lkml.org 
[lkml]   [2020]   [Mar]   [11]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
Subject[RFC PATCH 0/3] sched: Instrument sched domain flags
Date
Hi,

I've repeatedly stared at an SD flag and asked myself "how should that be
set up in the domain hierarchy anyway?". I figured that if we formalize our
flags zoology a bit, we could also do some runtime assertions on them -
this is what this series is all about.

Note that this is based on top of my select_task_rq() series [1], since it
removes SD_LOAD_BALANCE. If that other series dies I can go and rebase this
again on a branch that still has the flag.

Patches
=======

The idea is to associate the flags with a metatype that describes how they
should be set in a sched domain hierarchy - details are in the comments and
commit logs. For now it's just a simple parent/children relationship
description ("if this SD has it, all its {parents, children} have it").

The good thing is that this all goes away when CONFIG_SCHED_DEBUG isn't
set. The bad thing is that replaces SD_* flags definitions with some
unsavoury macros. This is mainly because I wanted to avoid having to
duplicate work between declaring the flags and declaring their types.

Discussion points
=================

I've gone with a flags field so that several behaviours can be associated
with a given SD flag, but right now they only get assigned one. An enum
could fit that job, although it's more constraining.

Naming is also a pain. I'm not really hot on "shared", but that's as
explicit as I managed to be.

I've inserted the reasoning behind the metaflag assignment in
comments. They might be a bit too wordy, so we may want to make them a bit
more broad to lessen the maintenance burden.

Lastly, since this adds an infrastructure to store flag names, we could use
that to pretty-print /proc/sys/kernel/sched_domain/cpu*/domain*/flags.

Some deltas
===========

I get a small codesize increase with SCHED_DEBUG=n due to the first
patch:

$ compare.sh before after vmlinux.o
SYMBOL BEFORE AFTER DELTA
build_sched_domains 4552 4588 +36

For instance, while my baseline would have this (this is all in sd_init()):

0078 a90c8: d63f0000 blr x0
007c a90cc: 1284b801 mov w1, #0xffffda3f // ~TOPOLOGY_SD_FLAGS
0080 a90d0: 6a010001 ands w1, w0, w1
0084 a90d4: 54000c81 b.ne a9264 <sd_init+0x214>

The change would have this:

0078 a90c8: d63f0000 blr x0
007c a90cc: 2a0003e1 mov w1, w0
0080 a90d0: 1284b800 mov w0, #0xffffda3f // ~TOPOLOGY_SD_FLAGS
0084 a90d4: 6a00003f tst w1, w0
0088 a90d8: 54000c81 b.ne a9268 <sd_init+0x218>

Sadly, the exact reasons why elude me.

[1]: https://lore.kernel.org/lkml/20200311181601.18314-1-valentin.schneider@arm.com/

Valentin Schneider (3):
sched/topology: Split out SD_* flags declaration to its own file
sched/topology: Define and assign sched_domain flag metadata
sched/topology: Verify SD_* flags setup when sched_debug is on

include/linux/sched/sd_flags.h | 139 +++++++++++++++++++++++++++++++++
include/linux/sched/topology.h | 29 +++----
kernel/sched/topology.c | 16 ++++
3 files changed, 170 insertions(+), 14 deletions(-)
create mode 100644 include/linux/sched/sd_flags.h

--
2.24.0

\
 
 \ /
  Last update: 2020-03-11 19:34    [W:0.082 / U:1.012 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site