Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Wed, 25 Jul 2018 13:11:15 -0500
From:   John Allen <jallen@linux.ibm.com>
To:     linux-kernel@vger.kernel.org, linuxppc-dev@lists.ozlabs.org
Cc:     jallen@linux.ibm.com, kamezawa.hiroyu@jp.fujitsu.com,
        n-horiguchi@ah.jp.nec.com, mgorman@suse.de, mhocko@suse.cz
Subject: Infinite looping observed in __offline_pages
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Disposition: inline
User-Agent: NeoMutt/20180622-63-e52393
Message-Id: <20180725181115.hmlyd3tmnu3mn3sf@p50.austin.ibm.com>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Hi All,

Under heavy stress and constant memory hot add/remove, I have observed 
the following loop to occasionally loop infinitely:

mm/memory_hotplug.c:__offline_pages

repeat:
        /* start memory hot removal */
        ret = -EINTR;
        if (signal_pending(current))
                goto failed_removal;

        cond_resched();
        lru_add_drain_all();
        drain_all_pages(zone);

        pfn = scan_movable_pages(start_pfn, end_pfn);
        if (pfn) { /* We have movable pages */
                ret = do_migrate_range(pfn, end_pfn);
                goto repeat;
        }

What appears to be happening in this case is that do_migrate_range 
returns a failure code which is being ignored. The failure is stemming 
from migrate_pages returning "1" which I'm guessing is the result of us 
hitting the following case:

mm/migrate.c: migrate_pages

	default:
		/*
		 * Permanent failure (-EBUSY, -ENOSYS, etc.):
		 * unlike -EAGAIN case, the failed page is
		 * removed from migration page list and not
		 * retried in the next outer loop.
		 */
		nr_failed++;
		break;
	}

Does a failure in do_migrate_range indicate that the range is 
unmigratable and the loop in __offline_pages should terminate and goto 
failed_removal? Or should we allow a certain number of retrys before we
give up on migrating the range?

This issue was observed on a ppc64le lpar on a 4.18-rc6 kernel.

-John