Re: [PATCH v2] mm,ksm: fix endless looping in allocating memory when ksm enable

[Michal Hocko <mhocko@suse.cz>]

On Tue 20-09-16 09:07:43, Michal Hocko wrote:
> [CCing Tetsuo again - please make sure you CC everybody who did respond
>  in earlier versions of the patch]

now for real

> 
> I am sorry to insist here but this doesn't address the previous review
> feedback. Let me try to show you what I would find much better. I do not
> insist on this precise wording of course but I do insist on mentioning
> the current state and making clear why GFP_NORETRY is really ok.
> 
> On Tue 20-09-16 13:50:13, zhongjiang wrote:
> > From: zhong jiang <zhongjiang@huawei.com>
> > 
> > I hit the following issue when run a OOM case of the LTP and
> > ksm enable.
> 
> "
> I hit the following hung task when running an OOM LTP test case with 4.1
> kernel.
> "
> 
> > 
> > Call trace:
> > [<ffffffc000086a88>] __switch_to+0x74/0x8c
> > [<ffffffc000a1bae0>] __schedule+0x23c/0x7bc
> > [<ffffffc000a1c09c>] schedule+0x3c/0x94
> > [<ffffffc000a1eb84>] rwsem_down_write_failed+0x214/0x350
> > [<ffffffc000a1e32c>] down_write+0x64/0x80
> > [<ffffffc00021f794>] __ksm_exit+0x90/0x19c
> > [<ffffffc0000be650>] mmput+0x118/0x11c
> > [<ffffffc0000c3ec4>] do_exit+0x2dc/0xa74
> > [<ffffffc0000c46f8>] do_group_exit+0x4c/0xe4
> > [<ffffffc0000d0f34>] get_signal+0x444/0x5e0
> > [<ffffffc000089fcc>] do_signal+0x1d8/0x450
> > [<ffffffc00008a35c>] do_notify_resume+0x70/0x78
> > 
> > it will leads to a hung task because the exiting task cannot get the
> > mmap sem for write. but the root cause is that the ksmd holds it for
> > read while allocateing memory which just takes ages to complete.
> > and ksmd will loop in the following path.
> 
> "
> The oom victim cannot terminate because it needs to take mmap_sem for
> write while the lock is held by ksmd for read which loops in the page
> allocator
> 
> ksm_do_scan
> 	scan_get_next_rmap_item
> 		down_read
> 		get_next_rmap_item
> 			alloc_rmap_item   #ksmd will loop permanently.
> 
> There is not way forward because the oom victim cannot release any
> memory in 4.1 based kernel. Since 4.6 we have the oom reaper which would
> solve this problem because it would release the memory asynchronously.
> Nevertheless we can relax alloc_rmap_item requirements and use
> __GFP_NORETRY because the allocation failure is acceptable as
> ksm_do_scan would just retry later after the lock got dropped.
> 
> Such a patch would be also easy to backport to older stable kernels
> which do not have oom_reaper.
> 
> While we are at it add GFP_NOWARN as the admin doesn't have to be
> alarmed by the allocation failure.
> > 
> > CC: <stable@vger.kernel.org>
> > Suggested-by: Hugh Dickins <hughd@google.com>
> > Suggested-by: Michal Hocko <mhocko@suse.cz>
> > Signed-off-by: zhong jiang <zhongjiang@huawei.com>
> > ---
> >  mm/ksm.c | 3 ++-
> >  1 file changed, 2 insertions(+), 1 deletion(-)
> > 
> > diff --git a/mm/ksm.c b/mm/ksm.c
> > index 73d43ba..5048083 100644
> > --- a/mm/ksm.c
> > +++ b/mm/ksm.c
> > @@ -283,7 +283,8 @@ static inline struct rmap_item *alloc_rmap_item(void)
> >  {
> >  	struct rmap_item *rmap_item;
> >  
> > -	rmap_item = kmem_cache_zalloc(rmap_item_cache, GFP_KERNEL);
> > +	rmap_item = kmem_cache_zalloc(rmap_item_cache, GFP_KERNEL |
> > +						__GFP_NORETRY | __GFP_NOWARN);
> >  	if (rmap_item)
> >  		ksm_rmap_items++;
> >  	return rmap_item;
> > -- 
> > 1.8.3.1
> 
> -- 
> Michal Hocko
> SUSE Labs

-- 
Michal Hocko
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>