Re: [patch net-next v2 4/9] devlink: remove reporters_lock

[Jiri Pirko <jiri@resnulli.us>]

Sun, Jan 08, 2023 at 05:28:50PM CET, idosch@nvidia.com wrote:
>On Sat, Jan 07, 2023 at 11:11:45AM +0100, Jiri Pirko wrote:
>> From: Jiri Pirko <jiri@nvidia.com>
>> 
>> Similar to other devlink objects, convert the reporters list to be
>> protected by devlink instance lock. Alongside add unlocked versions
>> of health reporter create functions and remove port-specific destroy
>> function which is no longer needed.
>> 
>> Signed-off-by: Jiri Pirko <jiri@nvidia.com>
>> ---
>>  .../ethernet/mellanox/mlx5/core/en/health.c   |  12 ++
>>  .../mellanox/mlx5/core/en/reporter_rx.c       |   6 +-
>>  .../mellanox/mlx5/core/en/reporter_tx.c       |   6 +-
>>  drivers/net/ethernet/mellanox/mlxsw/core.c    |   8 +-
>>  drivers/net/netdevsim/health.c                |  20 +--
>>  include/net/devlink.h                         |  20 +--
>>  net/devlink/core.c                            |   2 -
>>  net/devlink/devl_internal.h                   |   1 -
>>  net/devlink/leftover.c                        | 131 +++++++-----------
>>  9 files changed, 96 insertions(+), 110 deletions(-)
>
>This is quite difficult to review because there are multiple changes
>squashed into one patch:
>
>1. Addition of locked versions of both device and port health reporter
>while refactoring the code to share code paths.
>
>2. Removal of the reporters mutex.
>
>3. Partial conversion of drivers to use the locked APIs. The conversion
>of mlxsw and netdevsim is trivial because they hold the instance lock
>during probe, but the conversion of mlx5 is less trivial. I would split
>it into a separate patch.

Yeah, I was thinking about splitting this, however since the new lock
needs to be used on locked path with different helpers, it is not easy
to find a good split, was a bit lazy to do that. Anyway, let me think
and try again. Thanks!