Re: [PATCH] srcu: Fix flush sup work warning in cleanup_srcu_struct()

["Paul E. McKenney" <paulmck@kernel.org>]

On Wed, Mar 22, 2023 at 01:15:02AM +0000, Zhang, Qiang1 wrote:
> > insmod rcutorture.ko
> > rmmod rcutorture.ko
> > 
> > [  209.437327] WARNING: CPU: 0 PID: 508 at kernel/workqueue.c:3167 __flush_work+0x50a/0x540
> > [  209.437346] Modules linked in: rcutorture(-) torture [last unloaded: rcutorture]
> > [  209.437382] CPU: 0 PID: 508 Comm: rmmod Tainted: G  W  6.3.0-rc1-yocto-standard+
> > [  209.437406] RIP: 0010:__flush_work+0x50a/0x540
> > .....
> > [  209.437758]  flush_delayed_work+0x36/0x90
> > [  209.437776]  cleanup_srcu_struct+0x68/0x2e0
> > [  209.437817]  srcu_module_notify+0x71/0x140
> > [  209.437854]  blocking_notifier_call_chain+0x9d/0xd0
> > [  209.437880]  __x64_sys_delete_module+0x223/0x2e0
> > [  209.438046]  do_syscall_64+0x43/0x90
> > [  209.438062]  entry_SYSCALL_64_after_hwframe+0x72/0xdc
> > 
> > For srcu objects defined with DEFINE_SRCU() or DEFINE_STATIC_SRCU(),
> > when compiling and loading as modules, the srcu_module_coming() is invoked,
> > allocate memory for srcu structure's->sda and initialize sda structure,
> > due to not fully initialize srcu structure's->sup, so at this time the
> > sup structure's->delaywork.func is null, if not invoke init_srcu_struct_fields()
> > before unloading modules, in srcu_module_going() the __flush_work() find
> > work->func is empty, will raise the warning above.
> > 
> > This commit add init_srcu_struct_fields() to initialize srcu structure's->sup,
> > in srcu_module_coming().
> > 
> > Signed-off-by: Zqiang <qiang1.zhang@intel.com>
> >
> >Good catch, and thank you for testing the in-module case!
> >
> >One question below...
> >
> >							Thanx, Paul
> >
> > ---
> >  kernel/rcu/srcutree.c | 11 ++++++++---
> >  1 file changed, 8 insertions(+), 3 deletions(-)
> > 
> > diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c
> > index 1fb078abbdc9..42d8720e016c 100644
> > --- a/kernel/rcu/srcutree.c
> > +++ b/kernel/rcu/srcutree.c
> > @@ -1921,7 +1921,8 @@ static int srcu_module_coming(struct module *mod)
> >  		ssp->sda = alloc_percpu(struct srcu_data);
> >  		if (WARN_ON_ONCE(!ssp->sda))
> >  			return -ENOMEM;
> > -		init_srcu_struct_data(ssp);
> > +		if (WARN_ON_ONCE(init_srcu_struct_fields(ssp, true)))
> > +			return -ENOMEM;
> >
> >Wouldn't it be better to simply delete the init_srcu_struct_data()?
> >
> >Then the first call to check_init_srcu_struct() would take care of
> >the initialization, just as for the non-module case.  Or am I missing
> >something subtle?
> 
> Hi Paul
> 
> Maybe the check_init_srcu_struct() is always not invoked, for example,
> In rcutorture.c,   here is such a definition DEFINE_STATIC_SRCU(srcu_ctl),
> but we use torture_type=rcu to test,  there will not be any interface calling
> check_init_srcu_struct() to initialize srcu_ctl and set  structure's->delaywork.func 
> is process_srcu().
> when we unload the rcutorture module, invoke cleanup_srcu_struct()
> to flush sup structure's->delaywork.func, due to the func pointer is not initialize,
> it's null, will trigger warning.
> 
> About kernel/workqueue.c:3167
> 
> __flush_work
>      if (WARN_ON(!work->func))   <---------trigger waning
> 	return false;
> 
> 
> and  in  init_srcu_struct_fields(ssp, true), wil set srcu_sup->sda_is_static is true
> and set srcu_sup-> srcu_gp_seq_needed is zero,  after that when we call
>  check_init_srcu_struct() again, it not be initialized again.

Good point!  In the non-module statically allocated case there is never
a call to cleanup_srcu_struct().

So suppose the code in srcu_module_coming() only did the alloc_percpu(),
and then the code in srcu_module_going() only did the the matching
free_percpu() instead of the current cleanup_srcu_struct()?

							Thanx, Paul

> Thanks
> Zqiang
> 
> >
> >It should also be possible to eliminate duplicate code between the
> >in-module and non-module statically allocated initialization cases,
> >come to think of it.
> >
> >  	}
> >  	return 0;
> >  }
> > @@ -1931,9 +1932,13 @@ static void srcu_module_going(struct module *mod)
> >  {
> >  	int i;
> >  	struct srcu_struct **sspp = mod->srcu_struct_ptrs;
> > +	struct srcu_struct *ssp;
> >  
> > -	for (i = 0; i < mod->num_srcu_structs; i++)
> > -		cleanup_srcu_struct(*(sspp++));
> > +	for (i = 0; i < mod->num_srcu_structs; i++) {
> > +		ssp = *(sspp++);
> > +		cleanup_srcu_struct(ssp);
> > +		free_percpu(ssp->sda);
> > +	}
> >
> >And good catch on another memory leak with this one, looks proper
> >to me.
> >
> >  }
> >  
> >  /* Handle one module, either coming or going. */
> > -- 
> > 2.25.1
> >