linux-kernel - [PATCHSET sched_ext/for-6.12-fixes] sched_ext: Fix locking enable/disable path bugs includling locking order one

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: <20240923190020.1446325-1-tj@kernel.org>
Date: Mon, 23 Sep 2024 08:59:27 -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,
	aboorvad@...ux.ibm.com
Subject: [PATCHSET sched_ext/for-6.12-fixes] sched_ext: Fix locking enable/disable path bugs includling locking order one

Aboorva Devarajan reported an issue where sched_ext init code can
occasionally deadlock when scheduler loading races CPU hotplug. The deadlock
scenario is as follows:

       scx_ops_enable()                               hotplug

                                          percpu_down_write(&cpu_hotplug_lock)
   percpu_down_write(&scx_fork_rwsem)
   block on cpu_hotplug_lock
                                          kthread_create() waits for kthreadd
					  kthreadd blocks on scx_fork_rwsem

Note that this doesn't trigger lockdep because the hotplug side dependency
bounces through kthreadd.

This is primarily caused by SCX enable/disable paths grabbing big locks
together. This patchset updates the enable/disable paths to decouple the
locks. In the process, it also fixes several subtle bugs in the enable path.

This patchset contains the following patches:

 0001-sched_ext-Relocate-check_hotplug_seq-call-in-scx_ops.patch
 0002-sched_ext-Remove-SCX_OPS_PREPPING.patch
 0003-sched_ext-Initialize-in-bypass-mode.patch
 0004-sched_ext-Fix-SCX_TASK_INIT-SCX_TASK_READY-transitio.patch
 0005-sched_ext-Enable-scx_ops_init_task-separately.patch
 0006-sched_ext-Add-scx_cgroup_enabled-to-gate-cgroup-oper.patch
 0007-sched_ext-Decouple-locks-in-scx_ops_disable_workfn.patch
 0008-sched_ext-Decouple-locks-in-scx_ops_enable.patch

 0001-0002 are prep patches.

 0003 removes a race window in the enable path that can cause stalls and
 prepares for further locking updates.

 0004-0005 remove race windows in the enable path that can cause invalid task
 state transitions.

 0006 fixes a bug in cgroup enable path which can skip invocation of
 ops.cgroup_exit() and prepares for further locking updates.

 0007-0008 decouple the big locks and fix the deadlock.

This patchset can also be found in the following git branch:

 git://git.kernel.org/pub/scm/linux/kernel/git/tj/sched_ext.git scx-enable-locking-fix

diffstat follows. Thanks.

 kernel/sched/ext.c |  199 ++++++++++++++++++++++++++++++++------------------------------------
 1 file changed, 94 insertions(+), 105 deletions(-)

--
tejun