ZFS On Linux Landing Workaround For Linux 5.0 Kernel Support

[Weasel]

no, you are a wannabe clown giving me a lot of fun the past few days
not more and not less

> makes a retarded claim
> boasts about how much "experience" he has
> gets proven wrong
> comes up with 5yo bitch cop out

ok.

[starshipeleven]

I find it hard to believe Clear Linux for example doesn't use something similar. But of course it's a "patched" kernel and not "mainline", just like distro kernels are. If mainline was so good then nobody would patch it up.

Lol is this your only argument?

You know right that you can make a ton of wild dirty unsafe hacks to make any software go "faster", (or "look like faster" in synthetic benchmarks)?

Clear Linux is a development testbed (to see what makes sense to actually optimize in a safe way) and a way to show off, not a true production distro.

[starshipeleven]

YOU bring the numbers that it's not affected by this change.

It's you that say this is going to be bad for performance, it's you that need to back this up.

Actual scientific method says that we just don't know if this is good or bad for performance. We need to see benchmarking.

Changes require justification not the other way around, dummy.

This change is justified by the fact that this interface was not defined as stable, is not used by any other kernel module, and that the last kernel module using it was reworked to not use it anymore as part of a cleanup of its code.

ZOL performance is completely irrelevant as its out-of-tree so they get no say.

[starshipeleven]

I'm an actual developer not a piece of shit "IT guy" go take a piss with your shit "admin" job you clueless clown.

There is at least two of us! Two!

[starshipeleven]

I don't give a shit about Fedora but every proper distro usually has its own patched kernels, especially Debian based.

I find very hard to believe that distros maintain massive amounts of patches to have all relevant kernel modules use these interfaces instead of whatever system they are using upstream.

Could I remind you that the very mail of the patch that stopped exposing this functionality to modules cites the only user being some form of UEFI-related module, and says it is the ONLY user and that it has been like that for quite a while?

[Weasel]

i wonder why this clown of "an actual developer" just don't "devel" what he wants because all the bullshit he is posting here would have been enough for a prototype in number of code lines :-)

If systemd pays me at least $54 per hour then I might consider wasting my time on "devel" that alternative you speak of.

[Weasel]

because all of my systems

Dude NOBODY FUCKING CARES ABOUT **YOU**. Shut the fuck up already about your shitty systems or your pathetic life.

And just for the record, I DON'T GIVE A SHIT IF YOU DON'T CARE ABOUT ME. I did NOT post this for you. I will continue to whine because others can read it. I don't care about your life, just as you don't care about mine.

But I don't talk about my life, or my systems, you keep going on and on about it. NOBODY CARES.

So I will continue to whine, and frankly I couldn't care less if you reply. I usually don't block people, even clueless ones like oiaohm, but I'll make an exception here, because you're more degenerate than I can count, and argue like a fucking kid.

So you're on my block list. Feel free to block me too, I don't care nor give a shit what you think, and I don't post "whining" for you, so get fucking lost.

[Weasel]

Lol is this your only argument?

You know right that you can make a ton of wild dirty unsafe hacks to make any software go "faster", (or "look like faster" in synthetic benchmarks)?

No, and i don't care? I'm simply referring to the fact that using vector instructions and registers can be useful within the kernel. So those functions are useful.

Clear Linux is a development testbed (to see what makes sense to actually optimize in a safe way) and a way to show off, not a true production distro.

Citation? Or is this just your opinion?

[starshipeleven]

This is true too. It is rumor by the Linux base that it was intentional but Sun developers empathically claim the opposite. The "Oh they did this on purpose!" claim because they *knew*.. :: roll eyes :: Unix people tend not to like the restrictions of the GPL so the went with something else. Who knows the truth really..

Main difference between CDDL and MIT/BSD licenses is the patent granting, which is also something that is present in a slightly later license like the Apache 2.0.

It's entirely possible that they only genuinely wanted to have an Apache-like license (which is GPL-compatible) and screwed up, but this does not helm much more than deciding who to blame, the execs or the lawyers.

Oracle can be safely blamed because they could have re-licensed it to Apache when they bought Sun or something but they didn't as they are evil as we all know.

[starshipeleven]

No, and i don't care? I'm simply referring to the fact that using vector instructions and registers can be useful within the kernel. So those functions are useful.

Yes. The functions were not removed, they were refactored and are just not exported anymore so only the kernel itself can use them, but not the kernel modules.

And I'm saying that maybe you don't see the whole picture here because they clearly said they don't use them in kernel modules, and that's why they removed the interface.
Maybe the kernel modules do something similar to what userspace does, to the contrary of what you thought? I would not rip my clothes in awe if that was the case, you know.

Citation? Or is this just your opinion?

They are trying to make it look like it's good for production of course, but on their 01.org site they do say something https://01.org/clearlinux/

The Clear Linux* Project for Intel® Architecture is building a Linux OS distribution targeted for various cloud use cases. The goal of the Clear Linux OS is to showcase the best of Intel Architecture technology, from low-level kernel features to more complex use cases that span the entire operating system stack.

If that isn't enough, what about some independent research? Think with your head and all that?

For example, since Clear Linux has optimized packages (which do more than just an optimized kernel in many cases due to their instruction-detecting code path swapping, that does not rely on kernel) one would expect it to be common in containers like docker. Is it though?
containers based on clear linux https://hub.docker.com/search?q=clea...e=image&page=1 41 results
containers based on debian https://hub.docker.com/search?q=debian&type=image 14186 results
containers based on Alpine (which is the best imho) https://hub.docker.com/search?q=alpine&type=image 31022 results
containers based on the queen bitch of all Linux distros, Ubuntu https://hub.docker.com/search?q=ubuntu&type=image 56572 results

No it is not popular, like at all.

What about prerequisites for actual midrange server usage? Do they offer some kind of company support contracts which is the bare minimum for any company to consider them on midrange servers? https://clearlinux.org/
No they don't. There are free docs and that's it.
Meanwhile, even Debian has some basic paid consultant list for companies, which might still be a joke if compared to RHEL or SLES, but that's more or less how it works also for Windows servers too (my job irl) https://www.debian.org/consultants/

So what is left? Hobbist usage, and very high-end use, think like Facebook or other big company with large internal IT department that can deal with their own custom setup and needs no external assistence.

Does it make a true production distro? I'd say no.