[v2,0/4] CPU isolation improvements

Message

Leonardo Bras Oct. 13, 2022, 6:40 p.m. UTC

Patch 1 removes some noise from isolation.c

Patch 2 adds some information about the housekeeping flags and a short
description on what to expect from the HK functions. I would really like 
some feedback on this one, since I got all that from the flags usage, and 
maybe I am misreading stuff.

In patch 3, I am suggesting making isolcpus have both the _DOMAIN flag and 
the _WQ flag, so the _DOMAIN flag is not responsible for isolating cpus on 
workqueue operations anymore. This will avoid AND'ing both those bitmaps 
every time we need to check for Workqueue isolation, simplifying code and 
avoiding cpumask allocation in most cases. 

Maybe I am missing something in this move, so please provide feedback.

In patch 4 I use the results from patch 3 and I disallow pcrypt to schedule 
work in cpus that are not enabled for workqueue housekeeping, meaning there 
will be less work done in those isolated cpus.

Best regards,
Leo

Leonardo Bras (4):
  sched/isolation: Fix style issues reported by checkpatch
  sched/isolation: Improve documentation
  sched/isolation: Add HK_TYPE_WQ to isolcpus=domain
  crypto/pcrypt: Do not use isolated CPUs for callback

 crypto/pcrypt.c                 |  9 +++++---
 drivers/pci/pci-driver.c        | 13 +----------
 include/linux/sched/isolation.h | 38 ++++++++++++++++++++-------------
 kernel/sched/isolation.c        |  4 ++--
 kernel/workqueue.c              |  1 -
 net/core/net-sysfs.c            |  1 -
 6 files changed, 32 insertions(+), 34 deletions(-)