The 5.7 kernel is out

https://lwn.net/Articles/821829/

by thuva4 • 6 years ago

210 108 6 years ago

lwn.net

freedomben• 6 years ago

a new, Microsoft-blessed exFAT filesystem implementation

For those of us who lived through the "Linux is cancer" days, we are truly living in interesting times :-)

gsich• 6 years ago

Extend phase.

stefan_• 6 years ago

Ehh, Microsoft only opened exFAT after smartphones dropped all SD card slots and it was no longer possible for them to extort Android device manufacturers with patent licensing.

beefhash• 6 years ago

How long until the patents expire?

pjmlp• 6 years ago

Around here there are plenty of Android phones and tablets with SD card slots to chose from.

fierarul2• 6 years ago

I agree. It's a token gesture. Even if you do have SD card slots android offers to extends storage to that, which I think means it formats the card as EXT4 or something.

Maybe TV manufacturers care about this, but in mobile it's a non-event.

Maybe Microsoft could open-source MSDOS in order for FreeDOS to have a rejuvenating moment.

asveikau2• 6 years ago

after smartphones dropped all SD card slots

Last I checked, only at the high end and Google devices, and substantially less true in non-US markets.

Careful with proclaiming such a bubble to be universal. I am sure in mountain view some PMs make decisions based on "oh but we killed the SD card years ago" but they don't have actual power to do that.

aries1980• 6 years ago

Camera/camcorder makers are huge market, and they use Linux internally. E.g. Sony's stuff: https://oss.sony.net/Products/Linux/

amelius• 6 years ago

Patents should not be allowed on communication protocols and storage formats, or anything with lock-in potential really.

magicalhippo• 6 years ago

after smartphones dropped all SD card slots

So companies like Samsung[1], Sony[2], Motorola[3], Xiaomi[4] and LG[5] don't make smartphones anymore?

[1] https://www.gsmarena.com/samsung_galaxy_s20-10081.php

[2] https://www.gsmarena.com/sony_xperia_1_ii-10096.php

[3] https://www.gsmarena.com/motorola_edge-10133.php

[4] https://www.gsmarena.com/xiaomi_mi_a3-9769.php

[5] https://www.gsmarena.com/lg_k61-10088.php

nickysielicki• 6 years ago

Finally, we get a non-journaled semi-cross-platform filesystem. fat32 just wasn't cutting it... /s

When will we get a proper performant journaled cross platform filesystem that doesn't require additional software on Windows, MacOS, and Linux? I can't believe we still don't have anything like that.

cmurf3• 6 years ago

Btrfs and WinBtrfs. The file system is the journal. https://github.com/maharmstone/btrfs

Previous discussion on WinBtrfs https://news.ycombinator.com/item?id=15177002

Journal is about obviating fsck, just do journal replay to make the fs consistent again. I think for most media like USB sticks, the most platform interoperable file system is UDF. The gotcha that trips most people up I think, is that you shouldn't partition the media device. Just format the whole block device as UDF. I do this with mkuddf using defaults and it works fine across all three operating systems with similar performance as FAT32.

diegocg• 6 years ago

With WSL, shouldn't it be possible to use any native Linux file system?

phone8675309• 6 years ago

Finally, we get a non-journaled semi-cross-platform filesystem. fat32 just wasn't cutting it... /s

Hope you never had to store files larger than 4GB otherwise FAT32 would not work for you. (Before you respond with "that's what archive files / files splitting are for", not every device that takes files on an SD card and supports FAT32 has a way to join files on the fly/a way to read archived files.)

discreditable• 6 years ago

Does UDF count?

minedwiz• 6 years ago

My favourite part of this (source: https://kernelnewbies.org/Linux_5.7#New_exFAT_file_system) is that they originally went with an MS-produced exFAT driver, but dumped it because even MS said the Samsung one was better.

tw04• 6 years ago

Given that Samsung has had years to perfect theirs, and it was an ms licensed implementation, that’s not really surprising tbqh.

ComputerGuru4• 6 years ago

(To clarify, this is in reference to the Linux driver implementations from both and not in comparison to the Windows version.)

juergbi• 6 years ago

That's not accurate. The initial implementation (in Linux 5.4) was also a derivative of Samsung exFAT code, however, it was not contributed by Samsung directly and it was based on a much older version. Samsung then suggested¹ it would be better to upstream their latest exFAT driver (sdfat), which is what has happened now - after improving the code to be suitable for mainline.

¹ https://lore.kernel.org/linux-fsdevel/003701d56d04$470def50$...

bluedino• 6 years ago

Not to rain on the parade but it's exFAT, not NTFS.

Havoc• 6 years ago

How is this for shared storage among dual boot? Terrible idea?

LockAndLol2• 6 years ago

It's a business investment, nothing more. There was a recent story about how M$ treats open-source devs (AppGet vs WinGet). If they could EEE Linux, they'd do it.

KenoFischer2• 6 years ago

Oh, already? I got a patch in to fix PTRACE_SYSEMU on aarch64, but I still have a few patches queued up to fix a few more issues there. I had hoped to get them all into the same release so there is only broken and not broken releases, rather than broken, semi-broken and non-broken releases. Oh well, ptrace on aarch64 has a few other big issues anyway that wouldn't have been mergeable during the RC window, so I suppose I should just try to get that in for 5.8.

_-___________-_• 6 years ago

I hate to break it to you, but there are only semi-broken releases ;)

KenoFischer2• 6 years ago

Well aware, but I was hoping to say to people "Just use 5.7 and it'll be fine". At the moment I have to say "Don't use anything prior to 5.7, unless you have a recent stable backpoint. In either case, some things are fine others aren't - I'll hope to have it fixed by 5.8". Which is fine, but more complicated ;).

pabs3• 6 years ago

Just CC stable@ and get them into the next point release?

exikyut2• 6 years ago

Was curious; quoting https://www.man7.org/linux/man-pages/man2/ptrace.2.html:

PTRACE_SYSEMU, PTRACE_SYSEMU_SINGLESTEP (since Linux 2.6.14)
For PTRACE_SYSEMU, continue and stop on entry to the next system call, which will not be executed. See the documentation on syscall-stops below. For PTRACE_SYSEMU_SINGLESTEP, do the same but also singlestep if not a system call. This call is used by programs like User Mode Linux that want to emulate all the tracee's system calls. The data argument is treated as for PTRACE_CONT. The addr argument is ignored. These requests are currently supported only on x86.

That sounds like a fun pile of dragons. Good luck backspacing the last bit out soon ;)

chungy• 6 years ago

Not yet finished kernelnewbies summary of changes: https://kernelnewbies.org/Linux_5.7

srl• 6 years ago

Not mentioned in the change summaries as far as I can tell, but this contains a (for me critical) fix to the intel graphics drivers (i915): https://gitlab.freedesktop.org/drm/intel/-/issues/1600.

dnr4• 6 years ago

Will this fix my "i915 0000:00:02.0: Resetting rcs0 for hang on rcs0" that's makes my screen freeze for 5 seconds multiple times a day??? It's been driving me slightly crazy.

cmurf3• 6 years ago

Depends on the model. Quite a few were hit in 5.4, including mine, but was fixed in 5.5. I haven't seen a problem with 5.6 or 5.7 so far.

undecisive• 6 years ago

There is a fix for this available. I believe it's in 5.5. My distro (mint) doesn't have it in their kernels utility, so I'm sticking to 5.0.0 until they do.

(Note: It took me ages to find any information about this - https://gitlab.freedesktop.org/drm/intel/-/issues/673 is where the bug is supposedly fixed)

dyingkneepad3• 6 years ago

We can't know the answer. What you are describing is a GPU hang and GPU hangs are a whole class of bugs, it pretty much means "the hardware was misused and now it doesn't know how to move forward, so we reset it" (and the reset is why you get the 5s freeze, your graphics card was essentially rebooted). When this happens, i915.ko creates a nice file in your file system that you can use to create a bug report. Please check dmesg and do so.

paxys• 6 years ago

Linux* kernel

uwu• 6 years ago

*GNU/Linux kernel

ChrisRR• 6 years ago

No, the kernel is just linux

Tyr42• 6 years ago

Kernel of the GNU/Linux system. /s

SCdF• 6 years ago

OK so let's say that I want to run a relatively stable linux install (so not the arch bleeding edge methodology), but I do want to run the latest linux kernel, or as close to it as possible.

What are my choices? Or is the only choice to alter my interpretation of the word "stable"? :-)

pipeep• 6 years ago

See if your preferred distribution has a kernel backport available. Most have backports of the kernel because it has relatively few dependencies and is backwards compatible, making it easy to backport.

E.g. Debian currently has a 5.5 kernel in buster-backports[1]. However, Debian's backported kernel doesn't get the same level of attention and isn't subject to the same security policies as stable releases, so may be lacking some security patches that the stable kernel has.

[1] https://packages.debian.org/buster-backports/linux-image-amd...

spindle• 6 years ago

I find NixOS is good for this - but it depends partly on what userspace programs you need.

You can run NixOS Stable, which is on a six-monthly release and testing schedule, and add the latest kernel, very easily.

tennineeight• 6 years ago

+1. NixOS benefits especially, because a broken kernel/nvidia update never renders your system unusable. You can make changes to all your heart's desire, and if something does break, all you have to do is reboot and select previous configuration

tvb12• 6 years ago

I don't know enough about Linux to offer a complete answer, but I can offer the distribution I use as a suggestion. I've been using Solus (getsol.us/) for a few years and I think it comes close to what you're asking for. The kernel version is currently 5.6.13.

There's no i3 flavor installation image, but you can grab it from the software center/eopkg.

AsyncAwait• 6 years ago

Or is the only choice to alter my interpretation of the word "stable"?

Depends, if your "stable" means has been tested by tens of thousands of users under different scenarios, then running 5.7 right now is by definition not stable.

But if stable means that it is unlikely to cause problems, then even Arch fits that definition, which honestly is also suitable for Windows 10 and macOS at this point.

Anecdotally, I have an Arch install going strong on 5 years now, so even Arch isn't unstable, but it does require paying attention to the announcements.

I'd say a good compromise may be to wait a couple 5.7 point releases in. By then, practically any distro should have it.

Manjaro is a little behind Arch to offer more "stability", there's the OpenSUSE rolling-release option, Solus, Sabayon...

udlv6Yy• 6 years ago

Fedora is the distro you want. It's not bleeding edge, but everything is up-to-date and the kernel is 1-2 releases old. It has a 6 month release cycle and 13 month support window iirc.

I find it a lot more stable than arch and Debian Sid.

St_Alfonzo• 6 years ago

I use Fedora with dnf automatic[1] on all of my machines. Only problem is my Desktop PC, which has a nvidia gpu with the official nvidia driver installed. Every few weeks there is a new kernel version which often only works with the brand new nvidia driver version. I use grub to choose the previous working kernel version until I take the time to download and install the new nvidia driver.

[1]https://dnf.readthedocs.io/en/latest/automatic.html

k_bx• 6 years ago

Ubuntu should have a PPA for that

ronjouch• 6 years ago

Yup:

https://wiki.ubuntu.com/Kernel/MainlineBuilds

https://kernel.ubuntu.com/~kernel-ppa/mainline/?C=M;O=D

Technically, it's not a "PPA" and isn't usable as a source, but you can download builds and install them.

namibj• 6 years ago

Use the distribution of your choice, but with the most recent mainline kernel. The releases are already gated by the LKML process, so don't worry about actual, non-pre kernel releases being too unstable for any normal desktop/workstation usage.

If you don't lack the skills, and are ok with the system telling you when you need to spend some time on transitioning configurations/library versions, consider using Arch Linux. It will demand manual configuration to some extend, but that's mostly just enabling (and rebooting/manually-starting) systemd units for things like a GUI login manager. The benefit is, that you don't rely on backports for bug fixes, allowing things like youtube-dl (interacting with uncooperative websites by ~scraping) to work from the official repositories.

michaelcampbell• 6 years ago

Or is the only choice to alter my interpretation of the word "stable"? :-)

I'm not experienced enough to give you an answer, but a hearty "bravo" for recognizing this as an option. Too many do not.

LinuxBender• 6 years ago

For my own systems, I use CentOS + the ElRepo [1] ML kernel which is about a week behind upstream and uses the same compile options as Redhat uses. I would consider it stable for hobby use. I would not use the ML kernel in production systems unless you had a very specific niche need for it. It's assume you are updating the kernel weekly with ML.

[1] https://www.elrepo.org/tiki/tiki-index.php

panpanna• 6 years ago

Ubuntu HWE?

dcow• 6 years ago

Debian testing is on 5.6 right now. If you can handle a high update volume and are comfortable managing occasional packaging issues, it’s a rather nice “not arch but newish” experience.

fach• 6 years ago

Agreed. I was following Debian mainline releases but grew tired of out of date packages and putting my hopes and dreams into *-backports. After migrating to testing, it's been smooth sailing for well over a year and I've grown to appreciate a rolling release cycle.

leadingthenet2• 6 years ago

Why not Arch? It’s been super stable for me over the past months I’ve been using it, and I configured to have Btrfs take snapshops on every pacman update, for peace of mind.

Haven’t had to roll back once.

teraflop• 6 years ago

On Debian/Ubuntu, you can just download and configure the kernel as normal, and then run "make deb-pkg" to get a .deb that plays nicely with dpkg. You can run whatever kernel you want without having to upgrade any user-space packages or depend on third-party repositories.

ndesaulniers• 6 years ago

So I run Linux mint on my personal laptop, but boot a mainline kernel built from source (so more stable userspace, but cutting edge kernel, as I do a fair amount of kernel development). I think that sounds like they setup your asking about?

Building and installing your own kernel isn't too bad. If you want to give it a shot:

1. download mainline, ie torvalds/linux 2. make -j localmodconfig olddefconfig 3. make -j 4. make -j modules_install install

and reboot.

heavyset_go2• 6 years ago

If you apply the appropriate Debian[1] or Ubuntu kernel patch sets, you can build .deb packages of your kernel and ensure that the appropriate kernel settings and integrations are set for your system.

[1] https://salsa.debian.org/kernel-team/linux/tree/master/debia...

heavyset_go2• 6 years ago

Ubuntu + the latest Ubuntu kernel patch sets. I have a script that downloads the latest kernel source, applies the appropriate patches and builds it if you're interested.

tkuraku• 6 years ago

I would recommend the latest Fedora release. The kernel is updated frequently, seem to be only a couple weeks behind upstream. Fedora has been incredibly stable for me.

philsnow• 6 years ago

Piggybacking on this, do any of the modern distributions treat kexec as a first-class thing? I would prefer to never reboot, but I also don't want to keep running ancient kernels.

juangacovas• 6 years ago

I update CentOS boxes kernel using "ElRepo" and works out fine, for my use case (webservers, etc)

raverbashing• 6 years ago

was this one deprecating the long-standing 80-column limit for kernel source

Amen!

Let's try to get rid of the "80 column limit" dogma and move on to more flexible and realistic limits, in all languages.

dguest• 6 years ago

Apparently it's been moved up to 100, and 80 is still preferred [1].

I have a personal bias toward 80 because I like having multiple buffers side by side (100 would also work).

I always assumed that the hard limit on column length was as much excuse to reject sloppy merge requests as anything else: I don't mind a few 150 column lines, but if every single line has to wrap a 100 column buffer there's something wrong.

[1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/lin...

ed25519FUUU• 6 years ago

+However, never break user-visible strings such as printk messages because +that breaks the ability to grep for them.

Amazing it took this long for such a common sense formatting method to be acceptable.

dyingkneepad3• 6 years ago

This is not a new rule, it's just a rule that was not written there. You can probably find emails from Linus with nsfw-language bashing people for breaking messages.

sersnth• 6 years ago

The rule was already written there. The patch just reformatted things a bit.

badrabbit• 6 years ago

Bpf-lsm is a welcome addition. I can see a new generation of endpoint defense tools built on it.

[Meta] what's with all the auto-collapsed comments.

LockAndLol2• 6 years ago

And once again, no NVIDIA drivers but lots of AMDGPU stuff. How are people still buying NVIDIA?

teruakohatu• 6 years ago

Machine Learning.

einpoklum3• 6 years ago

A much farther developed ecosystem for most/all aspects of computation, including lots of free software. I know, nVIDIA are terrible with their close-sourcing of almost anything and their OpenCL betrayal. But AMD is too much of an inferior option for doing compute work.

The 5.7 kernel is out

Related Stories