Date:   Mon, 5 Nov 2018 13:38:37 +0100
From:   Michal Hocko <mhocko@...nel.org>
To:     Baoquan He <bhe@...hat.com>
Cc:     Andrew Morton <akpm@...ux-foundation.org>, linux-mm@...ck.org,
        LKML <linux-kernel@...r.kernel.org>,
        Stable tree <stable@...r.kernel.org>
Subject: Re: [PATCH] mm, memory_hotplug: teach has_unmovable_pages about of
 LRU migrateable pages

On Mon 05-11-18 18:25:20, Baoquan He wrote:
> Hi Michal,
> 
> On 11/05/18 at 10:28am, Michal Hocko wrote:
> > 
> > Or something like this. Ugly as hell, no question about that. I also
> > have to think about this some more to convince myself this will not
> > result in an endless loop under some situations.
> 
> It failed. Paste the log and patch diff here, please help check if I made
> any mistake on manual code change. The log is at bottom.

The retry patch is obviously still racy, it just makes the race window
slightly smaller and I hoped it would catch most of those races but this
is obviously not the case.

I was thinking about your MIGRATE_MOVABLE check some more and I still do
not like it much, we just change migrate type at many places and I have
hard time to actually see this is always safe wrt. to what we need here.

We should be able to restore the zone type check though. The
primary problem fixed by 15c30bc09085 ("mm, memory_hotplug: make
has_unmovable_pages more robust") was that early allocations made it to
the zone_movable range. If we add the check _after_ the PageReserved()
check then we should be able to rule all bootmem allocation out.

So what about the following (on top of the previous patch which makes
sense on its own I believe).


>From d7ffd1342529c892f1de8999c3a5609211599c9d Mon Sep 17 00:00:00 2001
From: Michal Hocko <mhocko@...e.com>
Date: Mon, 5 Nov 2018 13:28:51 +0100
Subject: [PATCH] mm, memory_hotplug: check zone_movable in has_unmovable_pages

Page state checks are racy. Under a heavy memory workload (e.g. stress
-m 200 -t 2h) it is quite easy to hit a race window when the page is
allocated but its state is not fully populated yet. A debugging patch to
dump the struct page state shows
: [  476.575516] has_unmovable_pages: pfn:0x10dfec00, found:0x1, count:0x0
: [  476.582103] page:ffffea0437fb0000 count:1 mapcount:1 mapping:ffff880e05239841 index:0x7f26e5000 compound_mapcount: 1
: [  476.592645] flags: 0x5fffffc0090034(uptodate|lru|active|head|swapbacked)

Note that the state has been checked for both PageLRU and PageSwapBacked
already. Closing this race completely would require some sort of retry
logic. This can be tricky and error prone (think of potential endless
or long taking loops).

Workaround this problem for movable zones at least. Such a zone should
only contain movable pages. 15c30bc09085 ("mm, memory_hotplug: make
has_unmovable_pages more robust") has told us that this is not strictly
true though. Bootmem pages should be marked reserved though so we can
move the original check after the PageReserved check. Pages from other
zones are still prone to races but we even do not pretend that memory
hotremove works for those so pre-mature failure doesn't hurt that much.

Reported-by: Baoquan He <bhe@...hat.com>
Signed-off-by: Michal Hocko <mhocko@...e.com>
---
 mm/page_alloc.c | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 48ceda313332..5b64c5bc6ea0 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -7788,6 +7788,14 @@ bool has_unmovable_pages(struct zone *zone, struct page *page, int count,
 		if (PageReserved(page))
 			goto unmovable;
 
+		/*
+		 * If the zone is movable and we have ruled out all reserved
+		 * pages then it should be reasonably safe to assume the rest
+		 * is movable.
+		 */
+		if (zone_idx(zone) == ZONE_MOVABLE)
+			continue;
+
 		/*
 		 * Hugepages are not in LRU lists, but they're movable.
 		 * We need not scan over tail pages bacause we don't
-- 
2.19.1

-- 
Michal Hocko
SUSE Labs

Powered by blists - more mailing lists