[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <4df63333-de57-4a58-a110-77b4fdfa6a9e@redhat.com>
Date: Mon, 30 Oct 2023 11:20:21 +0100
From: David Hildenbrand <david@...hat.com>
To: Vishal Verma <vishal.l.verma@...el.com>,
Andrew Morton <akpm@...ux-foundation.org>,
Oscar Salvador <osalvador@...e.de>,
Dan Williams <dan.j.williams@...el.com>,
Dave Jiang <dave.jiang@...el.com>
Cc: linux-kernel@...r.kernel.org, linux-mm@...ck.org,
nvdimm@...ts.linux.dev, linux-cxl@...r.kernel.org,
Huang Ying <ying.huang@...el.com>,
Dave Hansen <dave.hansen@...ux.intel.com>,
"Aneesh Kumar K.V" <aneesh.kumar@...ux.ibm.com>,
Michal Hocko <mhocko@...e.com>,
Jonathan Cameron <Jonathan.Cameron@...wei.com>,
Jeff Moyer <jmoyer@...hat.com>
Subject: Re: [PATCH v7 2/3] mm/memory_hotplug: split memmap_on_memory requests
across memblocks
On 26.10.23 00:44, Vishal Verma 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.
>
> This does preclude being able to use PUD mappings in the direct map; a
> proposal to how this could be optimized in the future is laid out
> here[1].
Almost there, I think :)
>
> +static int create_altmaps_and_memory_blocks(int nid, struct memory_group *group,
> + u64 start, u64 size)
> +{
> + unsigned long memblock_size = memory_block_size_bytes();
> + u64 cur_start;
> + int ret;
> +
> + for (cur_start = start; cur_start < start + size;
> + cur_start += memblock_size) {
> + struct mhp_params params = { .pgprot =
> + pgprot_mhp(PAGE_KERNEL) };
> + struct vmem_altmap mhp_altmap = {
> + .base_pfn = PHYS_PFN(cur_start),
> + .end_pfn = PHYS_PFN(cur_start + memblock_size - 1),
> + };
> +
> + mhp_altmap.free = memory_block_memmap_on_memory_pages();
> + params.altmap = kmemdup(&mhp_altmap, sizeof(struct vmem_altmap),
> + GFP_KERNEL);
> + if (!params.altmap)
> + return -ENOMEM;
Best to cleanup here instead of handling it in the caller [as noted by
Vishal, we might not be doing that yet]. Using
remove_memory_blocks_and_altmaps() on the fully processed range sounds
reasonable.
maybe something like
ret = arch_add_memory(nid, cur_start, memblock_size, ¶ms);
if (ret) {
kfree(params.altmap);
goto out;
}
ret = create_memory_block_devices(cur_start, memblock_size,
params.altmap, group);
if (ret) {
arch_remove_memory(cur_start, memblock_size, NULL);
kfree(params.altmap);
goto out;
}
if (ret && cur_start != start)
remove_memory_blocks_and_altmaps(start, cur_start - start);
return ret;
> +
> + /* call arch's memory hotadd */
> + ret = arch_add_memory(nid, cur_start, memblock_size, ¶ms);
> + if (ret < 0) {
> + kfree(params.altmap);
> + return ret;
> + }
> +
> + /* create memory block devices after memory was added */
> + ret = create_memory_block_devices(cur_start, memblock_size,
> + params.altmap, group);
> + if (ret) {
> + arch_remove_memory(cur_start, memblock_size, NULL);
> + kfree(params.altmap);
> + return ret;
> + }
> + }
> +
> + return 0;
> +}
> +
[...]
> static int check_cpu_on_node(int nid)
> {
> int cpu;
> @@ -2146,11 +2186,69 @@ void try_offline_node(int nid)
> }
> EXPORT_SYMBOL(try_offline_node);
>
> -static int __ref try_remove_memory(u64 start, u64 size)
> +static void __ref remove_memory_blocks_and_altmaps(u64 start, u64 size)
> {
> - struct memory_block *mem;
> - int rc = 0, nid = NUMA_NO_NODE;
> + unsigned long memblock_size = memory_block_size_bytes();
> struct vmem_altmap *altmap = NULL;
> + struct memory_block *mem;
> + u64 cur_start;
> + int rc;
> +
> + /*
> + * For memmap_on_memory, the altmaps could have been added on
> + * a per-memblock basis. Loop through the entire range if so,
> + * and remove each memblock and its altmap.
> + */
/*
* altmaps where added on a per-memblock basis; we have to process
* each individual memory block.
*/
> + for (cur_start = start; cur_start < start + size;
> + cur_start += memblock_size) {
> + rc = walk_memory_blocks(cur_start, memblock_size, &mem,
> + test_has_altmap_cb);
> + if (rc) {
> + altmap = mem->altmap;
> + /*
> + * Mark altmap NULL so that we can add a debug
> + * check on memblock free.
> + */
> + mem->altmap = NULL;
> + }
Simpler (especially, we know that there must be an altmap):
mem = find_memory_block(pfn_to_section_nr(cur_start));
altmap = mem->altmap;
mem->altmap = NULL;
I think we might be able to remove test_has_altmap_cb() then.
> +
> + remove_memory_block_devices(cur_start, memblock_size);
> +
> + arch_remove_memory(cur_start, memblock_size, altmap);
> +
> + /* Verify that all vmemmap pages have actually been freed. */
> + if (altmap) {
There must be an altmap, so this can be done unconditionally.
> + WARN(altmap->alloc, "Altmap not fully unmapped");
> + kfree(altmap);
> + }
> + }
> +}
--
Cheers,
David / dhildenb
Powered by blists - more mailing lists