Re: possible deadlock in lru_add_drain_all

[Michal Hocko]

On Mon 30-10-17 16:10:09, Peter Zijlstra wrote:
> On Mon, Oct 30, 2017 at 07:09:21PM +0900, Byungchul Park wrote:
> > On Mon, Oct 30, 2017 at 09:22:03AM +0100, Michal Hocko wrote:
> > > [Cc Byungchul. The original full report is
> > > http://lkml.kernel.org/r/089e0825eec8955c1f055c83d476@xxxxxxxxxx]
> > > 
> > > Could you have a look please? This smells like a false positive to me.
> > 
> > +cc peterz@xxxxxxxxxxxxx
> > 
> > Hello,
> > 
> > IMHO, the false positive was caused by the lockdep_map of 'cpuhp_state'
> > which couldn't distinguish between cpu-up and cpu-down.
> > 
> > And it was solved with the following commit by Peter and Thomas:
> > 
> > 5f4b55e10645b7371322c800a5ec745cab487a6c
> > smp/hotplug: Differentiate the AP-work lockdep class between up and down
> > 
> > Therefore, we can avoid the false positive on later than the commit.
> > 
> > Peter and Thomas, could you confirm it?
> 
> I can indeed confirm it's running old code; cpuhp_state is no more.

Does this mean the below chain is no longer possible with the current
linux-next (tip)?

> However, that splat translates like:
> 
> 	__cpuhp_setup_state()
> #0	  cpus_read_lock()
> 	  __cpuhp_setup_state_cpuslocked()
> #1	    mutex_lock(&cpuhp_state_mutex)
> 
> 
> 
> 	__cpuhp_state_add_instance()
> #2	  mutex_lock(&cpuhp_state_mutex)

this should be #1 right?

> 	  cpuhp_issue_call()
> 	    cpuhp_invoke_ap_callback()
> #3	      wait_for_completion()
> 
> 						msr_device_create()
> 						  ...
> #4						    filename_create()
> #3						complete()
> 
> 
> 
> 	do_splice()
> #4	  file_start_write()
> 	  do_splice_from()
> 	    iter_file_splice_write()
> #5	      pipe_lock()
> 	      vfs_iter_write()
> 	        ...
> #6		  inode_lock()
> 
> 
> 
> 	sys_fcntl()
> 	  do_fcntl()
> 	    shmem_fcntl()
> #5	      inode_lock()
> 	      shmem_wait_for_pins()
> 	        if (!scan)
> 		  lru_add_drain_all()
> #0		    cpus_read_lock()
> 
> 
> 
> Which is an actual real deadlock, there is no mixing of up and down.

thanks a lot, this made it more clear to me. It took a while to
actually see 0 -> 1 -> 3 -> 4 -> 5 -> 0 cycle. I have only focused
on lru_add_drain_all while it was holding the cpus lock.
-- 
Michal Hocko
SUSE Labs