Re: [PATCH net-next v2 1/2] selftests/ptr_ring: add benchmark application for ptr_ring

On 2021/6/27 14:09, Michael S. Tsirkin wrote:

On Fri, Jun 25, 2021 at 11:52:16AM +0800, Yunsheng Lin wrote:

quoted

On 2021/6/25 11:36, Jason Wang wrote:

quoted

在 2021/6/25 上午11:18, Yunsheng Lin 写道:

quoted

Currently ptr_ring selftest is embedded within the virtio
selftest, which involves some specific virtio operation,
such as notifying and kicking.

As ptr_ring has been used by various subsystems, it deserves
it's owner's selftest in order to benchmark different usecase
of ptr_ring, such as page pool and pfifo_fast qdisc.

So add a simple application to benchmark ptr_ring performance.
Currently two test mode is supported:
Mode 0: Both enqueuing and dequeuing is done in a single thread,
         it is called simple test mode in the test app.
Mode 1: Enqueuing and dequeuing is done in different thread
         concurrently, also known as SPSC(single-producer/
         single-consumer) test.

The multi-producer/single-consumer test for pfifo_fast case is
not added yet, which can be added if using CAS atomic operation
to enable lockless multi-producer is proved to be better than
using r->producer_lock.

Only supported on x86 and arm64 for now.

Signed-off-by: Yunsheng Lin <redacted>
---
  MAINTAINERS                                      |   5 +
  tools/testing/selftests/ptr_ring/Makefile        |   6 +
  tools/testing/selftests/ptr_ring/ptr_ring_test.c | 249 +++++++++++++++++++++++
  tools/testing/selftests/ptr_ring/ptr_ring_test.h | 150 ++++++++++++++
  4 files changed, 410 insertions(+)

Why can't you simply reuse tools/virtio/ringtest?

The main reason is stated in the commit log:
"Currently ptr_ring selftest is embedded within the virtio
selftest, which involves some specific virtio operation,
such as notifying and kicking.

As ptr_ring has been used by various subsystems, it deserves
it's owner's selftest in order to benchmark different usecase
of ptr_ring, such as page pool and pfifo_fast qdisc."

More specificly in tools/virtio/ringtest/main.c and
tools/virtio/ringtest/ptr_ring.c, there are a lot of operation
related to virtio usecase, such as start_guest(), start_host(),
poll_used(), notify() or kick() ....., so it makes more sense
to add a generic selftest for ptr ring as it is not only used
by virtio now.

Okay that answers why you didn't just run main.c
but why not add a new test under tools/virtio/ringtest/
reusing the rest of infrastructure that you currently copied?

Actually, my first attempt was to reuse the infrastructure in
tools/virtio/ or tools/virtio/ringtest/, and neither of them
was able to be compiled in the latest kernel.

And then I read through the code to try fixing the compile error,
I found that the testcase under tools/virtio/ is coupled deeply
to virtio as explained above, which was difficult to read for
someone who is not fimiliar with virtio.

So I searched for how testing is supposed to be added in the kernel,
it seems it is more common to add the testing in tools/testing or
tools/testing/selftest, and ptr ring is not only used by virtio now,
so it seems more appropriate to add a sperate testing for virtio by
instinct.

Most of tools/virtio/ is to do testing related to virtio testing, IMHO,
most of them are better to be in tools/testing/selftest. Even if most of
virtio testing is moved to tools/testing/selftest, I think it makes more
sense to decouple the virtio testing to ptr_ring testing too if we can
find some mechanism to share the abstract infrastructure in ptr_ring_test.h
for both virtio and ptr_ring testing.

quoted

quoted

Thanks


.

.