Re: [PATCH -V4 -mm] mm, swap: Fix race between swapoff and some swap operations

[Mel Gorman <mgorman@techsingularity.net>]

On Wed, Jan 03, 2018 at 08:42:15AM +0800, Huang, Ying wrote:
> Mel Gorman <mgorman@techsingularity.net> writes:
> 
> > On Tue, Jan 02, 2018 at 12:29:55PM +0100, Jan Kara wrote:
> >> On Tue 02-01-18 10:21:03, Mel Gorman wrote:
> >> > On Sat, Dec 23, 2017 at 10:36:53AM +0900, Minchan Kim wrote:
> >> > > > code path.  It appears that similar situation is possible for them too.
> >> > > > 
> >> > > > The file cache pages will be delete from file cache address_space before
> >> > > > address_space (embedded in inode) is freed.  But they will be deleted
> >> > > > from LRU list only when its refcount dropped to zero, please take a look
> >> > > > at put_page() and release_pages().  While address_space will be freed
> >> > > > after putting reference to all file cache pages.  If someone holds a
> >> > > > reference to a file cache page for quite long time, it is possible for a
> >> > > > file cache page to be in LRU list after the inode/address_space is
> >> > > > freed.
> >> > > > 
> >> > > > And I found inode/address_space is freed witch call_rcu().  I don't know
> >> > > > whether this is related to page_mapping().
> >> > > > 
> >> > > > This is just my understanding.
> >> > > 
> >> > > Hmm, it smells like a bug of __isolate_lru_page.
> >> > > 
> >> > > Ccing Mel:
> >> > > 
> >> > > What locks protects address_space destroying when race happens between
> >> > > inode trauncation and __isolate_lru_page?
> >> > > 
> >> > 
> >> > I'm just back online and have a lot of catching up to do so this is a rushed
> >> > answer and I didn't read the background of this. However the question is
> >> > somewhat ambiguous and the scope is broad as I'm not sure which race you
> >> > refer to. For file cache pages, I wouldnt' expect the address_space to be
> >> > destroyed specifically as long as the inode exists which is the structure
> >> > containing the address_space in this case. A page on the LRU being isolated
> >> > in __isolate_lru_page will have an elevated reference count which will
> >> > pin the inode until remove_mapping is called which holds the page lock
> >> > while inode truncation looking at a page for truncation also only checks
> >> > page_mapping under the page lock. Very broadly speaking, pages avoid being
> >> > added back to an inode being freed by checking the I_FREEING state.
> >> 
> >> So I'm wondering what prevents the following:
> >> 
> >> CPU1						CPU2
> >> 
> >> truncate(inode)					__isolate_lru_page()
> >>   ...
> >>   truncate_inode_page(mapping, page);
> >>     delete_from_page_cache(page)
> >>       spin_lock_irqsave(&mapping->tree_lock, flags);
> >>         __delete_from_page_cache(page, NULL)
> >>           page_cache_tree_delete(..)
> >>             ...					  mapping = page_mapping(page);
> >>             page->mapping = NULL;
> >>             ...
> >>       spin_unlock_irqrestore(&mapping->tree_lock, flags);
> >>       page_cache_free_page(mapping, page)
> >>         put_page(page)
> >>           if (put_page_testzero(page)) -> false
> >> - inode now has no pages and can be freed including embedded address_space
> >> 
> >> 						  if (mapping && !mapping->a_ops->migratepage)
> >> - we've dereferenced mapping which is potentially already free.
> >> 
> >
> > Hmm, possible if unlikely.
> >
> > Before delete_from_page_cache, we called truncate_cleanup_page so the
> > page is likely to be !PageDirty or PageWriteback which gets skipped by
> > the only caller that checks the mappping in __isolate_lru_page. The race
> > is tiny but it does exist. One way of closing it is to check the mapping
> > under the page lock which will prevent races with truncation. The
> > overhead is minimal as the calling context (compaction) is quite a heavy
> > operation anyway.
> >
> 
> I think another possible fix is to use call_rcu_sched() to free inode
> (and address_space).  Because __isolate_lru_page() will be called with
> LRU spinlock held and IRQ disabled, call_rcu_sched() will wait
> LRU spin_unlock and IRQ enabled.
> 

Maybe, but in this particular case, I would prefer to go with something
more conventional unless there is strong evidence that it's an improvement
(which I doubt in this case given the cost of migration overall and the
corner case of migrating a dirty page).

-- 
Mel Gorman
SUSE Labs