Re: [REGRESSION] mainline/master: (build) in vmlinux.unstripped (scripts/Makefile.vmlinux:72) [logspec:kbuil...

On 2/14/26 02:59, KernelCI bot wrote:
> - dashboard: https://d.kernelci.org/i/maestro:b5ed9b51212f5174ccbd9d812fc43e26fcc5361e
> - giturl: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
> - commit HEAD:  f50822fd8675c68d294e89bd102f7b487ca3acd3
> 
> Please include the KernelCI tag when submitting a fix:
> 
> Reported-by: kernelci.org bot <[email protected]>
> 
> 
> Log excerpt:
> =====================================================
>   LD      .tmp_vmlinux2
>   NM      .tmp_vmlinux2.syms
>   KSYMS   .tmp_vmlinux2.kallsyms.S
>   AS      .tmp_vmlinux2.kallsyms.o
>   LD      vmlinux.unstripped
>   NM      System.map
>   SORTTAB vmlinux.unstripped
> Inconsistent kallsyms data
> Try "make KALLSYMS_EXTRA_PASS=1" as a workaround
> 
> =====================================================
> 
> 
> # Builds where the incident occurred:
> 
> ## defconfig+CONFIG_ARM64_64K_PAGES=y on (arm64):

Not my area of expertise, so the following might be misleading, so
handle with care. But FWIW:

We had a similar issue recently in stable; back then Guenther wrote:

"""
> It is probably just a random error. I used to see this once in a while, though
> not recently. Sometimes extra pass(es) are required for kallsyms to stabilize.
> Some architectures are more affected than others, and it also depends on the
> compiler version since different compiler versions generate different sets of
> symbols. In my tests I have KALLSYMS_EXTRA_PASS=1 always enabled for alpha,
> arm, and m68k. I used to have it enabled for powerpc as well, but not anymore
> (if I recall correctly I stopped to see the problem there when I switched to
> gcc 10).
> 
> I even submitted a patch to fix the problem several years ago, but it was
> rejected.
"""https://lore.kernel.org/all/[email protected]/

Is that maybe something similar?

Ciao, Thorsten