Linux 3443 Published by

Three Linux kernel point releases, 7.2.11, 6.18.57, and 6.12.113, has been released today, within minutes of each other and signed off by Greg Kroah-Hartman. All three carry the same shared fixes for the Steam controller driver, sched_ext, AMD display, and the CAN subsystem, but each branch also brings its own standout patches. The newest LTS (6.18.57) leads with a fix for the XFS "refluxfs" data-corruption bug and s390 zcrypt memory-disclosure hardening, while 6.12.113 includes a netfilter fix that stops OpenWrt routers from dropping WiFi after ~300 seconds. The updates also carry "Assisted-by: Claude" tags, underscoring how AI-assisted review has become routine, and the advice is the same as always: patch the branch you run and reboot.





Three Linux Kernels Drop in Minutes, All Patched for One Day

Point releases for 7.2.11, 6.18.57, and 6.12.113 shipped today, carrying fixes from XFS data-corruption to a bug that drops OpenWrt routers.

Three separate Linux kernel releases went out the same morning of today. They landed within minutes of each other. Greg Kroah-Hartman signed off all three, with testing pulled from the usual roster of stable maintainers including Shuah Khan, Miguel Ojeda, and Florian Fainelli.

Here's the part that trips people up: "the Linux kernel" isn't a single version number floating around. It's several tracks running at the same time. There's mainline, the rolling release Linus Torvalds keeps up with a new major roughly every nine to ten weeks. There are LTS series, the even-numbered releases Greg's team patches for years. Then there are the point releases, the third number in versions like .11, stacking fixes on top of a base.

So three kernels in one morning isn't a fluke. It's the normal heartbeat. 7.2.11 rides the newest mainline, 6.18.57 is the freshest LTS, and 6.12.113 is the old standby. Each targets a different crowd.

Kernel

What actually changed

Some bugs hit multiple series at once, so the same upstream patch shows up in every changelog. The Steam controller driver got a multi-part cleanup from Vicki Pfau, reworking locking, registration, and teardown order in the HID:steam driver. Fairly funny that a gamepad driver lands inside server-focused LTS kernels, but that's the beauty of a shared codebase.

Kuba Piecuch added a per-task sequence counter to sched_ext, the BPF-extensible scheduler, so stale dispatches get ignored reliably. Adrian Betschart fixed a corrupted 4K sink problem in AMD display, and Oliver Hartkopp closed a CAN-frame duplicate-detection regression under RPS.

But the interesting stuff lives in what differs between the three.

The big one for 6.18.57, the newest LTS, is a bug people already nicknamed "refluxfs." An XFS problem in the exchange-range-to-eof feature could trick the filesystem into creating shared blocks past the end of a file, which is impossible for real written data, and the mis-swap let an extending write corrupt another file holding that block. Darrick J. Wong wrote the main fix, with Norbert Szetei adding a companion patch, both from a security researcher at doyensec. "Refluxfs" isn't a great name, but it's memorable.

Same release also gets four fixes on the s390 zcrypt driver from IBM's Harald Freudenberger. Up to three bytes of uninitialized kernel memory could leak to crypto-card firmware. Freudenberger padded the messages with zeros and swapped fragile struct overlays for explicit ASN.1 parsing plus real length and overflow checks. That closes an information-disclosure path and a buffer-overflow path in one go.

For 7.2.11, the newest mainline, the batch is a long stream of memory, I/O, and networking fixes. io_uring had a regression that made unprivileged users hit ENOMEM while creating many rings. PostgreSQL developers reportedly tripped this in testing, and the fix restores pre-6.14 behavior. Lorenzo Stoakes fixed an mremap locked_vm leak, and a networking haul spanning vsock, openvswitch, mlxsw, and veth rounds out the batch.

It's the same 7.2 base where Torvalds again complained the week's changes were bigger than he'd have wished, a size he puts down to AI-assisted review. That cycle pulled more than 400 fixes from over 230 contributors.

6.12.113, the workhorse, has its own crowd-pleasers, and one hits a very wide audience.

netfilter had a bug where conntrack timeouts for offloaded MediaTek WiFi links never got extended, so connections dropped after roughly 300 seconds on hardware using network offloading. Adrian Bente traced the root cause to a cmpxchg comparing a relative expiry against an absolute timestamp, which never matched. Florian Westphal added a companion fix for the offload bit. This is the one that matters if you run OpenWrt, where those WED-offloaded links bite home setups all the time.

Osama Abdelkader added a panthor bounds check. If firmware reported fewer queue slots than the maximum, unprivileged userspace could pull an uninitialized interface and crash the driver. A NULL deref in the mt7603 WiFi driver rounds out the GPU and wireless fixes here.

Why this is worth your attention

The security context makes all of this timely. Ubuntu rolled out several kernel USNs on October 9, days before the kernels shipped. Each covered a batch of CVEs alongside "hundreds of others."

Stable changelogs usually list commit hashes and Fixes tags without naming CVEs directly. That's why Ubuntu, SUSE, and Debian are the ones that map point releases to actual CVE numbers. The fix is simple enough: update to the newest point release for your branch, then reboot. Every issue here is already fixed somewhere.

Three things stand out from the batch.

First, the "new normal" Torvalds mentioned is real. Several fixes across these releases carry "Assisted-by: Claude" attributions. AI tooling is now baked into how the kernel gets reviewed, and that's a genuine cultural shift that isn't fading.

Second, storage and crypto integrity are the highest-stakes items. The XFS "refluxfs" corruption and the s390 zcrypt memory leak both touch data you can actually lose or expose. Those are the patches that keep enterprise and cloud teams up at night.

Third, remember to reboot. Kernel patches don't take effect until a restart. For systems where downtime costs money, live-patching is worth a real look.

Linux kernel 7.2.11 released

Linux kernel version 7.2.11 is now available:

Full source: https://cdn.kernel.org/pub/linux/kernel/v7.x/linux-7.2.11.tar.xz
Patch: https://cdn.kernel.org/pub/linux/kernel/v7.x/patch-7.2.11.xz
PGP Signature: https://cdn.kernel.org/pub/linux/kernel/v7.x/linux-7.2.11.tar.sign

You can view the summary of the changes at the following URL:
https://git.kernel.org/stable/ds/v7.2.11/v7.2.10

Linux kernel 6.18.57 released

Linux kernel version 6.18.57 is now available:

Full source: https://cdn.kernel.org/pub/linux/kernel/v6.x/linux-6.18.57.tar.xz
Patch: https://cdn.kernel.org/pub/linux/kernel/v6.x/patch-6.18.57.xz
PGP Signature: https://cdn.kernel.org/pub/linux/kernel/v6.x/linux-6.18.57.tar.sign

You can view the summary of the changes at the following URL:
https://git.kernel.org/stable/ds/v6.18.57/v6.18.56

Linux kernel 6.12.113 released

Linux kernel version 6.12.113 is now available:

Full source: https://cdn.kernel.org/pub/linux/kernel/v6.x/linux-6.12.113.tar.xz
Patch: https://cdn.kernel.org/pub/linux/kernel/v6.x/patch-6.12.113.xz
PGP Signature: https://cdn.kernel.org/pub/linux/kernel/v6.x/linux-6.12.113.tar.sign

You can view the summary of the changes at the following URL:
https://git.kernel.org/stable/ds/v6.12.113/v6.12.112