Linux 7.0-rc3 Released: Two Release Candidates Among “The Biggest in Recent History”
Linux 7.0-rc3 Released: Two Release Candidates Among “The Biggest in Recent History”
- Why Enterprise RAID Rebuilding Succeeds Where Consumer Arrays Fail?
- Linus Torvalds Rejects MMC Subsystem Updates for Linux 7.0: “Complete Garbage”
- The Man Who Maintained Sudo for 30 Years Now Struggles to Fund the Work That Powers Millions of Servers
- How Close Are Quantum Computers to Breaking RSA-2048?
- Why Windows 10 Users Are Flocking to Zorin OS 18 Instead of Linux Mint?
- How to Prevent Ransomware Infection Risks?
- What is the best alternative to Microsoft Office?
Linux 7.0-rc3 Released: Two Release Candidates Among “The Biggest in Recent History”
Linus Torvalds has dropped the third release candidate for Linux 7.0, and the patch set is even larger than the already-oversized rc2 — raising fresh questions about whether the 7.0 cycle will need to extend to an eighth release candidate before stable launch.
Linus Torvalds released Linux 7.0-rc3 on March 8–9, 2026, the third weekly test snapshot in the lead-up to a stable Linux 7.0 launch targeted for mid-April. The release is notable not for any single dramatic feature, but for its sheer volume of changes — a trait that Torvalds himself has flagged as unusual and mildly concerning for this stage of the development cycle.
In a typical kernel release cycle, the first release candidate absorbs the bulk of the merge window’s work, and subsequent candidates progressively shrink as developers focus narrowly on fixes. That pattern has not held for Linux 7.0. Both rc2 and rc3 are, by Torvalds’ own description, among the largest release candidates in recent kernel history. Torvalds attributes part of the bloat to a scheduling artifact: Linux 6.19 ran an extra week with an rc8 release, compressing the timeline and pushing more work into the early 7.0 candidates.
Size in Context: Selftests Drive Much of the Growth
The alarming headline number is softened by what the patch actually contains. According to Torvalds, roughly one-fifth of rc3 — approximately 20% — consists of kernel selftests: automated validation code that developers added to catch regressions before they reach end users. Beyond that, many of the remaining commits are described as trivial: small cleanups, hardware ID additions, and driver quirks. Torvalds was clear that nothing in the remaining changes “looks particularly scary.”
Still, the pure commit count remains higher than is the norm at this point in the cycle, and Torvalds expressed a hope that the pace of incoming patches slows significantly in the coming weeks. There is now a real possibility that the Linux 7.0 cycle will stretch to a rare eighth release candidate — something that would push the stable release beyond the originally projected mid-April window.
Security: IBPB-on-Entry for AMD SEV-SNP VMs
On the security front, rc3 integrates IBPB-on-Entry support for AMD SEV-SNP virtual machines. The feature — supported by AMD’s latest EPYC Zen 5 server processors — forces an Indirect Branch Predictor Barrier (IBPB) upon each entry into a guest virtual machine. This adds a meaningful layer of defense against speculative execution attacks that could otherwise leak sensitive data across VM boundaries. The implementation required only a small number of code changes, making it a compact but high-value addition. A separate patch also addresses AMD SEV guest boot failures that occurred under certain hardware configurations.
On the Intel side, the x86/urgent patches include fixes for Sub-NUMA Clustering (SNC) topology enumeration errors in modern processors such as Granite Rapids X and Clearwater Forest X. SNC topology directly influences how the kernel organizes memory and CPU cores, and the bugs were capable of causing incorrect scheduling and memory usage behavior on affected hardware.
Performance: Critical SLAB Allocator Regression Resolved
Perhaps the most practically significant fix in rc3 is a correction to a severe regression in the kernel’s SLAB memory allocator — the subsystem responsible for managing small, frequently reused blocks of kernel memory. The regression was introduced earlier in the 7.0 development cycle and was found to meaningfully degrade performance on workloads that rely on rapid, repeated kernel memory allocation, such as high-traffic network servers and I/O-intensive applications. The patch restores expected behavior and is considered critical for anyone planning to run Linux 7.0 in production server environments.
Alongside the SLAB fix, rc3 also includes an optimization to the epoll event-polling subsystem, which is widely used in network servers and applications that manage large numbers of simultaneous file descriptors. The change was motivated in part by issues surfaced by newer Intel CPUs, and benchmarks on AMD Zen 2 systems confirm a measurable — if modest — throughput improvement.
Hardware Support Expands Across Laptops and Handhelds
As is typical for each kernel cycle, rc3 extends Linux’s hardware compatibility across a range of consumer and professional devices. New platform driver support has been added for OneXPlayer handheld gaming PC models including the APEX, X1z, X1 Air, and Aokzoe A2 Pro. HP laptops — the Omen 14-fb1xxx and Victus 16-d0xxx — gain support, as do several ASUS models: the G733QS, GX650RX, and FA401UM. Dell 14 Plus and 16 Plus enhancements are also included, along with Tegra238 HDA codec work and G-Mode support for ASUS ROG M18 laptops. Lenovo devices receive additional IDs and quirks as well.
Stability Work: Race Conditions and Memory Safety
Beyond the headlining fixes, a batch of HID (Human Interface Device) patches addresses memory leaks, potential kernel panics, and null pointer dereferences across several input drivers. Bug fixes targeting use-after-free vulnerabilities and race conditions are scattered throughout the patch set as well. These changes are unglamorous but essential to the kind of robust stability that enterprise users and distribution maintainers expect from a kernel major version.
Nearly a fifth of rc3’s patch consists of kernel self-test code, blunting concerns about the raw commit count.
IBPB-on-Entry now enforced for AMD SEV-SNP VMs on Zen 5 EPYC, strengthening virtualization boundary security.
A severe performance regression in the SLAB memory allocator — critical for servers — is now resolved.
Dozens of new device IDs added for ASUS, Dell, HP, Lenovo, OneXPlayer, and Aokzoe hardware.
The elevated patch volume raises the possibility of an eighth release candidate, potentially delaying stable launch past mid-April.
Measurable I/O multiplexing gains on AMD Zen 2 after fixes motivated by Granite Rapids X and Clearwater Forest X bugs.
Outlook
Linux 7.0-rc3 leaves the kernel in a functional but busier-than-ideal state for this point in the release cycle. Torvalds remains publicly measured: the content of the patches does not alarm him, and he sees no single lurking catastrophe. His concern is one of volume and trajectory — a release cycle that has yet to show the traditional deceleration signal that precedes a stable launch. Kernel contributors have been asked to keep testing and, above all, to let the patch rate drop in the weeks ahead.
The stable Linux 7.0 kernel remains on track for a mid-April release if the next few release candidates behave as expected. If they don’t, users and distributions planning around that timeline should monitor the mailing list closely. Source tarballs and full diff patches for rc3 are available directly from the official kernel Git mirror at git.kernel.org.
