Re: [BUG v6.0-rc2] lockdep splat on ct_kernel_enter()

[Cheng-Jui Wang (王正睿)]

On Wed, 2023-03-01 at 13:52 +0100, gregkh@xxxxxxxxxxxxxxxxxxx wrote:
> On Wed, Mar 01, 2023 at 12:37:29PM +0000, Cheng-Jui Wang (王正睿) wrote:
> > On Mon, 2022-08-22 at 16:44 -0400, Steven Rostedt wrote:
> > > My tests are failing because of this splat:
> > > 
> > > [   16.073659] ------------[ cut here ]------------
> > > [   16.074407] bus: 'platform': add driver acpi-ged
> > > [   16.074424] DEBUG_LOCKS_WARN_ON(lockdep_hardirqs_enabled())
> > > [   16.074424] WARNING: CPU: 0 PID: 0 at
> > > kernel/locking/lockdep.c:5506 check_flags+0x114/0x1d0
> > > [   16.074424]  lock_is_held_type+0x6f/0x130
> > > [   16.186284]  rcu_read_lock_sched_held+0x4a/0x90
> > > [   16.186284]  trace_rcu_dyntick+0x3a/0xe0
> > > [   16.186284]  ct_kernel_enter.constprop.0+0x66/0xa0
> > > [   16.186284]  ct_idle_exit+0xd/0x30
> > > [   16.186284]  cpuidle_enter_state+0x28a/0x310
> > > [   16.186284]  cpuidle_enter+0x2e/0x50
> > > [   16.186284]  do_idle+0x1ec/0x280
> > 
> > Our test in v6.1 stable is failing due to this splat too. The v6.1
> > stable kernel still has this splat.
> > 
> > This splat can be fixed by Peter's patch 
> > https://lore.kernel.org/all/20220608144516.808451191@xxxxxxxxxxxxx/
> > , but the fix is part of a big patchset 
> > https://lore.kernel.org/all/20220608142723.103523089@xxxxxxxxxxxxx/
> > introduced in 6.2.
> > 
> > Could the fixes be backported to v6.1 stable?
> 
> What "fixes" exactly are you referring to?  
a01353cf1896 ("cpuidle: Fix ct_idle_*() usage") and other patches it
needs.

> Can you provide a series of
> git commit ids that cleanly apply or better yet, and series of
> patches
> that you have backported and tested to ensure that they work
> properly?
Suren has backported part of the patchset to fix this particular issue
in android14-6.1: 
https://android-review.googlesource.com/c/kernel/common/+/2449203 to 
https://android-review.googlesource.com/c/kernel/common/+/2449212. I
have tested these patches and confirmed these can fix this splat.

Following are the commit ids of the above series of patches in Linus'
tree:
5e26aa93391195a64871db5d96d7163f0062ca4f
8e9ab9e8da1eae61fdff35690d998eaf8cd527dc
5fca0d9f5d76664786ca6c09076341def165a677
e038f7b8028a1d1bc8ac82351c71ea538f19a879
b3f46658ce40a3467cda82f920dd9d5325ab0eaf
4d1be9e745382b41492b0cb9000829863db7133a
4ce40e9dbe83153f60d7e4ccd24a1eb4f8264f6a
c3d42418dca53d6c498a48c408f7a45289593650
0c5ffc3d7b15978c6b184938cd6b8af06e436424
a01353cf1896ea5b8a7bbc5e2b2d38feed8b7aaa

thanks,

cheng-jui