Skip to main content


Highlights from the # changes(¹) merged for # # 5.12:

* Support for bitwise atomics(²)

* Support for directly accessing kernel module variables via module BTF info(³)

* Allow bpf to attach to bare tracepoints(⁴)

* allow variable-offset stack access(⁵)

1/



[bookmark=https://twitter.com/maelver/status/1366308283557023744]KFENCE, a low-overhead sampling-based memory safety error detector for the Linux kernel, was merged for Linux 5.12. @Glider [/bookmark]



# # 5.12-rc1 is out:

"[…] without electricity for 6 days[…] considering just extending the merge window[…] didn't do that[…] To a large part because[…] everything was nicely lined up[…] partly this is also because 5.12 is a smaller[…]"

lore.kernel.org/lkml/CAHk-=wjU…



[bookmark=https://twitter.com/Friz_N/status/1366048944703340546]Wrote about a futex+vfs kernel UAF reported by @tehjh one year ago, in particular how huge the patch gap ended up being on "enterprise" distros: [/bookmark]



The merge window for # # 5.12 is not yet closed, but normally will end in a few hours from now.

Seems it'll be one of the smaller ones in recently history.

But I'd say that wasn't due to the power outage that Linus had to deal with; I'd say that's due to…



Highlights from the big # merge(¹) for # # 5.12:

* threaded NAPI polling patches, which can improve the performance(²)

* netfilter allows "ownership" of specific tables(³)

* various improvements for multipath TCP (MPTCP), like priorities for subflows(⁴)

1/



In # # 5.12, the integrity measurement architecture (IMA) thx to a Microsoft employee will be able to measure various bits of data within the kernel (like the SELinux policy) to ensure they have not been tampered with:

git.kernel.org/torvalds/c/d64…

lore.kernel.org/lkml/202101080…



Another new # feature got merged just hours after I sent below tweet:

"[…] This converts the io-wq workers to be forked off the tasks in question instead of being kernel threads that assume […] kills > 400 loc […]"

git.kernel.org/torvalds/c/569…

twitter.com/axboe/status/1…
[bookmark=https://twitter.com/kernellogger/status/1365607165545840644]Highlights from big # merge for # # 5.12:

* LOOKUP_CACHED, allowing VFS pathname lookups without blocking(¹)

* integration with memcg/memory control groups(²)

* "[…] 10-20% of speedups on workloads that mostly are inline […]"

1/

[/bookmark]



Highlights from big # merge for # # 5.12:

* LOOKUP_CACHED, allowing VFS pathname lookups without blocking(¹)

* integration with memcg/memory control groups(²)

* "[…] 10-20% of speedups on workloads that mostly are inline […]"

1/



Highlights from the big # merge for # 5.12:

* Scalability improvements for the new MMU [added in # 5.10(¹)]

* Support for userspace to emulate Xen hypercalls [allow hosting Xen HVM
guests in KVM(²)]

Changed were submitted by @fagiolinux:
git.kernel.org/torvalds/c/3e1…




[bookmark=https://twitter.com/0x00626f52/status/1365121045967958016]Upstream kernel support getting sorted for arm things before they even ship.. this is how it should have been years ago! [/bookmark]



The # article on
development statistics for # # 5.11 is now available for free, too:
lwn.net/Articles/84583…

"[…]# […] home to an increasing number of core-kernel developers […] contributes far more patches to the kernel directory than any other […]"




Highlights from the big graphic driver (aka #) merge merge for # # 5.12:

#: Sienna Cichild overdrive support, GPU reset on navy flounder/vangogh

#: BCM2711 (the one used on the # Model B): DSI1 support, enable 10/12 bpc outputs

(1/2)



The kcmp syscall in the # # 5.12 is now available independently of checkpoint/restore (aka @__criu__). The LWNnet text about this is now available for non-subscribers: lwn.net/Articles/84544…

git.kernel.org/torvalds/c/4b5…

git.kernel.org/torvalds/c/bfe… (tagged for stable backp.)



[bookmark=https://twitter.com/linuxfoundation/status/1364609779000311810]NEWS: @Google funds Linux kernel developers to focus exclusively on security: https://t.co/2KE3hDAYaM # #kernel # # # @nathanchance[/bookmark]

Security is wonderful. Given Google's reputation, what effect will their contribution have on PRIVACY?

They want to make Linux proprietary, I think.



With the "platform profile support" that was merged for # # 5.12 it becomes possible to tell certain systems if the user prefers things like "max performance", "max battery life", or "run quietly".

For details see this documentation change:
git.kernel.org/torvalds/c/8e0…

Jure Repinc reshared this.



[bookmark=https://twitter.com/qeole/status/1364535272613027841]Although the main focus is on eBPF news, the latest issue of the eBPF Updates (# has a “Did you know?” section about the size limit for eBPF programs. I hope you'll enjoy it! 🐝 [/bookmark]
# #4)



The idmapped mounts to expose files or directories with a different ownership to a container was now merged for # # 5.12:
git.kernel.org/torvalds/c/7d6…

For details see also this # article:
lwn.net/Articles/83756

Congrats and thx for your work on this, @brau_ner!
[bookmark=https://twitter.com/kernellogger/status/1354689489512390657]The # # might finally get a proper solution to expose files or directories with different ownership to a container: The idmapped mounts patches(¹) from @brau_ner are now in linux-next and he plans to submit them for inclusion in 5.12.

(¹) lore.kernel.org/linux-fsdevel/…

[/bookmark]