Re: [PATCH] arm64: perf: Fix the userspace counter example link

[James Clark]

On 06/11/2026 23:45, TAKANO Mitsuhiro wrote:
> Documentation/arch/arm64/perf.rst points at
> tools/perf/arch/arm64/tests/user-events.c, in the text and in the link
> target, but the file does not exist: no file of that name is in
> tools/perf/arch/arm64/tests/, and no commit in the history leading to
> this base has added one.
>
> The userspace register access described in this section is covered by
> the "User space counter reading of instructions" and "User space
> counter reading of cycles" cases in tools/perf/tests/mmap-basic.c.
> That file is the only in-tree userspace test that walks the core PMUs,
> pins the test task to each PMU's CPUs, and selects the PMU by its type
> number when the kernel accepts the extended type.
>
> perf test -v user runs the four cases in that file; the two reading
> cases ask for user access through config1.
>
> Point the text and the link target at that file.
>
> make htmldocs does not fetch external links, so a plain documentation
> build does not report it, but make linkcheckdocs does.  I checked that
> the old link target returns 404 and the new one returns 200.  No kernel
> was built.
>
> The edit and this changelog were written with the help of an LLM
> assistant (Hermes Agent).  I reviewed the diff and the reasoning, and I
> am responsible for the submission.
>
> Fixes: aa1005d15d2a ("Documentation: arm64: Document PMU counters access from userspace")
> Signed-off-by: TAKANO Mitsuhiro <takano32@xxxxxxxxx>
> ---
>   Documentation/arch/arm64/perf.rst | 10 +++++-----
>   1 file changed, 5 insertions(+), 5 deletions(-)
>
> diff --git a/Documentation/arch/arm64/perf.rst b/Documentation/arch/arm64/perf.rst
> index 997fd716b82f..236cb5da425a 100644
> --- a/Documentation/arch/arm64/perf.rst
> +++ b/Documentation/arch/arm64/perf.rst
> @@ -140,9 +140,9 @@ only be enabled when the tasks are pinned to a homogeneous subset of cores and
>   the corresponding PMU instance is opened by specifying the 'type' attribute.
>   The use of generic event types is not supported in this case.
>   
> -Have a look at `tools/perf/arch/arm64/tests/user-events.c`_ for an example. It
> -can be run using the perf tool to check that the access to the registers works
> -correctly from userspace:
> +Have a look at `tools/perf/tests/mmap-basic.c`_ for an example. The two
> +'User space counter reading' tests in it can be run using the perf tool to
> +check that the access to the registers works correctly from userspace:

The section above already says how to use it, and points to the libperf 
test as an example.

So this extra reference isn't necessary and is in the "About 
heterogeneous systems" anyway which doesn't make sense. Unless it's only 
to show how to call sched_setaffinity(), but that's not really in scope.

I would move the sentence about how to run the test into the "How to" 
block above and delete the sentence about user-events.c. mmap-basic.c is 
also a very confusing test IMO so isn't a good example. That saves one 
reference going stale again in the future.

>   
>   .. code-block:: sh
>   
> @@ -160,8 +160,8 @@ userspace must treat the upper 32-bits read from the counter as UNKNOWN. The
>   and should be used to mask the upper bits as needed.
>   
>   .. Links
> -.. _tools/perf/arch/arm64/tests/user-events.c:
> -   https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/tools/perf/arch/arm64/tests/user-events.c
> +.. _tools/perf/tests/mmap-basic.c:
> +   https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/tools/perf/tests/mmap-basic.c
>   .. _tools/lib/perf/tests/test-evsel.c:
>      https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/tools/lib/perf/tests/test-evsel.c
>   
>
> base-commit: d48723711bae9b8cdc56719c448ba42fef28cbee