Linux set to delete about 247,000 lines of deprecated 32-bit ARM platform code

A quarter-million-line deletion is queued

Linux kernel maintainer Arnd Bergmann has assembled a branch that would strip roughly 247,000 lines of code out of the kernel, targeting 32-bit ARM platforms that were deprecated in the 7.3 release. Phoronix covered the work on September 14, 2026, and the branch — named board-remove-7.3 — sits in the kernel's SoC tree, where it waits for a maintainer to pull it.

The change arrives as roughly 300 patches, and it is not purely subtractive: the series adds about 5,000 lines even as it deletes a quarter million. Set against a kernel source tree of roughly 41 million lines, the cut is modest in percentage terms but large in absolute terms.

The drivers are the bulk of it

The composition of the deletion is less obvious than it looks. According to Bergmann, the platform code itself accounts for only about 55,000 of the removed lines. The remaining roughly 192,000 lines are device drivers that become orphans once the platforms are gone, because no other hardware supported by the tree uses them.

"Following the deprecation in Linux 7.3, removing the old ARM platforms will lighten the kernel by around 55k lines of code," Bergmann wrote. "But removing now unused drivers only relevant to those platforms will mean roughly a quarter million lines of code can be removed."

The distinction explains the cleanup's logic. Platform code covers how the kernel boots on one family of chips or boards and how it reaches that hardware's timers, interrupt controllers and buses. Drivers sit above that layer and talk to individual devices. Once the last machine that needed a driver has left the tree, nothing calls it anymore — yet the code still exists and still has to be touched by every tree-wide refactor.

Which platforms are going

The branch removes support for A1100, Footbridge, RISCPC, the Orion, Dove and MV78xx0 family, OMAP24xx, i.MX31, LPC18xx, and the STM32 microcontroller ports. All are 32-bit ARM designs, and the oldest of them predate the era when ARM became the default architecture for phones and embedded boards.

None of this arrived without warning. The platforms were marked deprecated in Linux 7.3, which in kernel terms is the formal announcement that removal is coming.

Timing points to Linux 7.5

The target release is not settled. Phoronix reports that Bergmann raised Linux 7.4 or 7.5 as candidates on the kernel mailing list, with 7.5 the likelier destination. Until someone pulls the branch, it remains a proposal rather than a decision.

The removal also fits an established pattern, as the Tech AI Wire report on dev.to notes. Linux 7.4 dropped BFS, the UnixWare boot file system, on similar grounds: code with no remaining users still imposes real maintenance costs on the people who have to keep it building.

What embedded teams should check now

The practical advice from the reporting is to audit before the pull rather than after. Teams should search their kernel configurations and any out-of-tree drivers for the affected platform names and the matching mach- directories.

Those who do depend on these chips face a choice rather than a crisis. The options laid out are: stay on a long-term stable kernel, which keeps the code for years; carry the platform as an out-of-tree patch set, which means a rebase every release; or move the product to a supported chip. The first is the cheapest, and the third is the only one that ends the problem permanently.

Why it matters

For embedded developers on the affected silicon, this is a deadline: support is leaving mainline, and the 7.3 deprecation notice was the warning to act on. For everyone else the benefit is indirect but concrete — drivers nobody builds still get dragged through every API rename, tree-wide change and static-analysis pass, so deleting them removes work that previously landed on maintainers with no stake in that hardware. The episode also illustrates how the kernel manages its own history: deprecation comes a release in advance, removal arrives as a reviewable branch, and long-term stable kernels remain the escape hatch for anyone who needs old hardware to keep working.