netdev - Re: [PATCH v2 net-next 1/2] net: make default_rps

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] [day] [month] [year] [list]

Message-ID: <Y/TnEUMQ9BT1rG3u@corigine.com>
Date:   Tue, 21 Feb 2023 16:45:21 +0100
From:   Simon Horman <simon.horman@...igine.com>
To:     Paolo Abeni <pabeni@...hat.com>
Cc:     netdev@...r.kernel.org, "David S. Miller" <davem@...emloft.net>,
        Jakub Kicinski <kuba@...nel.org>,
        Eric Dumazet <edumazet@...gle.com>,
        Shuah Khan <shuah@...nel.org>
Subject: Re: [PATCH v2 net-next 1/2] net: make default_rps_mask a per netns
 attribute

On Fri, Feb 17, 2023 at 01:28:49PM +0100, Paolo Abeni wrote:
> That really was meant to be a per netns attribute from the beginning.
> 
> The idea is that once proper isolation is in place in the main
> namespace, additional demux in the child namespaces will be redundant.
> Let's make child netns default rps mask empty by default.
> 
> To avoid bloating the netns with a possibly large cpumask, allocate
> it on-demand during the first write operation.
> 
> Signed-off-by: Paolo Abeni <pabeni@...hat.com>

...

> diff --git a/net/core/sysctl_net_core.c b/net/core/sysctl_net_core.c
> index 7130e6d9e263..74842b453407 100644
> --- a/net/core/sysctl_net_core.c
> +++ b/net/core/sysctl_net_core.c
> @@ -74,24 +74,47 @@ static void dump_cpumask(void *buffer, size_t *lenp, loff_t *ppos,
>  #endif
>  
>  #ifdef CONFIG_RPS
> -struct cpumask rps_default_mask;
> +
> +static struct cpumask *rps_default_mask_cow_alloc(struct net *net)
> +{
> +	struct cpumask *rps_default_mask;
> +
> +	if (net->core.rps_default_mask)
> +		return net->core.rps_default_mask;
> +
> +	rps_default_mask = kzalloc(cpumask_size(), GFP_KERNEL);
> +	if (!rps_default_mask)
> +		return NULL;
> +
> +	/* pairs with READ_ONCE in rx_queue_default_mask() */
> +	WRITE_ONCE(net->core.rps_default_mask, rps_default_mask);
> +	return rps_default_mask;
> +}
>  
>  static int rps_default_mask_sysctl(struct ctl_table *table, int write,
>  				   void *buffer, size_t *lenp, loff_t *ppos)
>  {
> +	struct net *net = (struct net *)table->data;
>  	int err = 0;
>  
>  	rtnl_lock();
>  	if (write) {
> -		err = cpumask_parse(buffer, &rps_default_mask);
> +		struct cpumask *rps_default_mask = rps_default_mask_cow_alloc(net);
> +
> +		err = -ENOMEM;

nit: Would it be nicer to set err to -ENOMEM inside the if clause?
     I think that is the only path where it is used.

> +		if (!rps_default_mask)
> +			goto done;
> +
> +		err = cpumask_parse(buffer, rps_default_mask);
>  		if (err)
>  			goto done;
>  
> -		err = rps_cpumask_housekeeping(&rps_default_mask);
> +		err = rps_cpumask_housekeeping(rps_default_mask);
>  		if (err)
>  			goto done;
>  	} else {
> -		dump_cpumask(buffer, lenp, ppos, &rps_default_mask);
> +		dump_cpumask(buffer, lenp, ppos,
> +			     net->core.rps_default_mask ? : cpu_none_mask);
>  	}
>  
>  done:

...