Re: [PATCH -mm -v6 3/9] mm, THP, swap: Add swap cluster allocate/free functions

["Huang\, Ying" <ying.huang@intel.com>]

Tim Chen <tim.c.chen@linux.intel.com> writes:

> On Wed, 2017-03-15 at 09:19 +0800, Huang, Ying wrote:
>> Tim Chen <tim.c.chen@linux.intel.com> writes:
>> 
>> > 
>> > On Wed, 2017-03-08 at 15:26 +0800, Huang, Ying wrote:
>> > > 
>> > > From: Huang Ying <ying.huang@intel.com>
>> > > 
>> > > The swap cluster allocation/free functions are added based on the
>> > > existing swap cluster management mechanism for SSD.  These functions
>> > > don't work for the rotating hard disks because the existing swap cluster
>> > > management mechanism doesn't work for them.  The hard disks support may
>> > > be added if someone really need it.  But that needn't be included in
>> > > this patchset.
>> > > 
>> > > This will be used for the THP (Transparent Huge Page) swap support.
>> > > Where one swap cluster will hold the contents of each THP swapped out.
>> > > 
>> > > Cc: Andrea Arcangeli <aarcange@redhat.com>
>> > > Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
>> > > Cc: Hugh Dickins <hughd@google.com>
>> > > Cc: Shaohua Li <shli@kernel.org>
>> > > Cc: Minchan Kim <minchan@kernel.org>
>> > > Cc: Rik van Riel <riel@redhat.com>
>> > > Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
>> > > ---
>> > >  mm/swapfile.c | 217 +++++++++++++++++++++++++++++++++++++++++-----------------
>> > >  1 file changed, 156 insertions(+), 61 deletions(-)
>> > > 
>> > > diff --git a/mm/swapfile.c b/mm/swapfile.c
>> > > index a744604384ff..91876c33114b 100644
>> > > --- a/mm/swapfile.c
>> > > +++ b/mm/swapfile.c
>> > > @@ -378,6 +378,14 @@ static void swap_cluster_schedule_discard(struct swap_info_struct *si,
>> > >  	schedule_work(&si->discard_work);
>> > >  }
>> > >  
>> > > +static void __free_cluster(struct swap_info_struct *si, unsigned long idx)
>> > > +{
>> > > +	struct swap_cluster_info *ci = si->cluster_info;
>> > > +
>> > > +	cluster_set_flag(ci + idx, CLUSTER_FLAG_FREE);
>> > > +	cluster_list_add_tail(&si->free_clusters, ci, idx);
>> > > +}
>> > > +
>> > >  /*
>> > >   * Doing discard actually. After a cluster discard is finished, the cluster
>> > >   * will be added to free cluster list. caller should hold si->lock.
>> > > @@ -398,10 +406,7 @@ static void swap_do_scheduled_discard(struct swap_info_struct *si)
>> > >  
>> > >  		spin_lock(&si->lock);
>> > >  		ci = lock_cluster(si, idx * SWAPFILE_CLUSTER);
>> > > -		cluster_set_flag(ci, CLUSTER_FLAG_FREE);
>> > > -		unlock_cluster(ci);
>> > > -		cluster_list_add_tail(&si->free_clusters, info, idx);
>> > > -		ci = lock_cluster(si, idx * SWAPFILE_CLUSTER);
>> > > +		__free_cluster(si, idx);
>> > >  		memset(si->swap_map + idx * SWAPFILE_CLUSTER,
>> > >  				0, SWAPFILE_CLUSTER);
>> > >  		unlock_cluster(ci);
>> > The __free_cluster definition and the above change to eliminate
>> > the extra unlock_cluster and lock_cluster can perhaps be broken up
>> > as a separate patch.  It can be independent of THP changes.
>> I think the change may have no value by itself without THP changes.
>> There will be only 1 user of __free_cluster() and the lock change is
>> trivial too.  So I think it may be better just to keep it as that?
>> 
>
> Seems like the extra unlock and lock of cluster in existing code should be taken out
> irrespective of the THP changes:
>  
> 		cluster_set_flag(ci, CLUSTER_FLAG_FREE);
> -		unlock_cluster(ci);
> 		cluster_list_add_tail(&si->free_clusters, info, idx);
> -		ci = lock_cluster(si, idx * SWAPFILE_CLUSTER);
> 		memset(si->swap_map + idx * SWAPFILE_CLUSTER,
>   				0, SWAPFILE_CLUSTER);
>

This is not a functionality fix or performance optimization.  Because
the lock on the swap_info_struct is held during the operation and there
are no operations on cluster with index "idx" in
cluster_list_add_tail().  The change here is just to make the resulting
code a little simpler.  Is this deserved a separate patch?

Best Regards,
Huang, Ying