From mboxrd@z Thu Jan  1 00:00:00 1970
From: Cyril Hrubis <chrubis@suse.cz>
Date: Wed, 10 May 2017 14:21:40 +0200
Subject: [LTP] [RFC] [PATCH] move_pages12: Allocate and free hugepages
 prior	the test
In-Reply-To: <510505896.9544779.1494406598104.JavaMail.zimbra@redhat.com>
References: <20170509140458.26343-1-chrubis@suse.cz>
 <510505896.9544779.1494406598104.JavaMail.zimbra@redhat.com>
Message-ID: <20170510122130.GD29838@rei.suse.de>
List-Id: <ltp.lists.linux.it>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
To: ltp@lists.linux.it

Hi!
> I'm still getting sporadic failures with 4.11 kernel. It's freshly
> booted system, so I would expect fragmentation to be low:
> 
> # numactl -H; ./move_pages12 
> available: 2 nodes (0-1)
> node 0 cpus: 0 2 4 6 8 10 12 14 16 18 20 22
> node 0 size: 31963 MB
> node 0 free: 31600 MB
> node 1 cpus: 1 3 5 7 9 11 13 15 17 19 21 23
> node 1 size: 32251 MB
> node 1 free: 31915 MB
> node distances:
> node   0   1 
>   0:  10  20 
>   1:  20  10
> 
> tst_test.c:847: INFO: Timeout per run is 0h 05m 00s
> move_pages12.c:184: INFO: Free RAM 65040204 kB
> move_pages12.c:139: INFO: Allocating and freeing 2 hugepages on node 0
> move_pages12.c:139: INFO: Allocating and freeing 2 hugepages on node 1
> nodes: 0 1
> move_pages12.c:87: FAIL: move_pages failed: ENOMEM

Hmm, reproduced here as well. One of the 100 runs failed for me as well.

I've looked at the code in mm/migrate.c but so far I could not spot a
place where it could fail with ENOMEM (apart from very unlikely single
page allocation which is used to store the parameters passed to the
syscall).

If I comment out the memset() in the test main loop the failure does not
seem to happen. So my guess is that the failure happens when we try to
isolate a page that is in the process of teardown and we get the ENOMEM
somewhere from the memory management internals and that it's OK to
ignore this failure. But that is just a wild guess.

-- 
Cyril Hrubis
chrubis@suse.cz