netdev - Re: [PATCH 2/2] net/mlx5e: Leverage sched_numa_hop

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: <xhsmhleruc0uk.mognet@vschneid.remote.csb>
Date:   Thu, 11 Aug 2022 15:26:59 +0100
From:   Valentin Schneider <vschneid@...hat.com>
To:     Tariq Toukan <ttoukan.linux@...il.com>, netdev@...r.kernel.org,
        linux-kernel@...r.kernel.org, Jakub Kicinski <kuba@...nel.org>
Cc:     Tariq Toukan <tariqt@...dia.com>,
        "David S. Miller" <davem@...emloft.net>,
        Saeed Mahameed <saeedm@...dia.com>,
        Ingo Molnar <mingo@...hat.com>,
        Peter Zijlstra <peterz@...radead.org>,
        Juri Lelli <juri.lelli@...hat.com>,
        Eric Dumazet <edumazet@...gle.com>,
        Paolo Abeni <pabeni@...hat.com>, Gal Pressman <gal@...dia.com>,
        Vincent Guittot <vincent.guittot@...aro.org>
Subject: Re: [PATCH 2/2] net/mlx5e: Leverage sched_numa_hop_mask()

On 10/08/22 15:57, Tariq Toukan wrote:
> On 8/10/2022 1:51 PM, Valentin Schneider wrote:
>> Signed-off-by: Valentin Schneider <vschneid@...hat.com>
>> ---
>>   drivers/net/ethernet/mellanox/mlx5/core/eq.c | 16 ++++++++++++++--
>>   1 file changed, 14 insertions(+), 2 deletions(-)
>>
>
> Missing description.
>
> I had a very detailed description with performance numbers and an
> affinity hints example with before/after tables. I don't want to get
> them lost.
>

Me neither! This here is just a stand-in to show how the interface would be
used, I'd much rather have someone who actually knows the code and can
easily test it do it :)

>
>> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eq.c b/drivers/net/ethernet/mellanox/mlx5/core/eq.c
>> index 229728c80233..2eb4ffd96a95 100644
>> --- a/drivers/net/ethernet/mellanox/mlx5/core/eq.c
>> +++ b/drivers/net/ethernet/mellanox/mlx5/core/eq.c
>> @@ -809,9 +809,12 @@ static void comp_irqs_release(struct mlx5_core_dev *dev)
>>   static int comp_irqs_request(struct mlx5_core_dev *dev)
>>   {
>>      struct mlx5_eq_table *table = dev->priv.eq_table;
>> +	const struct cpumask *mask;
>>      int ncomp_eqs = table->num_comp_eqs;
>> +	int hops = 0;
>>      u16 *cpus;
>>      int ret;
>> +	int cpu;
>>      int i;
>>
>>      ncomp_eqs = table->num_comp_eqs;
>> @@ -830,8 +833,17 @@ static int comp_irqs_request(struct mlx5_core_dev *dev)
>>              ret = -ENOMEM;
>>              goto free_irqs;
>>      }
>> -	for (i = 0; i < ncomp_eqs; i++)
>> -		cpus[i] = cpumask_local_spread(i, dev->priv.numa_node);
>> +
>> +	rcu_read_lock();
>> +	for_each_numa_hop_mask(dev->priv.numa_node, hops, mask) {
>
> We don't really use this 'hops' iterator. We always pass 0 (not a
> valuable input...), and we do not care about its final value. Probably
> it's best to hide it from the user into the macro.
>

That's a very valid point. After a lot of mulling around, I've found some
way to hide it away in a macro, but it's not pretty :-) cf. other email.