Re: [PATCH] arch: fix broken BuildID for arm64 and riscv

From: Conor Dooley
Date: Mon Dec 26 2022 - 17:10:20 EST


Hey Masahiro,

On Tue, Dec 27, 2022 at 04:06:35AM +0900, Masahiro Yamada wrote:
> On Mon, Dec 26, 2022 at 12:18 AM Conor Dooley <conor@xxxxxxxxxx> wrote:
> > On Sun, Dec 25, 2022 at 04:27:51AM +0900, Masahiro Yamada wrote:
> > > Dennis Gilmore reports that the BuildID is missing in the arm64 vmlinux
> > > since commit 994b7ac1697b ("arm64: remove special treatment for the
> > > link order of head.o").
> > >
> > > The issue is that the type of .notes section, which contains the BuildID,
> > > changed from NOTES to PROGBITS.
> > >
> > > Ard Biesheuvel figured out that whichever object gets linked first gets
> > > to decide the type of a section, and the PROGBITS type is the result of
> > > the compiler emitting .note.GNU-stack as PROGBITS rather than NOTE.
> > >
> > > While Ard provided a fix for arm64, I want to fix this globally because
> > > the same issue is happening on riscv since commit 2348e6bf4421 ("riscv:
> > > remove special treatment for the link order of head.o"). This problem
> > > will happen in general for other architectures if they start to drop
> > > unneeded entries from scripts/head-object-list.txt.
> > >
> > > Discard .note.GNU-stack in include/asm-generic/vmlinux.lds.h.
> > >
> > > riscv needs to change its linker script so that DISCARDS comes before
> > > the .notes section.
> >
> > No idea why I decided to look at patchwork today, but this seems to
> > break the build on RISC-V, there's a whole load of the following in the
> > output:
> > `.LPFE4' referenced in section `__patchable_function_entries' of kernel/trace/trace_selftest_dynamic.o: defined in discarded section `.text.exit' of kernel/trace/trace_selftest_dynamic.o
> >
> > I assume that's what's doing it, but given the day that's in it - I
> > haven't looked into this any further, nor gone and fished the logs out of
> > the builder.
>
>
> arch/riscv/kernel/vmlinux.lds.S clearly says:
> /* we have to discard exit text and such at runtime, not link time */
>
> riscv already relies on the linker not discarding EXIT_{TEXT,DATA}
> so riscv should define RUNTIME_DISCARD_EXIT like x86, arm64.

Huh, fair enough. The diff for that appears to be trivial, but I was not
able to correctly determine a fixes tag. I may have erred in my
history-diving, but it's a wee bit hard to determine the correct fixes
tag. That comment about runtime discards appears to date back to
commit fbe934d69eb7 ("RISC-V: Build Infrastructure") in 2017 -
apparently pre-dating the addition of the define in the first place.

Commit 84d5f77fc2ee ("x86, vmlinux.lds: Add RUNTIME_DISCARD_EXIT to
generic DISCARDS") added it to x86 but not to arm64 - but it seems like
it was added to arm during a later reword. Does that make 84d5f77fc2ee
the correct one to mark it as a fix of & riscv was just overlooked when
the define was added?


> Anyway, I came up with a simpler patch, so I do not need to
> touch around arch linker scripts.
>
> I sent v2.
> https://lore.kernel.org/lkml/20221226184537.744960-1-masahiroy@xxxxxxxxxx/T/#u

Sweet, thanks. Hopefully the automation likes that version better :)

Attachment: signature.asc
Description: PGP signature