Re: [PATCH 0/2] fix devm_memremap_pages() mem hotplug locking

From: Dan Williams
Date: Sun Feb 12 2017 - 20:35:06 EST


On Sun, Feb 12, 2017 at 5:25 PM, Masayoshi Mizuma
<m.mizuma@xxxxxxxxxxxxxx> wrote:
> Hi Dan,
>
> On Sun, 12 Feb 2017 14:34:11 -0800 Dan Williams wrote:
>>
>> Ben notes that commit f931ab479dd2 "mm: fix devm_memremap_pages crash,
>> use mem_hotplug_{begin, done}" is incomplete and broken. Writes to
>> mem_hotplug.active_writer need to be coordinated under the device
>> hotplug lock. Otherwise, we can potentially corrupt mem_hotplug.refcount
>> leading to soft lockups.
>
>
> I think mem_hotplug_{begin,done} is not suitable to exclude
> devm_memremap_pages()
> because it seems that memory hotplug is not related to this context.
> How about using pgmap_lock instead?

The problem with that switch is the race in
kernel_physical_mapping_init(), see commit f931ab479dd2. As far as I
can see we need all paths that call kernel_physical_mapping_init() to
agree on the same lock, and can't use a private lock in
devm_memremap_pages().