linux-kernel - Re: [PATCH v8 6/6] cpufreq: powernv: Add sysfs attributes to show throttle stats

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]

Date:	Wed, 3 Feb 2016 14:33:57 +0530
From:	Viresh Kumar <viresh.kumar@...aro.org>
To:	Shilpasri G Bhat <shilpa.bhat@...ux.vnet.ibm.com>
Cc:	linuxppc-dev@...abs.org, linux-kernel@...r.kernel.org,
	rjw@...ysocki.net, linux-pm@...r.kernel.org, pc@...ibm.com,
	anton@...ba.org, ego@...ux.vnet.ibm.com,
	shreyas@...ux.vnet.ibm.com, bsingharora@...il.com,
	mpe@...erman.id.au, linux-api@...r.kernel.org
Subject: Re: [PATCH v8 6/6] cpufreq: powernv: Add sysfs attributes to show
 throttle stats

On 03-02-16, 14:12, Shilpasri G Bhat wrote:
> I need the chip-id in the <attr>_show(). With just sysfs_create_group() I will
> get the cpufreq_global_kobject in the <attr>_show() and I will not be able to
> figure out the chip-id.

The more I look at it, the more I am convinced that keeping this
'chip' directory in /sys/devices/system/cpu/cpuX/cpufreq/ makes sense.

So, here is the deal:
- A 'chip' on your platforms can contain multiple group of CPUs, which
  are represented by policies in cpufreq core. i.e. A chip can have
  multiple policies.
- All CPUs present on the same chip are subject to same throttling
  outcomes.
- Right now you are putting the 'chip' directory in cpu/cpufreq/
  directory. Because that directory isn't specific to a policy, but
  entire cpufreq subsystem, you can't get a policy->cpu in the code
  for the kobject in question. And so you are *forced* to create a
  kobject, so that you can do container_of() and get chip->id.
- And then you also need to unnecessarily add another field in the
  chip directory 'chip_mask', that is nothing but an bitwise OR
  operation on policy->related_cpus, so that userspace can know which
  policies/CPUs are managed by the 'chip'.

What I can suggest is:
- Move this directory inside cpuX/cpufreq/ directory, in a similar way
  as to how we create 'stats' directory today.
- You can then get policy->cpu, to get chip->id out of it.
- The only disadvantage here is that the same chip directory will be
  replicated in multiple policies, but that makes it more readable.

Thoughts ?

-- 
viresh