Six Features Not In The Mainline Linux 5.6 Kernel

[hax0r]

aufs, why won't they mainline that, pretty much every live cd distro uses squashfs...

[CommunityMember]

Did Samsung's Exfat driver get merged?

Missed it. The patch series is now up to version 13 (and has no further comments at this point), but missed the linux-next soak test requirements for inclusion into 5.6.

[duby229]

Yes! Must get that futex work in! The problem is that most people don't do mutexes right, it's just waaay beyond the scope of what a game engine -should- do and most don't bother. The few engines that do mutexes in user space pretty much exclusively get it wrong.

[CommunityMember]

aufs, why won't they mainline that, pretty much every live cd distro uses squashfs...

The aufs code base had been criticized as being "dense, unreadable, uncommented code", and the author gave up trying to mainline it quite some time ago. TTBOMK no one else has stepped up to resolve the various issues in the code base that would make it acceptable for mainline.

[indepe]

Yes! Must get that futex work in! The problem is that most people don't do mutexes right, it's just waaay beyond the scope of what a game engine -should- do and most don't bother. The few engines that do mutexes in user space pretty much exclusively get it wrong.

So it can be implemented , for example, in glibc in user space, just like glibc implements semaphores in user space on top of the futex API.

Why in the kernel? Who would call the kernel directly? If they "get it wrong" outside the kernel, they will also "get it wrong" inside the kernel. Better outside.

The whole point of the futex APi is to support implementations in user space, to avoid unnecessary kernel calls.

[ptrwis]

It's a shame inkernel DBUS has stalled, have been looking forward to it for a long time.

Varlink has been merged into systemd (https://github.com/systemd/systemd/c...2e79bf33ff7ae0) and there is already some api for user lookup https://systemd.io/USER_GROUP_API/
Podman is using varlink too. There is also kernel module for varlink on github, but I think the protocol needs some time to mature to even try mainlining it, if ever

[caligula]

The aufs code base had been criticized as being "dense, unreadable, uncommented code", and the author gave up trying to mainline it quite some time ago. TTBOMK no one else has stepped up to resolve the various issues in the code base that would make it acceptable for mainline.

Didn't Docker use aufs as an optional backend?

[Danny3]

Another very important feature missing form 5.6 kernel is network connection over Wifi, which is missing for some high-end motherboards.
Like for example:

https://bbs.archlinux.org/viewtopic.php?id=252676

https://bbs.archlinux.org/viewtopic.php?id=252623

Also I'm wondering why only Arch people are talking about this problem even though this clearly exist in Ubuntu kernels too.
Up until Now I see both Arch for version 19 and Ubuntu for version 20.04 are going back to 5.4.
I hope not everybody is solving this problem by avoiding it (using the older kernel).

[c117152]

Another very important feature missing form 5.6 kernel is network connection over Wifi, which is missing for some high-end motherboards.
Like for example:

https://bbs.archlinux.org/viewtopic.php?id=252676

https://bbs.archlinux.org/viewtopic.php?id=252623

1. It's not a feature. It's a driver regression that got mistakenly backported to stable 5.4.18.
2. It has nothing to do with the release cycle since patches can and should be written and applied by the maintainers.

Also I'm wondering why only Arch people are talking about this problem even though this clearly exist in Ubuntu kernels too.
Up until Now I see both Arch for version 19 and Ubuntu for version 20.04 are going back to 5.4.
I hope not everybody is solving this problem by avoiding it (using the older kernel).

1. It's niche hardware.
2. It doesn't affect the older kernels Ubuntu is packaging.
3. People who really need desktop wifi buy a $10 usb dongle at wallmart so what's the rush?

https://packages.ubuntu.com/search?keywords=linux-image-generic

https://bugzilla.kernel.org/show_bug.cgi?id=206329

[doublez13]

Obligatory ZFS comment.