Re: Fwd: a8d5875ce5 ("Default enable RCU list lockdep debugging with .."): WARNING: suspicious RCU usage

[Mimi Zohar <zohar@linux.ibm.com>]

On Thu, 2020-04-30 at 18:37 +0530, Madhuparna Bhowmik wrote:
> On Wed, Apr 29, 2020 at 03:40:58PM -0700, Paul E. McKenney wrote:
> > You do of course need the code to use the RCU variants of list_add*().
> > And also list_for_each_entry_rcu(), as in the current code.
> > 
> > There are several options, none of them perfect:
> > 
> > 1.	Add (not otherwise needed) calls to rcu_read_lock() and
> > 	rcu_read_unlock() and leave list_for_each_entry_rcu() unchanged.
> > 
> > 2.	Add "true" for the optional fourth argument to
> > 	list_for_each_entry_rcu().  This will suppress the complaints,
> > 	but would (incorrectly) continue to do so should this code change
> > 	so as to be able to delete form this list.
> > 
> > 3.	Switch from list_for_each_entry_rcu() to its lockless counterpart,
> > 	list_for_each_entry_lockless().  This is simiar to #2 above, but
> > 	at least the name lets people know that something unusual is up.
> > 
> > If it was my code, I would take door #3.  ;-)
> >
> Thanks a lot for your inputs on this. I will send a patch soon.

Please remember to expand the "mutex" comment in evm_write_xattrs() to
reflect the reason why using list_for_each_entry_lockless() is safe.

thanks,

Mimi