linux-kernel - [PATCHSET sched_ext/for-6.12-fixes] sched_ext: Fix RCU and other stalls while iterating tasks during enable/disable

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: <20241009214411.681233-1-tj@kernel.org>
Date: Wed,  9 Oct 2024 11:40:56 -1000
From: Tejun Heo <tj@...nel.org>
To: void@...ifault.com
Cc: kernel-team@...a.com,
	linux-kernel@...r.kernel.org,
	sched-ext@...a.com
Subject: [PATCHSET sched_ext/for-6.12-fixes] sched_ext: Fix RCU and other stalls while iterating tasks during enable/disable

The enable/disable paths walk all tasks a couple times in bypass mode. There
are a couple problems:

- Bypass mode incorrectly depends on ops.select_cpu() which must not be
  trusted in bypass mode.

- scx_tasks_lock is held while walking all tasks. This can lead to RCU and
  other stalls on a large heavily contended system with many tasks.

Fix the former by always using the default select_cpu() in bypass mode and
the latter by periodically dropping scx_tasks_lock while iterating tasks.

This patchset contains the following patches:

 0001-Revert-sched_ext-Use-shorter-slice-while-bypassing.patch
 0002-sched_ext-Start-schedulers-with-consistent-p-scx.sli.patch
 0003-sched_ext-Move-scx_buildin_idle_enabled-check-to-scx.patch
 0004-sched_ext-bypass-mode-shouldn-t-depend-on-ops.select.patch
 0005-sched_ext-Move-scx_tasks_lock-handling-into-scx_task.patch
 0006-sched_ext-Don-t-hold-scx_tasks_lock-for-too-long.patch

 0001 reverts an unnecessary earlier change.

 0002 makes sure that p->scx.slice is consistent on scheduler load.

 0003-0004 make bypass mode always use the default select_cpu().

 0005-0006 make task iteration drop scx_tasks_lock periodically.

and is also available in the following git branch:

 git://git.kernel.org/pub/scm/linux/kernel/git/tj/sched_ext.git scx-fix-task_iter-stalls

diffstat follows. Thanks.

 kernel/sched/ext.c |  183 +++++++++++++++++++++++++++++++++++++-------------------------------
 1 file changed, 100 insertions(+), 83 deletions(-)

--
tejun