[PATCH v2 2/3] net/mlx5: fix sync meter processing in HWS setup
Raslan Darawsheh
rasland at nvidia.com
Tue Mar 12 09:03:43 CET 2024
Hi,
> -----Original Message-----
> From: Gregory Etelson <getelson at nvidia.com>
> Sent: Thursday, March 7, 2024 12:19 PM
> To: dev at dpdk.org
> Cc: Gregory Etelson <getelson at nvidia.com>; Maayan Kashani
> <mkashani at nvidia.com>; Raslan Darawsheh <rasland at nvidia.com>;
> stable at dpdk.org; Dariusz Sosnowski <dsosnowski at nvidia.com>; Slava
> Ovsiienko <viacheslavo at nvidia.com>; Ori Kam <orika at nvidia.com>;
> Suanming Mou <suanmingm at nvidia.com>; Matan Azrad
> <matan at nvidia.com>; Alexander Kozyrev <akozyrev at nvidia.com>
> Subject: [PATCH v2 2/3] net/mlx5: fix sync meter processing in HWS setup
>
> Synchronous calls for meter ASO try to pull pending completions from CQ,
> submit WR and return to caller. That avoids delays between WR post and HW
> response.
> If the template API was activated, PMD will use control queue for sync
> operations.
>
> PMD has different formats for the `user_data` context in sync and async meter
> ASO calls.
> PMD port destruction procedure submits async operations to the port control
> queue and polls the queue CQs to clean HW responses.
>
> Port destruction can pull a meter ASO completion from control CQ.
> Such completion has sync format, but was processed by async handler.
>
> The patch implements sync meter ASO interface with async calls in the
> template API environment.
>
> Fixes: 48fbb0e93d06 ("net/mlx5: support flow meter mark indirect action
> with HWS")
>
> Cc: stable at dpdk.org
>
> Signed-off-by: Gregory Etelson <getelson at nvidia.com>
> Acked-by: Dariusz Sosnowski <dsosnowski at nvidia.com>
> ---
> drivers/net/mlx5/mlx5.h | 35 +++++-
> drivers/net/mlx5/mlx5_flow_aso.c | 178 ++++++++++++++++++-----------
> drivers/net/mlx5/mlx5_flow_hw.c | 99 ++++++++--------
> drivers/net/mlx5/mlx5_flow_meter.c | 27 +++--
> 4 files changed, 216 insertions(+), 123 deletions(-)
Series applied to next-net-mlx,
Kindest regards
Raslan Darawsheh
More information about the stable
mailing list