Re: [PATCH v11 0/9] lib/pmu: cleanups and trace integration

[David Marchand <david.marchand@redhat.com>]

Hello,

On Fri, 24 Oct 2025 at 07:49, Tomasz Duszynski <tduszynski@marvell.com> wrote:
>
> This series does some cleanup and refactoring around the rc1 code like: trimming unused headers,
> switching to callbacks for per-arch handling, and adding trace support. It also re-enables existing
> base test to help catch reported issues on some architectures.
>
> v11:
> - rebase series
> - hide calls to experimental syms in inline helpers
> v10:
> - fix build without ALLOW_EXPERIMENTAL_API
> - move rte_pmu_tread_read() registration to avoid MSVC linker issues
> v9:
> - properly rebase patch integrating pmu and trace
> v8:
> - export __rte_pmu_trace_read from library itself to avoid build issues
>   with msvc linker
> v7:
> - change test return value
> v6:
> - add more logs to functional test
> - skip test in case of setup failure, user must make sure
>   system is properly configured to get valid results
> v5:
> - add missing patch that quiesces chincs check
> v4:
> - change fast test so that it won't fail on misconfigured system
> - fix compilation on windows
> v3:
> - do not export __rte_pmu_trace_read because that breaks compilation
>   on windows - script generating map files does not handle conditional
>   compilation
> - skip testing if paranoia is at wrong level
> v2:
> - explicitly check against NULL
> - make pmu lib optional by checking if dpdk config has RTE_LIB_PMU
>

Strange output in the cover letter.

Shortlog shows 10 patches:
> Tomasz Duszynski (10):

Followed by 8 lines:
>   trace: change scope of conditional block
>   lib/pmu: export only necessary arch headers
>   lib/pmu: reimplement per-arch ops as callbacks
>   lib/pmu: do not try enabling perf counter access on arm64
>   lib/pmu: use build system defined RTE_LIB_PMU macro
>   test/pmu: enable test
>   trace: add PMU
>   lib/pmu: fix out-of-bound access

But the series has 9 patches.

In any case,
- patch 2 "trace: change scope of conditional block" is unneeded, I
see nothing wrong with current code. I tried stopping at various
points of the series, no build issue,
- patch 7 has a comment from Morten,
- patch 8 has comments from me,
- patch 9 is vague, what is this about? Fixing coverity or some static
analysis tool bug report?

I applied the rest of the series, as other patches look valid fixes / cleanups.


-- 
David Marchand