On Wed, 2023-08-02 at 15:20 +0100, Jonathan Cameron wrote:
> On Tue, 01 Aug 2023 23:55:37 -0600
> Vishal Verma <vishal.l.verma@xxxxxxxxx> wrote:
>
> > The MHP_MEMMAP_ON_MEMORY flag for hotplugged memory is restricted to
> > 'memblock_size' chunks of memory being added. Adding a larger span of
> > memory precludes memmap_on_memory semantics.
> >
> > For users of hotplug such as kmem, large amounts of memory might get
> > added from the CXL subsystem. In some cases, this amount may exceed the
> > available 'main memory' to store the memmap for the memory being added.
> > In this case, it is useful to have a way to place the memmap on the
> > memory being added, even if it means splitting the addition into
> > memblock-sized chunks.
> >
> > Change add_memory_resource() to loop over memblock-sized chunks of
> > memory if caller requested memmap_on_memory, and if other conditions for
> > it are met. Teach try_remove_memory() to also expect that a memory
> > range being removed might have been split up into memblock sized chunks,
> > and to loop through those as needed.
> >
> > Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
> > Cc: David Hildenbrand <david@xxxxxxxxxx>
> > Cc: Michal Hocko <mhocko@xxxxxxxx>
> > Cc: Oscar Salvador <osalvador@xxxxxxx>
> > Cc: Dan Williams <dan.j.williams@xxxxxxxxx>
> > Cc: Dave Jiang <dave.jiang@xxxxxxxxx>
> > Cc: Dave Hansen <dave.hansen@xxxxxxxxxxxxxxx>
> > Cc: Huang Ying <ying.huang@xxxxxxxxx>
> > Suggested-by: David Hildenbrand <david@xxxxxxxxxx>
> > Signed-off-by: Vishal Verma <vishal.l.verma@xxxxxxxxx>
>
> A couple of trivial comments inline.
Hi Jonathan,
Thanks for taking a look.
>
> > ---
> > mm/memory_hotplug.c | 150 ++++++++++++++++++++++++++++++++--------------------
> > 1 file changed, 93 insertions(+), 57 deletions(-)
> >
> > diff --git a/mm/memory_hotplug.c b/mm/memory_hotplug.c
> > index d282664f558e..cae03c8d4bbf 100644
> > --- a/mm/memory_hotplug.c
> > +++ b/mm/memory_hotplug.c
> > @@ -1383,6 +1383,44 @@ static bool mhp_supports_memmap_on_memory(unsigned long size)
> > return arch_supports_memmap_on_memory(vmemmap_size);
> > }
> >
> > +static int add_memory_create_devices(int nid, struct memory_group *group,
> > + u64 start, u64 size, mhp_t mhp_flags)
> > +{
> > + struct mhp_params params = { .pgprot = pgprot_mhp(PAGE_KERNEL) };
> > + struct vmem_altmap mhp_altmap = {
> > + .base_pfn = PHYS_PFN(start),
> > + .end_pfn = PHYS_PFN(start + size - 1),
> > + };
> > + int ret;
> > +
> > + if ((mhp_flags & MHP_MEMMAP_ON_MEMORY)) {
> > + mhp_altmap.free = memory_block_memmap_on_memory_pages();
> > + params.altmap = kmalloc(sizeof(struct vmem_altmap), GFP_KERNEL);
> > + if (!params.altmap)
> > + return -ENOMEM;
> > +
> > + memcpy(params.altmap, &mhp_altmap, sizeof(mhp_altmap));
> > + }
> > +
> > + /* call arch's memory hotadd */
> > + ret = arch_add_memory(nid, start, size, ¶ms);
> > + if (ret < 0)
> > + goto error;
> > +
> > + /* create memory block devices after memory was added */
> > + ret = create_memory_block_devices(start, size, params.altmap, group);
> > + if (ret) {
> > + arch_remove_memory(start, size, NULL);
>
> Maybe push this down to a second label?
Yep will do.
>
<snip>
> > +
> > +static int __ref try_remove_memory(u64 start, u64 size)
> > +{
> > + int ret, nid = NUMA_NO_NODE;
>
> I'm not overly keen to see the trivial rename of rc -> ret in here.
> Just makes it ever so slightly harder to compare old code and new code.
Yep - this was to work around the patches I was based on, which added
both a ret and left the original rc [1]. Aneesh will stick to 'rc' so
my next revision should sort this out naturally.
[1]: https://lore.kernel.org/all/715042319ceb86016a4986862a82756e5629d725.camel@xxxxxxxxx/
>
