net/mlx5e: RX, Fix releasing page_pool pages twice for striding RQ
authorDragos Tatulea <dtatulea@nvidia.com>
Thu, 30 Mar 2023 16:37:00 +0000 (19:37 +0300)
committerSaeed Mahameed <saeedm@nvidia.com>
Fri, 21 Apr 2023 01:35:49 +0000 (18:35 -0700)
commitc8e9090233a70b8bbd8a73521a7a81856695732f
treee5331c75227c5df56b661cc6690bf327762802c6
parentcf14af140a5ad0937d385ce693100f33f02e9c54
net/mlx5e: RX, Fix releasing page_pool pages twice for striding RQ

mlx5e_free_rx_descs is responsible for calling the dealloc_wqe op which
returns pages to the page_pool. This can happen during flush or close.
For XSK, the regular RQ is flushed (when replaced by the XSK RQ) and
also closed later. This is normally not a problem as the wqe list is
empty on a second call to mlx5e_free_rx_descs. However, for striding RQ,
the previously released wqes from the list will appear as missing
and will be released a second time by mlx5e_free_rx_missing_descs.

This patch sets the no release bits on the striding RQ wqes in the
dealloc_wqe op to prevent releasing the pages a second time.

Please note that the bits are set only in the control path during
close and not in the data path.

Fixes: 4c2a13236807 ("net/mlx5e: RX, Defer page release in striding rq for better recycling")
Signed-off-by: Dragos Tatulea <dtatulea@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
drivers/net/ethernet/mellanox/mlx5/core/en_rx.c