Linux 4.16 Released (phoronix.com)

An anonymous reader writes: Linus Torvalds has released Linux 4.16. Linux 4.16 integrates more of the VirtualBox guest drivers into the kernel, provides AMDGPU DC multi-display synchronization, continues with mitigation improvements for Spectre and Meltdown mitigation, tightens up access to /dev/mem by default, and many other improvements and changes.

Comments

Worst day to release a new kernel

[Anonymous Coward]

Is this release an April Fools joke?

Re: Worst day to release a new kernel

[ThePhish]

That's because no one wants to try to top OMGPonies!

Re: Worst day to release a new kernel

[Anonymous Coward]

I thank you for the no jokes this year. I know it was a "thing" but it sucked so much.

Re: Worst day to release a new kernel

[antdude]

Maybe the lack of jokes IS the April 1st joke to /. since it was a tradition!

Re: Worst day to release a new kernel

[johnhennessy]

Well, i assume the %pCOW in printk.c is a joke?

But no one on slashdot reads the articles - so I guess itâ(TM)s pretty well hidden.

So much bloat

[Anonymous Coward]

Monolithic kernels will eventually get too big for their britches. Whatâ(TM)s a decent alternative?

Re:So much bloat

[fibonacci8]

Mac OS L, HURD 3.0, or, BeOS Warp will probably exist around the time that monolithic kernels get too big for their britches. Get back to us when you've tried them, if you please.

Re:So much bloat

[F.Ultra]

Not discovered kernel modules yet or are you simply trolling?

Re: So much bloat

[q_e_t]

The user land tools in "Linux" are generally GNU, and GCC and associated tools hgaved been very successful, so your criticism seems off-base. If you had criticised GNU's terrible and ugly website, along with the failure of HURD, you might have a point.

Re: So much bloat

[Anonymous Coward]

But... But... HURD *identifies* as a kernel! We therefore must use the politically compelled pronoun, the hapless 'whirrrrr" sound as it tries to find any of its own hardware to confirm its own identity!

Come to think of it, it is just like Leah Rowe of the libreboot project. I thought I saw traces of the identity confused politics.

Re: So much bloat

[byrtolet]

The user land tools in "Linux" are generally GNU, and GCC and associated tools hgaved been very successful, so your criticism seems off-base. If you had criticised GNU's terrible and ugly website, along with the failure of HURD, you might have a point.

I like GNU's website. It is fast.

I would take fast and ugly any time!

Re: So much bloat

[q_e_t]

Fast is good, but it could be fast and less ugly. I'd accept some CSS and a few images, but it would be nice to avoid fetching half the world's entire Javascript code base like some websites seem to do. I would think there is some middle ground between pretty and fast GNU could use. Some structural changes would be good too, but it's certainly not the worst structured site I've seen.

Linux 4.16

[Anonymous Coward]

For Linus so loved the world, that he gave his only begotten Kernel that whosoever uses Linux should not perish, but have everlasting uptime.

improvements

[Hognoxious]

Is systemd part of the kernel yet? Or is it the other way round?

Re:improvements

[Billly Gates]

SystemD is a great OS. It just has a crappy startup daemon.

Re:improvements

[F.Ultra]

Haven't you guys moved over to BSD or Devuan yet?

Re:improvements

[BadDreamer]

Corner cases, mostly.

Init is fine with daemons that work as intended, and are reasonably stable. It suffers with broken code, and most competing replacements are trying to address that, not any actual flaws in init itself.

Re:improvements

[Antique Geekmeister]

SysV init had a number of critical issues. One was the failure to successfully and automatically restart a crashed daemon. There have been several lighterweight, safer init systems. Upstart worked well, but never became popular.

Re:improvements

[KiloByte]

indeed, that restart feature is a bandage for shoddy software

It's worse than that: it's a security hole with no excuse whatsoever.

A daemon never has a reason to crash. I don't recall the last case a daemon crashed for me (not counting failing to start), and I run unstable on a number of non-production machines. Code that goes into a proper distribution these decades is not hopelessly buggy.

If your company has some local daemon that requires a band-aid, you need to debug crashes ASAP as there's above 50% chance the crash is exploitable. Obviously, usually your management is criminally negligent and wants you to add features instead, but in such case it takes a single line to use runit or some other daemon supervisor.

Thus, auto-restarting crashed daemons is such a bad idea it must never be a part of an init system. Doing this anyway should be considered exceptional, and in such cases you need to take some extra precautions anyway.

Re:improvements

[q_e_t]

It depends on why they are failing. If they are failing due to a lack of a resource, that's not unreasonable. Ideally you would provide a suitable exit code on a crash and have a suitable monitoring system pick that up, and allow a configuration for that daemon decide on the next action.

Re:improvements

[gweihir]

Indeed. I use it. It works and not only can I fully understand it, I can also easily modify and extend it. Of course, it does not instill the irrational happiness of being protected by an all-seeing all-knowing benevolent demon that the systemd-fanatics seem to experience in its presence. Init just works and works well. For some people that does not seem to be enough though.

Re:improvements

[gweihir]

Big egos that want to leave their mark. Some people (bad engineers, all of them) cannot understand "if it is not broken, do not fix it". They think that it must be _them_ that obviously can do things massively better. Incidentally, that is pretty much what drives the systemd train-wreck. (Well, for variable values of "train-wreck". It will still take a long time until the systemd-morons have degraded Linux to the level of Windows. But why should I use something worse just to support the mental masturbation of some fuckup?)

Re:improvements

[gweihir]

A direct lie. It is not the task of an init-system to restart a crashed demon. That is what you use a restart-wrapper for and it often needs to be customized for the demon in question in order to work well. And there are already such generic wrappers (where that is enough) that work well.

Re:improvements

[gweihir]

Whoever modded this down is apparently completely clueless. If you do automatic restart when demons fail because of resource starvation, you may well make the complete machine unresponsive and make it impossible to log into it too. That is a worst-case scenario.

Re:improvements

[gweihir]

Thanks to whoever modded this down for further demonstrating the utter cluelessness of the systemd morons.

Re:improvements

[HiThere]

Well, it's an overstatement. But if a daemon crashes without being asked to then there's clearly an error. Either a software error or a hardware error, but an error. And errors are frequently the kind of place that a malicious actor can use...so trusting them to restart is unwise...unless you told it to crash for some reason.

So. Overstatement, but not actually wrong. Now the question is, what do you do if you need to run the daemon, but it has errors that occasionally cause it to crash without being asked? The best thing to do would clearly be to fix the error, but this may be infeasible, so the next best thing is to have a monitor process that restarts it. But this probably shouldn't be a part of the init system, and certainly shouldn't be applied to all daemons by default.

Re:improvements

[HiThere]

Why so much hate? Because systemD is being shoved down people's throats. I know of nobody who asked for it. I sure didn't. I haven't yet switched to a distribution that doesn't use it, but that sure doesn't mean that I wanted it. It doesn't even mean I won't switch in the future.

Re:improvements

[gweihir]

You really, really have no experience with modern systems. Or you may just be completely clueless.

Re:improvements

[q_e_t]

Yes, resource starvation is definitely a time you'd not want a restart, unless it's a case of something like a lack of licences served by FlexLM, but you wouldn't really want to design a system with a daemon depending on an externally served licence that is pooled with other consumers of that licence type.

Re:improvements

[q_e_t]

The issue is not necessarily the daemon that is failing using the resource, but potentially some other process. So daemon quits, resources freed, other process uses the resources.

However, not all OSes free all resources on all crashes.

Re:improvements

[iggymanz]

If there is not enough resource a human should be looking at that system, not just restarting a daemon

instead the majority of the use cases is unstable garbage like mysql that doesn't even belong in a business.

Re:improvements

[Hognoxious]

You could say that about anything. Ask yourself why there isn't only one car manufacturer or one brewery.

The answer, for the hard of understanding, is that some things suit some people and other things suit others. On top of that there's ego (NIH syndrome).

However all the other competing init systems have one thing in common; you can pick and choose them at will. They aren't forcefully bundled by building in an arbitrary dependency with other random components.

Re:improvements

[Hognoxious]

Sounds like you're advocating tools that do one thing and do it well.

It'll never catch on.

Re:improvements

[The Evil Atheist]

So?

The claim was made that init was a "perfectly good one".

The fact that any other init system exists alludes to the fact that, no, init was not perfectly good.

It's like the hatred for systemd is so irrational that people would go the other extreme and say init is "perfectly good" that other people couldn't see shortcomings in init.

Re:improvements

[Joey Vegetables]

Excluded middle fallacy here. That initd needs replacement does not imply that systemd should be that replacement.

Re:improvements

[iggymanz]

Yes it can, you are speaking in ignorance, I admin hundreds of systems with 1.5 million users. The systemd way is wrong.

Re:improvements

[gweihir]

Funny. And completely off the mark. But I guess you are a paid, honor-less propaganda shill, so you cannot actual acknowledge the truth anyways.

Re:improvements

[gweihir]

Indeed, thanks. The systemd propaganda lies are as transparent as always. Also, that init is limited in some scenarios does not imply it needs replacement. It implies that it needs an alternative for those scenarios. Unless init is broken in most or all scenarios (which it is not), there is no reason to do a replacement. It is not a problem with clean startup scripts to support several different init-systems.

Re:improvements

[q_e_t]

If there is not enough resource a human should be looking at that system, not just restarting a daemon

Ideally, yes. Sometimes scale or staff availability makes that difficult, though. But it might lead to instability without analysis. Whether that is acceptable depends on the business's risk appetite and risk management processes, although analysing the risk can be non-trivial, especially in complex architectures that can, more-of-less, exhibit emergent behaviour.

Re:Windows 10 interface + Linux kernel

[Billly Gates]

When?

How Microsoft made Linux run on Windows was by adding hooks through an abstract layer. Unfortunately, Windows was designed with a HAL (hardware abstract layer) from the beggining when David Cutler wanted to make it portable across hardware. Win64 and wow32 (win32onwin64) are really layers on top of the kernel for runtimes. Linux is another one.

Linux is a macro-kernel so this would be messy (reminds me of the old Linux is obsolete it isn't a micro kernel debate from Andy Tannabum) but could be I guess possible if someone wanted to a winapi including NDIS and lord knows what else hooks into the linux kernel itself.

Then a daemon could use those hooks and launch the inverse of WSL that is on Windows 10 to run binaries.

Also keep in mind at this time only console linux apps work on Windows. This is because the OpenGL and device driver API and ABI's have not been ported yet. On Windows everything is gui based :-(

So this would not be easy or possible unless one wants to just run win32 powershell scripts and dos commands.

Re:Windows 10 interface + Linux kernel

[F.Ultra]

WSL is not Linux on Windows. It's GNU/Windows, nothing more.

Re: Windows 10 interface + Linux kernel

[Anonymous Coward]

You've obviously never had to admin windows networks. There's a lot of bubblegum and nailing wire between you, your products, and your customers if the entire ecosystem between them is windows based.

Re:Windows 10 interface + Linux kernel

[Anonymous Coward]

WSL is not "GNU" anything. It's proprietary, it's a compatibility layer, and it's not complete enough to actually make most Linux based code work. If you think I'm kidding, try any X based application, sshd, and httpd.

Re:Windows 10 interface + Linux kernel

[Kjella]

It depends on what the question is, if it's Microsoft doing a kernel swap while keeping all their binary blobs nVidia-style I imagine they could do it fairly easily. Like if all you want from the Linux kernel is totally generic functionality like allocating memory, mapping IO, catching interrupts etc. with blob drivers and blob libraries providing most the API, kinda like using winetricks to install the real DirectX libraries from Windows. It wouldn't really be Linux as you know it though, it'd be more like that deep, deep buried in OS X is the BSD kernel.

Re:Windows 10 interface + Linux kernel

[Antique Geekmeister]

I'm afraid that this is not remotely possible. The kernel and the critical libraries, for Windows 10 are proprietary and Microsoft does not follow the published API's.

Re:Windows 10 interface + Linux kernel

[Anonymous Coward]

WSL is not "GNU" anything. It's proprietary, it's a compatibility layer, and it's not complete enough to actually make most Linux based code work. If you think I'm kidding, try any X based application, sshd, and httpd.

httpd compiles and runs fine on WSL

Every x apps I tried runs fine as well, didnt try sshd though.

No AI?

[110010001000]

What? No AI engine? What about a smart assistant? This Linus guy obviously doesn't know much about operating systems.

Re:No AI?

[gweihir]

I want cloud integration, were the kernel runs in the cloud!

Re:Windows 10 interface + Linux kernel

[Billly Gates]

WSL is not Linux on Windows. It's GNU/Windows, nothing more.

You know there is more to FOSS than Gnu? It is popular on Debian circles and in groups here but Apache, LibreOffice, clang, and others use more free MIT/BSD style licenses.

Re:Windows 10 interface + Linux kernel

[ArchieBunker]

Give me the FreeBSD kernel with the Windows 7 interface. Clean and no systemd bullshit.

Re:Windows 10 interface + Linux kernel

[q_e_t]

The essential core of the userland tools in what people call Linux us from GNU, which is the point the OP was making.

Re:Forget Linux

[q_e_t]

Freer in what way?

Re: Windows 10 interface + Linux kernel

[spongman]

the ubuntu 16.04 sshd works just fine on wsl. You do have to open a port in the firewall.

Re:Windows 10 interface + Linux kernel

[MSG]

You know there is more to FOSS than Gnu?

The POSIX operating system you are familiar with is GNU/Linux. GNU and Linux provide the (mostly compliant) interfaces that POSIX requires. Everything else is an application.

If you're running GNU on WIndows, then the POSIX OS is GNU/Windows.

Yes, there are applications on top of GNU. There are also applications on top of Windows. Their presence doens't change the name of the OS.

Re: Windows 10 interface + Linux kernel

[Anonymous Coward]

We get Windows with a Linux shell. (Windows Subsystem for Linux).

Re:Windows 10 interface + Linux kernel

[serviscope_minor]

Give me the FreeBSD kernel with the Windows 7 interface. Clean and no systemd bullshit.

Give me the GNU/HURD kernel with the OS/2 interface.

Re:Windows 10 interface + Linux kernel

[Joey Vegetables]

That *BSD* style licenses are or are not "more free" is a matter of perspective. My view is that non-copylefted licenses may do too much to preserve the freedom of those who would embrace/extend/extinguish Free Software, while copylefted licenses encourage the protection of the users' freedoms instead. This distinction would be moot if there were not evil, rent-seeking organizations and individuals who wish to extort the users of formerly free software, plus "laws" that empower and embolden them. Free would simply mean Free. However, there are, and, for as long as there are, I do view copyleft as an vital mechanism to ensure that the most important pieces of software remain free.

Re:Forget Linux

[UnknownSoldier]

> When was the last time FreeBSD came up as the OS of a device?

Sony PS3: FreeBSD
Sony PS4: FreeBSD
Sony Vita: FreeBSD / NetBSD
Nintendo Switch: FreeBSD