lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <5c4604cd-1cfc-6116-12e6-95054e77736a@mellanox.com>
Date:   Wed, 26 Feb 2020 11:53:04 +0200
From:   Paul Blakey <paulb@...lanox.com>
To:     Saeed Mahameed <saeedm@...lanox.com>,
        Roi Dayan <roid@...lanox.com>,
        "saeedm@....mellanox.co.il" <saeedm@....mellanox.co.il>,
        "xiangxia.m.yue@...il.com" <xiangxia.m.yue@...il.com>,
        "gerlitz.or@...il.com" <gerlitz.or@...il.com>
Cc:     "netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: Re: [net-next] net/mlx5e: Remove the unnecessary parameter


On 2/26/2020 12:50 AM, Saeed Mahameed wrote:
> On Wed, 2020-02-26 at 00:03 +0800, xiangxia.m.yue@...il.com wrote:
>> From: Tonghao Zhang <xiangxia.m.yue@...il.com>
>>
>> The parameter desired_size is always 0, and there is only one
>> function calling the mlx5_esw_chains_get_avail_sz_from_pool.
>> Deleting the parameter desired_size.
> Paul, what is the reasoning behind desired size, i confirm that it  is
> not actually used right now, do we have a pending patch that needs it
> ? 
> if this is not going to happen in the near future i vote to apply this
> patch and bring it back when needed.

Right, it will be used in a following patch that reduces the size given for nft flow tables.

I planned on submitting it after connection tracking offload is complete, but it can be sent now.

This is the patch:

>From 66d3cb9706ed09f00150a42f555a51404602bba4 Mon Sep 17 00:00:00 2001
From: Paul Blakey <paulb@...lanox.com>
Date: Wed, 8 Jan 2020 14:31:53 +0200
Subject: [PATCH] net/mlx5: Allocate smaller size tables for ft offload

Instead of giving ft tables one of the largest tables available - 4M,
give it a more reasonable size - 64k. Especially since it will
always be created as a miss hook in the following patch.

Signed-off-by: Paul Blakey <paulb@...lanox.com>
Reviewed-by: Mark Bloch <markb@...lanox.com>
---
 drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads_chains.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads_chains.c b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads_chains.c
index 3990066..dabbc05 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads_chains.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads_chains.c
@@ -39,6 +39,7 @@
                                          1 * 1024 * 1024,
                                          64 * 1024,
                                          128 };
+#define ESW_FT_TBL_SZ (64 * 1024)

 struct mlx5_esw_chains_priv {
        struct rhashtable chains_ht;
@@ -205,7 +206,9 @@ static unsigned int mlx5_esw_chains_get_level_range(struct mlx5_eswitch *esw)
                ft_attr.flags |= (MLX5_FLOW_TABLE_TUNNEL_EN_REFORMAT |
                                  MLX5_FLOW_TABLE_TUNNEL_EN_DECAP);

-       sz = mlx5_esw_chains_get_avail_sz_from_pool(esw, POOL_NEXT_SIZE);
+       sz = (chain == mlx5_esw_chains_get_ft_chain(esw)) ?
+            mlx5_esw_chains_get_avail_sz_from_pool(esw, ESW_FT_TBL_SZ) :
+            mlx5_esw_chains_get_avail_sz_from_pool(esw, POOL_NEXT_SIZE);
        if (!sz)
                return ERR_PTR(-ENOSPC);
        ft_attr.max_fte = sz;
--
1.8.3.1


>
> Thanks,
> Saeed.
>
>> Signed-off-by: Tonghao Zhang <xiangxia.m.yue@...il.com>
>> ---
>>  .../net/ethernet/mellanox/mlx5/core/eswitch_offloads_chains.c | 11
>> +++--------
>>  1 file changed, 3 insertions(+), 8 deletions(-)
>>
>> diff --git
>> a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads_chains.c
>> b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads_chains.c
>> index c5a446e..ce5b7e1 100644
>> ---
>> a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads_chains.c
>> +++
>> b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads_chains.c
>> @@ -134,19 +134,14 @@ static unsigned int
>> mlx5_esw_chains_get_level_range(struct mlx5_eswitch *esw)
>>  	return FDB_TC_LEVELS_PER_PRIO;
>>  }
>>  
>> -#define POOL_NEXT_SIZE 0
>>  static int
>> -mlx5_esw_chains_get_avail_sz_from_pool(struct mlx5_eswitch *esw,
>> -				       int desired_size)
>> +mlx5_esw_chains_get_avail_sz_from_pool(struct mlx5_eswitch *esw)
>>  {
>>  	int i, found_i = -1;
>>  
>>  	for (i = ARRAY_SIZE(ESW_POOLS) - 1; i >= 0; i--) {
>> -		if (fdb_pool_left(esw)[i] && ESW_POOLS[i] >
>> desired_size) {
>> +		if (fdb_pool_left(esw)[i])
>>  			found_i = i;
>> -			if (desired_size != POOL_NEXT_SIZE)
>> -				break;
>> -		}
>>  	}
>>  
>>  	if (found_i != -1) {
>> @@ -198,7 +193,7 @@ static unsigned int
>> mlx5_esw_chains_get_level_range(struct mlx5_eswitch *esw)
>>  		ft_attr.flags |= (MLX5_FLOW_TABLE_TUNNEL_EN_REFORMAT |
>>  				  MLX5_FLOW_TABLE_TUNNEL_EN_DECAP);
>>  
>> -	sz = mlx5_esw_chains_get_avail_sz_from_pool(esw,
>> POOL_NEXT_SIZE);
>> +	sz = mlx5_esw_chains_get_avail_sz_from_pool(esw);
>>  	if (!sz)
>>  		return ERR_PTR(-ENOSPC);
>>  	ft_attr.max_fte = sz;

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ