Re: [PATCH v3 2/4] mm/sparsemem: Defer the ms->section_mem_map clearing

From: Dave Hansen
Date: Fri Apr 06 2018 - 10:23:52 EST


On 02/27/2018 07:26 PM, Baoquan He wrote:
> In sparse_init(), if CONFIG_SPARSEMEM_ALLOC_MEM_MAP_TOGETHER=y, system
> will allocate one continuous memory chunk for mem maps on one node and
> populate the relevant page tables to map memory section one by one. If
> fail to populate for a certain mem section, print warning and its
> ->section_mem_map will be cleared to cancel the marking of being present.
> Like this, the number of mem sections marked as present could become
> less during sparse_init() execution.
>
> Here just defer the ms->section_mem_map clearing if failed to populate
> its page tables until the last for_each_present_section_nr() loop. This
> is in preparation for later optimizing the mem map allocation.
>
> Signed-off-by: Baoquan He <bhe@xxxxxxxxxx>
> ---
> mm/sparse-vmemmap.c | 1 -
> mm/sparse.c | 12 ++++++++----
> 2 files changed, 8 insertions(+), 5 deletions(-)
>
> diff --git a/mm/sparse-vmemmap.c b/mm/sparse-vmemmap.c
> index bd0276d5f66b..640e68f8324b 100644
> --- a/mm/sparse-vmemmap.c
> +++ b/mm/sparse-vmemmap.c
> @@ -303,7 +303,6 @@ void __init sparse_mem_maps_populate_node(struct page **map_map,
> ms = __nr_to_section(pnum);
> pr_err("%s: sparsemem memory map backing failed some memory will not be available\n",
> __func__);
> - ms->section_mem_map = 0;
> }

I think you might have been trying to say this in the description, but I
was not able to parse it out of there. What is in ms->section_mem_map
that needs to get cleared?

It *looks* like memory_present() uses ms->section_mem_map to just mark
which sections are online relatively early in boot. We need this
clearing to mark that they are effectively *not* present any longer.
Correct?

I guess the concern here is that if you miss any of the error sites,
we'll end up with a bogus, non-null ms->section_mem_map. Do we handle
that nicely?

Should the " = 0" instead be clearing SECTION_MARKED_PRESENT or
something? That would make it easier to match the code up with the code
that it is effectively undoing.