lkml.org 
[lkml]   [2018]   [Nov]   [13]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
Date
SubjectRe: [PATCH 2/5] mm/memory_hotplug: Create add/del_device_memory functions
On Mon, Oct 15, 2018 at 8:31 AM Oscar Salvador
<osalvador@techadventures.net> wrote:
>
> From: Oscar Salvador <osalvador@suse.de>
>
> HMM/devm have a particular handling of memory-hotplug.
> They do not go through the common path, and so, they do not
> call either offline_pages() or online_pages().
>
> The operations they perform are the following ones:
>
> 1) Create the linear mapping in case the memory is not private
> 2) Initialize the pages and add the sections
> 3) Move the pages to ZONE_DEVICE
>
> Due to this particular handling of hot-add/remove memory from HMM/devm,
> I think it would be nice to provide a helper function in order to
> make this cleaner, and not populate other regions with code
> that should belong to memory-hotplug.
>
> The helpers are named:
>
> del_device_memory
> add_device_memory
>
> The idea is that add_device_memory will be in charge of:
>
> a) call either arch_add_memory() or add_pages(), depending on whether
> we want a linear mapping
> b) online the memory sections that correspond to the pfn range
> c) call move_pfn_range_to_zone() being zone ZONE_DEVICE to
> expand zone/pgdat spanned pages and initialize its pages
>
> del_device_memory, on the other hand, will be in charge of:
>
> a) offline the memory sections that correspond to the pfn range
> b) call shrink_zone_pgdat_pages(), which shrinks node/zone spanned pages.
> c) call either arch_remove_memory() or __remove_pages(), depending on
> whether we need to tear down the linear mapping or not
>
> The reason behind step b) from add_device_memory() and step a)
> from del_device_memory is that now find_smallest/biggest_section_pfn
> will have to check for online sections, and not for valid sections as
> they used to do, because we call offline_mem_sections() in
> offline_pages().
>
> In order to split up better the patches and ease the review,
> this patch will only make a) case work for add_device_memory(),
> and case c) for del_device_memory.
>
> The other cases will be added in the next patch.
>
> These two functions have to be called from devm/HMM code:
>
> dd_device_memory:
> - devm_memremap_pages()
> - hmm_devmem_pages_create()
>
> del_device_memory:
> - hmm_devmem_release
> - devm_memremap_pages_release
>
> One thing I do not know is whether we can move kasan calls out of the
> hotplug lock or not.
> If we can, we could move the hotplug lock within add/del_device_memory().
>
> Signed-off-by: Oscar Salvador <osalvador@suse.de>
> ---
> include/linux/memory_hotplug.h | 11 +++++++++++
> kernel/memremap.c | 11 ++++-------
> mm/hmm.c | 33 +++++++++++++++++----------------
> mm/memory_hotplug.c | 41 +++++++++++++++++++++++++++++++++++++++++
> 4 files changed, 73 insertions(+), 23 deletions(-)

This collides with the refactoring of hmm, to be done in terms of
devm_memremap_pages(). I'd rather not introduce another common
function *beneath* hmm and devm_memremap_pages() and rather make
devm_memremap_pages() the common function.

I plan to resubmit that cleanup after Plumbers. So, unless I'm
misunderstanding some other benefit a nak from me on this patch as it
stands currently.

\
 
 \ /
  Last update: 2018-11-12 20:22    [W:0.176 / U:0.472 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site