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-next>] [day] [month] [year] [list]
Message-Id: <20220804143609.515789-1-qais.yousef@arm.com>
Date:   Thu,  4 Aug 2022 15:36:00 +0100
From:   Qais Yousef <qais.yousef@....com>
To:     Ingo Molnar <mingo@...nel.org>,
        "Peter Zijlstra (Intel)" <peterz@...radead.org>,
        Vincent Guittot <vincent.guittot@...aro.org>,
        Dietmar Eggemann <dietmar.eggemann@....com>
Cc:     linux-kernel@...r.kernel.org, Xuewen Yan <xuewen.yan94@...il.com>,
        Lukasz Luba <lukasz.luba@....com>, Wei Wang <wvw@...gle.com>,
        Jonathan JMChen <Jonathan.JMChen@...iatek.com>,
        Hank <han.lin@...iatek.com>, Qais Yousef <qais.yousef@....com>
Subject: [PATCH v2 0/9] Fix relationship between uclamp and fits_capacity()

Relationship between uclamp and fits_capacity() is currently broken. Mostly due
to how uclamp should interact with migration margin and capacity pressure. But
also due not all users were converted to consider uclamp before calling
fits_capacity(). Namely cpu_overutilized().

The meat of the series is patch 1 where we introduce a new function,
util_fits_cpu(), that takes uclamp into account. The new function should
replace all call sits to fits_capacity(), which what subsequent patches do.
Except for patch 7 where we fix handling of early exit condition in
find_energy_efficient_cpu(AKA feec()) that must be uclamp aware too.

util_fits_cpu() will revert back to a simple call to fits_capacity() if uclamp
is not being used on the system.

Last two patches still need testing and verification, but they address the
various thermal handling issues raised in v1. We can re-order the patches, but
added at the end to facilitate review at this stage. Hope the approach and the
logic make sense.

I'll be on holidays, so if I don't respond to review comments fast enough,
apologies in advance.

Changes in v2:

	* use uclamp_rq_is_idle() in uclamp_rq_util_with() (Xuewen)

	* Simplify logic in update_sg_wakeup_stats() when converting
	  task_fits_cpu() (Vincent)

	* Better handling of thermal pressure (Vincent)

	  - We consider thermal pressure for uclamp_min in patch 1
	  - 2 new patches to handle capacity inversion which improves handling
	    when:

		+ There're multiple big cpus on separate perf domains.
		+ A task is boosted by uclamp_min and inspite of thermal
		  pressure, the big cpu is still the best placement.

		  e.g: p0->util_avg = 300, p0->uclamp_min = 1024

		  This task should stay on big CPU until thermal pressure is in
		  capacity inversion.

	v1 discussion: https://lore.kernel.org/lkml/20220629194632.1117723-1-qais.yousef@arm.com/

	Thanks for all reviewers on v1!


Qais Yousef (9):
  sched/uclamp: Fix relationship between uclamp and migration margin
  sched/uclamp: Make task_fits_capacity() use util_fits_cpu()
  sched/uclamp: Fix fits_capacity() check in feec()
  sched/uclamp: Make select_idle_capacity() use util_fits_cpu()
  sched/uclamp: Make asym_fits_capacity() use util_fits_cpu()
  sched/uclamp: Make cpu_overutilized() use util_fits_cpu()
  sched/uclamp: Cater for uclamp in find_energy_efficient_cpu()'s early
    exit condition
  sched/fair: Detect capacity inversion
  sched/fair: Consider capacity inversion in util_fits_cpu()

 kernel/sched/core.c  |  10 +-
 kernel/sched/fair.c  | 290 ++++++++++++++++++++++++++++++++++++++-----
 kernel/sched/sched.h |  70 ++++++++++-
 3 files changed, 329 insertions(+), 41 deletions(-)

-- 
2.25.1

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ