Re: [PATCH 4/4] selftests/resctrl: Adjust SNC support messages

From: Reinette Chatre <reinette.chatre@intel.com>
To: "Luck, Tony" <tony.luck@intel.com>,
	"Wieczor-Retman, Maciej" <maciej.wieczor-retman@intel.com>
Cc: "Yu, Fenghua" <fenghua.yu@intel.com>,
	Shuah Khan <shuah@kernel.org>,
	"james.morse@arm.com" <james.morse@arm.com>,
	"ilpo.jarvinen@linux.intel.com" <ilpo.jarvinen@linux.intel.com>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"linux-kselftest@vger.kernel.org"
	<linux-kselftest@vger.kernel.org>
Subject: Re: [PATCH 4/4] selftests/resctrl: Adjust SNC support messages
Date: Thu, 7 Mar 2024 14:39:08 -0800	[thread overview]
Message-ID: <55a55960-8bb1-4ce2-a2c7-68e167da8bcc@intel.com> (raw)
In-Reply-To: <SJ1PR11MB6083AACB10645E41DD3F9639FC202@SJ1PR11MB6083.namprd11.prod.outlook.com>

Hi Tony,

On 3/7/2024 1:14 PM, Luck, Tony wrote:
>> Thinking about it even differently. The goal is to give information
>> to userspace so we need to think about what would help user space?
>> For example, what if there is a file in info that shows 
>> which CPUs are associated with each domain?
> 
> Reinette,
> 
> Interesting idea. That would save users from having to chase through
> /sys/devices/system/cpu/cpu*/cache/index?/* to figure out what the domain
> numbers in schemata files and the mon_data/mon_L3_XX values mean.
> 
> May be extra useful for ARM which seems to have big random-looking numbers
> for domains that came out of ACPI tables.
> 
> For SNC it would get the user directly to what they probably care about
> (which CPUs are in which domain).

I agree.

> 
> So something like this for an SNC 2 system:
> 
> $ cat /sys/fs/resctrl/info/L3/cpus
> 0: 0-35,72-107
> 1: 36-71,108-143
> 
> $ cat /sys/fs/resctrl/info/L3_MON/cpus
> 0: 0-17,72-89
> 1: 18-35,90-107
> 2: 36-53,108-125
> 3: 54-71,126-143
> 
> [maybe there is a better name than "cpus" for this file?]

Thank you for the example. I find that significantly easier to
understand than a single number in a generic "nodes_per_l3_cache".
Especially with potential confusion surrounding inconsistent "nodes"
between allocation and monitoring. 

How about domain_cpu_list and domain_cpu_map ?

Reinette