Re: swap: nfs: Sleeping function called from an rcu read section in nfs_swap_activate

[Mel Gorman <mgorman@suse.de>]

On Thu, May 28, 2015 at 07:24:34AM -0400, Jeff Layton wrote:
> > > 1) this is not done under a lock, so the non-atomic ++/-- is racy if
> > > there are multiple swapons/swapoffs running concurrently on the same
> > > xprt. Shouldn't those use an atomic?
> > > 
> > 
> > It would be more appropriate to use atomics. It's a long time ago but I
> > doubt I considered the possibility of multiple swapons racing at the
> > time of implementation. Activation is typically a serialised task run
> > from init.
> > 
> > > 2) on enable, "swapper" is incremented and memalloc is set on the
> > > socket. Do we need to do xs_set_memalloc every time swapon is called,
> > > or only on a 0->1 swapper transition.
> > > 
> > 
> > Every time because the static_key_slow_inc call is for the total number
> > of connections.
> > 
> 
> That still seems wrong. The static_key would still be active even if
> you just did it once per xprt.
> 

True. As long as it is active while one swapfile exists then it's good.

-- 
Mel Gorman
SUSE Labs

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>