# Re: \[PATCH v2 16/58\] objtool/klp: Add test for cold function halves

- From: Song Liu \<song@kernel.org\>
- Date: 2026-09-14 20:55:15 
- List: live-patching
- Page: https://ratatoskr.run/live-patching/2026/09/17575680
- Thread: https://ratatoskr.run/sashiko-reviews/2026/09/17575680/t.md

```text
On Sun, Sep 13, 2026 at 11:38 PM [ref] wrote:
>
> Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
> - [Medium] The test assertion for the carried cold function incorrectly passes with `llvm-readelf` when the cold half is left behind.
> --
>
> commit a04e65cdf7036d7ab3a10a22fc0efb8a0c3873bd
> Author: Puranjay Mohan [ref]
>
> objtool/klp: Add test for cold function halves
>
> The compiler splits unlikely code into a separate foo.cold symbol. Both
> halves are the same function and both belong in the livepatch. This adds
> a test to ensure cold paths branching into unpatched code are handled.
>
> > diff --git a/tools/objtool/tests/generic/test-cold-function.sh b/tools/objtool/tests/generic/test-cold-function.sh
> > new file mode 100755
> > index 0000000000000..dd4a7297af044
> > --- /dev/null
> > +++ b/tools/objtool/tests/generic/test-cold-function.sh
>
> [ ... ]
>
> > +assert_patched target
> > +# Not a bare name match: had the cold half been left behind, the branch to
> > +# it would appear as an undefined .klp.sym.vmlinux.target.cold, whose name
> > +# contains the one being looked for.  Require a defined symbol.
> > +out_symbols | awk '$7 != "UND" && $8 ~ /target\.cold/ { found = 1 }
> > +     END { exit !found }' ||
> > +     fail "cold half was not carried into the patch"
>
> [Severity: Medium]
> Will this awk script correctly detect missing cold halves when the test is
> run with llvm-readelf?
>
> When the cold half is left behind, klp diff converts unresolved references
> into SHN_LIVEPATCH symbols. llvm-readelf formats these as:
>
> OS[0xff20] .klp.sym.vmlinux.target.cold,0
>
> Because there is no space between the section name and the symbol name in
> this output format, $7 evaluates to "OS[0xff20]" (which is not "UND") and $8
> matches the /target\.cold/ regex.
>
> Does this condition evaluate to true and cause the test to falsely pass when
> the cold half is actually left behind?

Seems to be an issue introduced in v2. Fixing it in v3.

Thanks,
Song
```
