[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <ed7a2ffa-bfc9-4276-8776-89cafa546ef8@amd.com>
Date: Thu, 14 Dec 2023 16:05:45 -0800
From: "Nelson, Shannon" <shannon.nelson@....com>
To: edward.cree@....com, linux-net-drivers@....com, davem@...emloft.net,
kuba@...nel.org, pabeni@...hat.com, edumazet@...gle.com
Cc: Edward Cree <ecree.xilinx@...il.com>, netdev@...r.kernel.org,
habetsm.xilinx@...il.com, Jonathan Cooper <jonathan.s.cooper@....com>
Subject: Re: [PATCH net-next 6/7] sfc: add debugfs entries for filter table
status
On 12/11/2023 9:18 AM, edward.cree@....com wrote:
>
> From: Edward Cree <ecree.xilinx@...il.com>
>
> Filter table management is complicated by the possibility of overflow
> kicking us into a promiscuous fallback for either unicast or multicast.
> Expose the internal flags that drive this.
> Since the table state (efx->filter_state) has a separate, shorter
> lifetime than struct efx_nic, put its debugfs nodes in a subdirectory
> (efx->filter_state->debug_dir) so that they can be cleaned up easily
> before the filter_state is freed.
>
> Reviewed-by: Jonathan Cooper <jonathan.s.cooper@....com>
> Signed-off-by: Edward Cree <ecree.xilinx@...il.com>
> ---
> drivers/net/ethernet/sfc/debugfs.h | 4 ++++
> drivers/net/ethernet/sfc/mcdi_filters.c | 18 ++++++++++++++++++
> drivers/net/ethernet/sfc/mcdi_filters.h | 4 ++++
> 3 files changed, 26 insertions(+)
>
> diff --git a/drivers/net/ethernet/sfc/debugfs.h b/drivers/net/ethernet/sfc/debugfs.h
> index 3e8d2e2b5bad..7a96f3798cbd 100644
> --- a/drivers/net/ethernet/sfc/debugfs.h
> +++ b/drivers/net/ethernet/sfc/debugfs.h
> @@ -39,6 +39,10 @@
> * index. (This may differ from both the kernel core TX queue index and
> * the hardware queue label of the TXQ.)
> * The directory will contain a symlink to the owning channel.
> + *
> + * * "filters/" (&efx_mcdi_filter_table.debug_dir).
> + * This contains parameter files for the NIC receive filter table
> + * (@efx->filter_state).
> */
>
> void efx_fini_debugfs_netdev(struct net_device *net_dev);
> diff --git a/drivers/net/ethernet/sfc/mcdi_filters.c b/drivers/net/ethernet/sfc/mcdi_filters.c
> index 4ff6586116ee..a4ab45082c8f 100644
> --- a/drivers/net/ethernet/sfc/mcdi_filters.c
> +++ b/drivers/net/ethernet/sfc/mcdi_filters.c
> @@ -1348,6 +1348,20 @@ int efx_mcdi_filter_table_probe(struct efx_nic *efx, bool multicast_chaining)
> INIT_LIST_HEAD(&table->vlan_list);
> init_rwsem(&table->lock);
>
> +#ifdef CONFIG_DEBUG_FS
> + table->debug_dir = debugfs_create_dir("filters", efx->debug_dir);
> + debugfs_create_bool("uc_promisc", 0444, table->debug_dir,
> + &table->uc_promisc);
> + debugfs_create_bool("mc_promisc", 0444, table->debug_dir,
> + &table->mc_promisc);
> + debugfs_create_bool("mc_promisc_last", 0444, table->debug_dir,
> + &table->mc_promisc_last);
> + debugfs_create_bool("mc_overflow", 0444, table->debug_dir,
> + &table->mc_overflow);
> + debugfs_create_bool("mc_chaining", 0444, table->debug_dir,
> + &table->mc_chaining);
> +#endif
It would be good to continue the practice of using the debugfs_*
primitives in your debugfs.c and just make a single call here that
doesn't need the ifdef
> +
> efx->filter_state = table;
>
> return 0;
> @@ -1518,6 +1532,10 @@ void efx_mcdi_filter_table_remove(struct efx_nic *efx)
> return;
>
> vfree(table->entry);
> +#ifdef CONFIG_DEBUG_FS
> + /* Remove debugfs entries pointing into @table */
> + debugfs_remove_recursive(table->debug_dir);
> +#endif
> kfree(table);
> }
>
> diff --git a/drivers/net/ethernet/sfc/mcdi_filters.h b/drivers/net/ethernet/sfc/mcdi_filters.h
> index c0d6558b9fd2..897843ade3ec 100644
> --- a/drivers/net/ethernet/sfc/mcdi_filters.h
> +++ b/drivers/net/ethernet/sfc/mcdi_filters.h
> @@ -91,6 +91,10 @@ struct efx_mcdi_filter_table {
> bool vlan_filter;
> /* Entries on the vlan_list are added/removed under filter_sem */
> struct list_head vlan_list;
> +#ifdef CONFIG_DEBUG_FS
> + /* filter table debugfs directory */
> + struct dentry *debug_dir;
> +#endif
> };
>
> int efx_mcdi_filter_table_probe(struct efx_nic *efx, bool multicast_chaining);
>
Powered by blists - more mailing lists