Re: [PATCH] dlm: wait for outstanding SRCU callbacks to complete in exit paths

[Alexander Aring]

Hi,

On Wed, Sep 2, 2026 at 10:49 AM Zqiang <qiang.zhang@xxxxxxxxx> wrote:
>
> >
> > Hi,
> >
> > On Wed, Sep 2, 2026 at 10:30 AM Zqiang <qiang.zhang@xxxxxxxxx> wrote:
> >
> > >
> > > >
> > >  > Hi,
> > >  >
> > >  > On Tue, Sep 1, 2026 at 5:19 AM Zqiang <qiang.zhang@xxxxxxxxx> wrote:
> > >  >
> > >  >
> > >  > The dlm_lowcomms_exit() and dlm_midcomms_exit() iterate over the
> > >  > srcu protected connection and node hash tables and hand each
> > >  > element to call_srcu() for deferred freeing (connection_release()
> > >  > and midcomms_node_release()). call_srcu() is asynchronous: the
> > >  > callbacks are invoked only after an SRCU grace period, which may
> > >  > happen after the exit function has already returned.
> > >  >
> > >  > These exit functions are reached from exit_dlm() on module unload.
> > >  > Once they return, module teardown continues and the module text
> > >  > may be unloaded while call_srcu() callbacks are still pending. When
> > >  > such a callback finally runs, it executes freed module code and
> > >  > touches the static SRCU domains that are being torn down, resulting
> > >  > in a use-after-free.
> > >
> > >  I should rephrase it:
> > >
> > >  Due to missed call to the srcu_barrier(), when invoke cleanup_srcu_struct()
> > >
> > Then there need to be a MUST call of srcu_barrier() before
> > cleanup_srcu_struct() every time.
>
> If there are some call_srcu() call, the srcu_barrier() should be called before cleanup_srcu_struct().
> usually, when the cleanup_srcu_struct() run, we should ensure that there are no pending SRCU callbacks
> or ongoing SRCU grace periods at this time.
>

For me, this is a MUST because who is maintaining something to know if
it was ever called?

> there are another reason:
>
> The midcomms_node_release() srcu callback can call call_rcu(), we must wait the all
> call_rcu() to complete, because rcu_barrier() which in dlm_memory_exit() is required
> to successfully intercept these RCU callbacks.
>

Yes, I can see why "nodes_srcu" is necessary as this has something to
do with kmem_cache is being involed.

The "connections_srcu" is necessary for the above reason that the srcu
API requires you to call srcu_barrier() before cleanup_srcu_struct()
because for whatever reason we don't do that in cleanup_srcu_struct()
as I understand it now.

- Alex