Joined: 05 Jul 2003
Posts: 44182
Location: 56N 3W

PostPosted: Sat Apr 18, 2015 6:50 pm    Post subject: Reply with quote

steveL,

Technial merits was probably the wrong term. Self consistent would have been better.

Its perfectly possible to make a self consistent argument that is based on one or more false premises.
Like this.

dbus and its would be successor kdbus are both a load of dingos kidneys.
Thats an assertion based on my opinion, which may or may not have any basis in fact. As its an opinion, I'm not required to support it or defend it.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.
Back to top
View user's profile Send private message
steveL
Watchman
Watchman


Joined: 13 Sep 2006
Posts: 5153
Location: The Peanut Gallery

PostPosted: Sun Apr 19, 2015 2:10 am    Post subject: Reply with quote

Neddy: indeed; insanity is often self-consistent. It's only when you question the flawed premise, typically a part of the background that everyone takes as read, that you trigger the defensive strikes.

In systemd's case the biggest flawed premise is that "shell is bad" (leading to the dumbass conclusion: "let's use javascript instead".)
Back to top
View user's profile Send private message
ulenrich
Veteran
Veteran


Joined: 10 Oct 2010
Posts: 1389

PostPosted: Sun Apr 19, 2015 2:21 am    Post subject: Reply with quote

krinn wrote:
ulenrich wrote:
[edit] to not provoke the Gentoo user forum community as Neddy requested I remove my trolling in the last line of this post

If you post that trolling message because you think i was trolling, i wasn't, but i really don't think that many bmw ppt exist with benchmark related to lib/k/dbus ; and GHK might have done himself the same error as me and refer really to that doc.
At least, please take time to reread it, you'll see i have put some warning that i'm of course unsure it is really the doc he is speaking about.

No, no I didn't talk about you in the last sentence I removed. And for sure GKH has done the same error "remembering" that power presentation. I was just doing a little detectives work by looking up the years (2012 of the presentation and 2013 of the LWN article).

Today sitting in the sun thinking,
@krinn, what performance test do you want: If kdbus reduces copying the messages from four to just one copy, the test depends on the size of the messages. If you benchmark empty ping messages, you probably will have less performance difference to user space dbus. If you take big messages your benchmark probably will show a four times faster kdbus. Not much news a benchmark would tell, what do you think should be tested?
The little thing trollish of your post was at the end saying in seven sentences : kernel dbus not in the kernel is not kernel dbus.

It seems quiet clear to me what motivations play the game here:
- The BMW, automotive guys want
a) big messages
b) extra secured by kernel means for their new live saving techniques in cars
- The Gnome wants
c) responsiveness of cut down in little pieces desktop parts (symbolize a dbus action with a pipe and imagine the messages pure ascii and the unix veteran becomes a Gnome fanboi)
- Poettering wants to
d) get rid of his "special" units by providing that typical systemD mechanism he would like to proudly present from the first init part at boot.
- The container buisiness wants
e) a more general solution kdbus cannot deliver
- Everyone wants a
f) secure kernel. But having so many different security/policy subsystems - mostly layerd orthogonal to each other - shows: We have a problem not solved in sober principles.
- /me wants
g) a first step to the foundations of artificial intelligence: A way all the different units lying around me and in my body, in my car and cloth can work together in a secure way (... coming up next more soonishly you may think)

... it's hot in here, the roof is on fire.
_________________
the thread ain't easily find an end
Back to top
View user's profile Send private message
krinn
Watchman
Watchman


Joined: 02 May 2003
Posts: 7197

PostPosted: Sun Apr 19, 2015 5:01 am    Post subject: Reply with quote

ulenrich wrote:
@krinn, what performance test do you want

Well, actually none, i'm not the one asking any tests ; but GHK gives tests results to backup his claims of better performance.
If you start using test result, i will ask you the test procedure to be able to check your claim, else everyone could claim random result, and while you say you found 100x better speed, i could answer i found myself 10000x time slower result testing the very same part as you.
I don't think Linus is asking anything else, GHK claim something, Linus wants see how to check it's true.

ulenrich wrote:
The little thing trollish of your post was at the end saying in seven sentences : kernel dbus not in the kernel is not kernel dbus.

Ah ok, that's the part you find trolling, well, here's why i said that:
Nowhere to be found anything about security of kdbus, except because it is in kernel ; so the kdbus security might be 0, and its security is only given by the kernel. So if kdbus isn't in kernel, the security argument is gone, and without this one, kdbus looks empty or at best, poor.

You may find that trolling to say kdbus have 0 security, but can you tell me about kdbus security then? Because all i see (of course i didn't dig it, as i'm not interest in kdbus myself) is that kdbus is secure by being in kernel. This might not be the reality, but this is how they are selling kdbus.
So, if kdbus primary selling argument is security, and that argument is strong because it is base on kernel security, if you can't get kdbus in kernel, that primary selling argument is gone, and kdbus itself might be dead then. In order to not let it die, you would do anything you can to let it be in kernel.
Back to top
View user's profile Send private message
ulenrich
Veteran
Veteran


Joined: 10 Oct 2010
Posts: 1389

PostPosted: Sun Apr 19, 2015 10:06 am    Post subject: Reply with quote

krinn wrote:
ulenrich wrote:
@krinn, what performance test do you want

Well, actually none, i'm not the one asking any tests ; but GHK gives tests results to backup his claims of better performance.
If you start using test result, i will ask you the test procedure to be able to check your claim, else ...

Well, for sure a very legitimite question Linus asked. I just showed GHK didn't gave it one correct thought at that point in the kernel mailing discussion. Otherwise he would have asked my question: What messages to test? Empty ping messages wouldn't show the memfd effects but would openly show other issues in the new code.

GKH may have been deeply frustrated at that point in the discussion, because as I just pointed out in my post, he himself had thought and spoken out all of it two years ago, what his critics say today. And he thinks they will come to the same conclusions once they got familiar with it. But I like it time being given for just the chance other ideas showing up.
_________________
the thread ain't easily find an end
Back to top
View user's profile Send private message
NeddySeagoon
Administrator
Administrator


Joined: 05 Jul 2003
Posts: 44182
Location: 56N 3W

PostPosted: Sun Apr 19, 2015 2:21 pm    Post subject: Reply with quote

ulenrich,

Benchmarks are worderful things but often flawed. They only tell you how good the code is running benchmarks, which are often difficult to relate to real world use cases.

An example from my day job a few years ago. A real time application runs an interrupt every 20ms. The Interrupt code takes 18ms to run, leaving 2ms every 20ms for the main loop. It works well as intended. One day someone makes some changes to the code (I forget why) The interrupt code now runs in 16ms leaving 4ms for the main loop. The perpertrator, who should have known better, claims that they have made a factor of two speed improvement.
Its left as an exercse for the reader to work out how the speed measuremet was being made.

Some that we are all fanilliar with - Graphics card performance :)

Hyperthreading. I'm aware of a few applications where enabling hyperthreading halves performance but Intel won't tell you that.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.
Back to top
View user's profile Send private message
Naib
Watchman
Watchman


Joined: 21 May 2004
Posts: 5710
Location: Removed by Neddy

PostPosted: Sun Apr 19, 2015 2:37 pm    Post subject: Reply with quote

NeddySeagoon wrote:

An example from my day job a few years ago. A real time application runs an interrupt every 20ms. The Interrupt code takes 18ms to run, leaving 2ms every 20ms for the main loop. It works well as intended. One day someone makes some changes to the code (I forget why) The interrupt code now runs in 16ms leaving 4ms for the main loop. The perpertrator, who should have known better, claims that they have made a factor of two speed improvement.
Its left as an exercse for the reader to work out how the speed measuremet was being made.
Sounds as bad as an issue we had at work with the main powerup sequencer being event driven that triggered cascaded timers... fell apart when a "28V avail" edge was significantly delayed - system would power ready at 14V, 28V ready interupt thrown at 16V :)

but yes benchmarks are ... subjective, just like stats "Do not trust any statistics you did not fake yourself." I have used "benchmarks" to force IT to upgrade hardware above and beyond what they had on their books purely because I wanted simulation machines to have Core2 chips and I knew how to weight a simulink sim to get the best outi of it
_________________
The best argument against democracy is a five-minute conversation with the average voter
Great Britain is a republic, with a hereditary president, while the United States is a monarchy with an elective king
Back to top
View user's profile Send private message
steveL
Watchman
Watchman


Joined: 13 Sep 2006
Posts: 5153
Location: The Peanut Gallery

PostPosted: Sun Apr 19, 2015 4:28 pm    Post subject: Reply with quote

ulenrich wrote:
GKH may have been deeply frustrated at that point in the discussion, because as I just pointed out in my post, he himself had thought and spoken out all of it two years ago, what his critics say today. And he thinks they will come to the same conclusions once they got familiar with it.

Nonsense on both fronts. We've discussed the underlying issues to death on these forums, and they haven't even factored into GKH's discussion yet.
Quote:
But I like it time being given for just the chance other ideas showing up.

Which is down to the kernel process, not kdbus numpties.
Back to top
View user's profile Send private message
gwr
Apprentice
Apprentice


Joined: 19 Nov 2014
Posts: 194

PostPosted: Sun Apr 19, 2015 6:17 pm    Post subject: Reply with quote

steveL wrote:
We've discussed the underlying issues to death on these forums, and they haven't even factored into GKH's discussion yet.


The lkml can't even get an answer about whether it needs to exist, let alone whether it has technical deficiencies.
Back to top
View user's profile Send private message
NeddySeagoon
Administrator
Administrator


Joined: 05 Jul 2003
Posts: 44182
Location: 56N 3W

PostPosted: Sun Apr 19, 2015 8:45 pm    Post subject: Reply with quote

Tony Hoare wrote:
“There are two ways of constructing a piece of software: One is to make it so simple that there are obviously no errors, and the other is to make it so complicated that there are no obvious errors.”


... and systemd fails on both counts.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.
Back to top
View user's profile Send private message
ulenrich
Veteran
Veteran


Joined: 10 Oct 2010
Posts: 1389

PostPosted: Sun Apr 19, 2015 9:07 pm    Post subject: Reply with quote

gwr wrote:
steveL wrote:
We've discussed the underlying issues to death on these forums, and they haven't even factored into GKH's discussion yet.


The lkml can't even get an answer about whether it needs to exist, let alone whether it has technical deficiencies.

You cannot point any interested party to this five month long thread, because it is a wild discussion mixing Dbus,kDbus,systemD and Redhat commercial strategies. When talking about some of dbus technical defiencies as if they still exist in kdbus, when talking about older network based ipc like this could provide dbus stateful features, when talking about sudo as ultima ratio like it has not given us a constant stream of exploits, I doubt any kernel developer would read more than half a page of this thread.
_________________
the thread ain't easily find an end
Back to top
View user's profile Send private message
steveL
Watchman
Watchman


Joined: 13 Sep 2006
Posts: 5153
Location: The Peanut Gallery

PostPosted: Mon Apr 20, 2015 2:57 am    Post subject: Reply with quote

ulenrich wrote:
I doubt any kernel developer would read more than half a page of this thread.

Yeah, not with spurious page-long "debates" which simply regurgitate what we already know in a Markov-chain response to anyone and everyone.

In any event, that's irrelevant; the same points were raised in 2012, and they still haven't been addressed.

So nice try to flame us all, but completely off the mark.
Back to top
View user's profile Send private message
mv
Watchman
Watchman


Joined: 20 Apr 2005
Posts: 6296

PostPosted: Mon Apr 20, 2015 3:25 pm    Post subject: Reply with quote

ulenrich wrote:
sudo as ultima ratio like it has not given us a constant stream of exploits

Nice try. Try again with a sane argument (or give references to the "constant stream of exploits" which I was able to do for policykit in a few minutes).
Hint: Most sudo bugs do not allow for any exploit - usually, the opposite is true (i.e. under certain cases you do not get the required permissions).
Of course, this does not mean that sudo cannot contain such a bug or never did contain, but the chances are practically a low as you can get it, because in contrast to policykit it satisfies the main necessary condition for security: To keep the privileged code as simple as possible.

BTW: I was not claiming that using sudo is without any danger. Everything which raises privileges is an extreme danger.

But having such a complex thing running when it is even unneeded (as for all "home" systems - from desktop to laptop) is just plain stupidity.
Back to top
View user's profile Send private message
tld
Veteran
Veteran


Joined: 09 Dec 2003
Posts: 1452

PostPosted: Mon Apr 20, 2015 5:47 pm    Post subject: Reply with quote

I'm convinced that the dbus/systemd proponents simply don't believe that anything can be good if it's simple, which seems to mirror the way MS approaches everything. Code that allows privilege escalation is the last place you want complex meta-data and layers of abstraction to the point were it becomes impossible to determine if it's safe or not...but Lord knows, it can't possibly be "flexible" enough without that right? (Queue mandatory fabricated requirements nobody needs)... :roll:
Back to top
View user's profile Send private message
Anon-E-moose
Advocate
Advocate


Joined: 23 May 2008
Posts: 4336
Location: Dallas area

PostPosted: Tue Apr 21, 2015 9:17 pm    Post subject: Reply with quote

Now there is some point being raised that maybe, just maybe, kdbus should be withdrawn from the merge request
until it can be properly reviewed. And some people still think that at most a new IPC should be put into the kernel
not "dbus".
_________________
Asus m5a99fx, FX 8320 - nouveau, oss4, rx550 for qemu passthrough
Acer laptop E5-575, i3-7100u - i965, alsa
---both---
5.0.13 zen kernel, profile 17.1 (no-pie & modified) amd64-no-multilib
gcc 8.2.0, eudev, openrc, openbox, palemoon
Back to top
View user's profile Send private message
Shamus397
Apprentice
Apprentice


Joined: 03 Apr 2005
Posts: 218
Location: Ur-th

PostPosted: Wed Apr 22, 2015 1:53 pm    Post subject: Reply with quote

Wow, sanity in the kernel development process, who would've thought it? ;)

I bet steam is coming out of Poettering's ears right about now. :D
Back to top
View user's profile Send private message
gwr
Apprentice
Apprentice


Joined: 19 Nov 2014
Posts: 194

PostPosted: Wed Apr 22, 2015 3:12 pm    Post subject: Reply with quote

Anon-E-moose wrote:
Now there is some point being raised that maybe, just maybe, kdbus should be withdrawn from the merge request
until it can be properly reviewed. And some people still think that at most a new IPC should be put into the kernel
not "dbus".


How has it not been properly reviewed before? Weren't all of the previous submissions that were criticized proper reviews? It should be "withdrawn until you solve the problems already brought up".
Back to top
View user's profile Send private message
tld
Veteran
Veteran


Joined: 09 Dec 2003
Posts: 1452

PostPosted: Wed Apr 22, 2015 4:15 pm    Post subject: Reply with quote

Anon-E-moose wrote:
And some people still think that at most a new IPC should be put into the kernel
not "dbus".
Which of course is the sane thing to do if there's a need.

I guarantee you this approach won't fly with the dbus/systemd crowd for two reasons: a) It's an approach that actually just might be beneficial to other totally unrelated projects, and b) it does nothing to promote the dbus/systemd lock-in that they're really after in all this. I can only imagine the excuses they'll dredge up to cloak all that.
Back to top
View user's profile Send private message
tld
Veteran
Veteran


Joined: 09 Dec 2003
Posts: 1452

PostPosted: Wed Apr 22, 2015 4:27 pm    Post subject: Reply with quote

gwr wrote:
How has it not been properly reviewed before? Weren't all of the previous submissions that were criticized proper reviews? It should be "withdrawn until you solve the problems already brought up".
From the way it looks based on the lkml posts it appears that it's hugely complex making it very difficult to review in any detail, even for experienced kernel devs...which is pretty scary right there. However the kernel devs primarily appear to be questioning the entire design...that is, whether or not it even makes sense to be doing what they're trying to do, no matter how the code is doing it...let alone doing so in the kernel. The dbus folks just don't want to hear that and will just endlessly talk around it demanding specifics around what's wrong with the code.

I don't see the kernel devs ever letting anything get in the kernel forever with such high level questions still open...at least I'd sure hope not.
Back to top
View user's profile Send private message
NeddySeagoon
Administrator
Administrator


Joined: 05 Jul 2003
Posts: 44182
Location: 56N 3W

PostPosted: Wed Apr 22, 2015 7:05 pm    Post subject: Reply with quote

Well, Red Hat already maintains a huge kernel patch set. Whats one more patch to them?
Or Red Hat could follow through an the systemd April Fool and just fork the kernel.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.
Back to top
View user's profile Send private message
miket
Guru
Guru


Joined: 28 Apr 2007
Posts: 412
Location: Gainesville, FL, USA

PostPosted: Wed Apr 22, 2015 9:38 pm    Post subject: kdbuswreck Reply with quote

Over at LWN, where you find a lot of systemd/kdbus/*kit cheerleading, comes an article about the delays in getting kdbus merged into the kernel. They call the situation "The kdbuswreck".

Actually, I think that is an apt term to describe the project itself.

There you go. systemdamage and kdbuswreck.
Back to top
View user's profile Send private message
Anon-E-moose
Advocate
Advocate


Joined: 23 May 2008
Posts: 4336
Location: Dallas area

PostPosted: Thu Apr 23, 2015 12:13 am    Post subject: Re: kdbuswreck Reply with quote

miket wrote:
There you go. systemdamage and kdbuswreck.


LoL
_________________
Asus m5a99fx, FX 8320 - nouveau, oss4, rx550 for qemu passthrough
Acer laptop E5-575, i3-7100u - i965, alsa
---both---
5.0.13 zen kernel, profile 17.1 (no-pie & modified) amd64-no-multilib
gcc 8.2.0, eudev, openrc, openbox, palemoon
Back to top
View user's profile Send private message
depontius
Advocate
Advocate


Joined: 05 May 2004
Posts: 3406

PostPosted: Thu Apr 23, 2015 12:21 am    Post subject: Re: kdbuswreck Reply with quote

miket wrote:
Over at LWN, where you find a lot of systemd/kdbus/*kit cheerleading, comes an article about the delays in getting kdbus merged into the kernel. They call the situation "The kdbuswreck".

Actually, I think that is an apt term to describe the project itself.

There you go. systemdamage and kdbuswreck.


Naaah, it's just those haters who work on the kernel. THEY'RE the problem. There are no problems with systemd - NOTBUG!

EDIT - while this is said in all snark, I'm sure there are those who will espouse this opinion.
_________________
.sigs waste space and bandwidth


Last edited by depontius on Thu Apr 23, 2015 3:33 pm; edited 1 time in total
Back to top
View user's profile Send private message
Display posts from previous:   
Reply to topic    Gentoo Forums Forum Index Kernel & Hardware All times are GMT
Goto page Previous  1, 2, 3 ... 12, 13, 14 ... 25, 26, 27  Next
Page 13 of 27

 
Jump to:  
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