Message-ID: <55B64F1D.8090807@citrix.com>
Date: Mon, 27 Jul 2015 16:32:45 +0100
From: David Vrabel <david.vrabel@citrix.com>
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:31.0) Gecko/20100101 Icedove/31.7.0
MIME-Version: 1.0
To: Mel Gorman <mgorman@suse.de>
CC: "linux-mm@kvack.org" <linux-mm@kvack.org>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Subject: vmemmap_verify() BUGs during memory hotplug (4.2-rc1 regression)
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1260
Lines: 29

Mel,

As of commit 8a942fdea560d4ac0e9d9fabcd5201ad20e0c382 (mm: meminit: make
__early_pfn_to_nid SMP-safe and introduce meminit_pfn_in_nid)
vmemmap_verify() will BUG_ON() during memory hotplug because of its use
of early_pfn_to_nid().  Previously, it would have reported bogus (or
failed to report valid) warnings.

I believe this does not affect memory hotplug on most x86 systems
because vmemmap_populate() would normally call
vmemmap_populate_hugepages() which avoids calling vmemmap_verify() in
the common case (no existing mappings covering the new area).

I'm triggering the early_pfn_to_nid() BUG_ON() with the Xen balloon
driver in a PV guest which will always call vmemmap_populate_basepages()
(since Xen PV guests lack superpage support).

Not really sure what the best way to resolve this is.  Presumably
vmmemmap_verify() needs to switch to using pfn_to_nid() after the
initial initialization but there doesn't appear to be anything suitable
to distinguish between the early and hotplug cases.

David

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/