What the kernel can answer about RVA23 today

What the apply notice took

On October 9, 2026, a kernel bot reported that Paul Walmsley had applied part of Guodong Xu's version-8 series, riscv: hwprobe: Expose RVA23U64 base behavior, to riscv/for-next. The notice lists commits for the first eight patches, including work on the B extension, Zic64b, and a getter for user pointer-masking length. It then lists three patches with no matching commit: the ISA-bases bitmap and RVA23U64 detection, the cpuinfo lines, and the hwprobe bit itself. The page filing that split isFutureweb.

What those three patches were for

The series cover, as the apply notice quotes it, says a program should be able to check RVA23U64 conformance in one call instead of walking hwprobe and prctl across every mandatory extension. The missing patches are the ones that would build that answer and show it. Taking the earlier patches into for-next does not, on this notice, publish the bit. The desk isOpen Source.

What the October 11 note adds

A speaker summary from the RISC-V microconference at Linux Plumbers, written as a personal note rather than official minutes and updated as of October 10, says the first eight of the eleven patches were picked up on October 9 and that the three core patches remain for version 9. It also says two review points were still open: support semantics that can fall back through older profiles, and a written statement that the new bit does not inherit older implicit behavior of the IMA bits. An interface standard, the standing question under a bit that is not yet a contract, iswhat an interface standard fixes.

Why it matters on this desk

The dependency is a public kernel interface for a ratified profile. A distribution that wants to ship an RVA23U64 build of a library still cannot, on these documents, read one kernel bit and trust it as merged. The software desk, for a compiler or a runtime rather than this patch status, is The short answer this page will stand on is this: Not with the commits the October 9 apply notice lists. That notice took the first eight patches of an eleven-patch series into riscv/for-next. The three that would expose the profile bit have no matching commit.Software.

What Phoronix is, and is not

Phoronix's October 14 headline says Linux is finally making it easier to answer whether a RISC-V system is RVA23. This page treats that headline as an account of the series, not as a commit. The apply notice is the commit list. Where the headline can be read as if the answer already exists, the notice says the answering patches have no matching commit. A version number, which a profile bit is not yet, iswhat a version number has to mean.

What this page will not treat as shipped

for-next is not a release tarball. A patch that is applied there can still be rewritten before a kernel version ships. This page does not name a kernel version that contains RISCV_HWPROBE_BASE_BEHAVIOR_RVA23U64, because the notice does not. It does not tell a user-space program which call to make. The call is the point of the unmerged patches.

What to watch next

A version-9 posting, or an apply notice that assigns commits to the three core patches, would replace the not-yet sentence. A distro announcement that it will ship profile-based library builds only after the bit lands would be a different document. Another writeup of the September 20 cover letter would not. The kernel prepatch desk, for a different tree, isthe 7.3-rc7 prepatch.

What would change the record

A maintainer reply that says the three patches were applied under different subjects would change it, and this page would follow that later notice. A decision to drop the single bit and keep per-extension queries would change it too. A headline that drops the word making would not become the commit list. The October 9 notice is the record of what was taken.

What stays unmerged

What stays specific is an eleven-patch series, eight patches applied to riscv/for-next on October 9, 2026, and three left without commits: detection, cpuinfo, and the hwprobe bit a program would read. The October 11 speaker note agrees that those three remain for a later version. Phoronix's October 14 headline is the account that the work is aimed at the question. It is not, on the documents here, a statement that the question already has a merged answer. Userspace still reconstructs the profile until those commits exist. A board that implements the extensions is a hardware fact this page has not measured. The kernel interface is the only fact these three documents share, and they share it as unfinished. The apply notice also names commits for earlier subjects, including Zic64b and board descriptions. Those commits are in the list. They do not expose the profile. A program that parses cpuinfo text is doing the job the unmerged patches were written to retire, and this page will not pretend that job is over because a headline says the kernel is making it easier. Easier is not merged. for-next is not a tarball. The next document is a commit id for the bit, or a maintainer note that the bit was refused. Until one of those appears, the answer a caller can rely on is the one the October 9 bot printed: no matching commit for the three patches that would have given it.