Re: [PATCH v5 7/9] igb: Introduce qtest for igb device

[Akihiko Odaki]

On 2023/01/31 22:04, Thomas Huth wrote:
> On 30/01/2023 15.08, Akihiko Odaki wrote:
> > This change is derived from qtest for e1000e device.
> >
> > Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
> > ---
> >   MAINTAINERS                             |   2 +
> >   tests/qtest/fuzz/generic_fuzz_configs.h |   5 +
> >   tests/qtest/igb-test.c                  | 243 ++++++++++++++++++++++++
> >   tests/qtest/libqos/igb.c                | 185 ++++++++++++++++++
> >   tests/qtest/libqos/meson.build          |   1 +
> >   tests/qtest/meson.build                 |   1 +
> >   6 files changed, 437 insertions(+)
> >   create mode 100644 tests/qtest/igb-test.c
> >   create mode 100644 tests/qtest/libqos/igb.c
>
> The igb test code still looks very, very similar to the e1000e test code 
> ... would it be feasible to extend the e1000e code to support the igb 
> device instead of duplicating the code base here?
>
>   Thomas

I doubt it is worth as most of the duplicate code is more like 
boilerplate like registering qos nodes and tests. While such boilerplate 
code is currently identical except renamed identifiers, sharing it with 
e1000e and igb will prevent from changing the device configuration for 
one device (e.g., adding igbvf) or from adding tests specific to one device.

The core of the testing code which e.g. accesses registers is somewhat 
similar but diverged to the extent that it is hard to extract the 
commonalities. It is because the design of igb is diverged from e1000e 
to support more Rx/Tx queues and it changed the format of Rx/Tx descriptor.

Regards,
Akihiko Odaki