Re: [Patch v4 08/12] memory-hotplug: remove memmap of sparse-vmemmap

[Wen Congyang <wency@cn.fujitsu.com>]

At 11/28/2012 05:40 PM, Jianguo Wu Wrote:
> Hi Congyang,
> 
> I think vmemmap's pgtable pages should be freed after all entries are cleared, I have a patch to do this.
> The code logic is the same as [Patch v4 09/12] memory-hotplug: remove page table of x86_64 architecture.
> 
> How do you think about this?
> 
> Signed-off-by: Jianguo Wu <wujianguo@huawei.com>
> Signed-off-by: Jiang Liu <jiang.liu@huawei.com>
> ---
>  include/linux/mm.h  |    1 +
>  mm/sparse-vmemmap.c |  214 +++++++++++++++++++++++++++++++++++++++++++++++++++
>  mm/sparse.c         |    5 +-
>  3 files changed, 218 insertions(+), 2 deletions(-)
> 
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> index 5657670..1f26af5 100644
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -1642,6 +1642,7 @@ int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
>  void vmemmap_populate_print_last(void);
>  void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
>  				  unsigned long size);
> +void vmemmap_free(struct page *memmap, unsigned long nr_pages);
>  
>  enum mf_flags {
>  	MF_COUNT_INCREASED = 1 << 0,
> diff --git a/mm/sparse-vmemmap.c b/mm/sparse-vmemmap.c
> index 1b7e22a..242cb28 100644
> --- a/mm/sparse-vmemmap.c
> +++ b/mm/sparse-vmemmap.c
> @@ -29,6 +29,10 @@
>  #include <asm/pgalloc.h>
>  #include <asm/pgtable.h>
>  
> +#ifdef CONFIG_MEMORY_HOTREMOVE
> +#include <asm/tlbflush.h>
> +#endif
> +
>  /*
>   * Allocate a block of memory to be used to back the virtual memory map
>   * or to back the page tables that are used to create the mapping.
> @@ -224,3 +228,213 @@ void __init sparse_mem_maps_populate_node(struct page **map_map,
>  		vmemmap_buf_end = NULL;
>  	}
>  }
> +
> +#ifdef CONFIG_MEMORY_HOTREMOVE
> +static void vmemmap_free_pages(struct page *page, int order)
> +{
> +	struct zone *zone;
> +	unsigned long magic;
> +
> +	magic = (unsigned long) page->lru.next;
> +	if (magic == SECTION_INFO || magic == MIX_SECTION_INFO) {
> +		put_page_bootmem(page);
> +
> +		zone = page_zone(page);
> +		zone_span_writelock(zone);
> +		zone->present_pages++;
> +		zone_span_writeunlock(zone);
> +		totalram_pages++;
> +	} else {
> +		if (is_vmalloc_addr(page_address(page)))
> +			vfree(page_address(page));

Hmm, vmemmap doesn't use vmalloc() to allocate memory.

> +		else
> +			free_pages((unsigned long)page_address(page), order);
> +	}
> +}
> +
> +static void free_pte_table(pmd_t *pmd)
> +{
> +	pte_t *pte, *pte_start;
> +	int i;
> +
> +	pte_start = (pte_t *)pmd_page_vaddr(*pmd);
> +	for (i = 0; i < PTRS_PER_PTE; i++) {
> +		pte = pte_start + i;
> +		if (pte_val(*pte))
> +			return;
> +	}
> +
> +	/* free a pte talbe */
> +	vmemmap_free_pages(pmd_page(*pmd), 0);
> +	spin_lock(&init_mm.page_table_lock);
> +	pmd_clear(pmd);
> +	spin_unlock(&init_mm.page_table_lock);
> +}
> +
> +static void free_pmd_table(pud_t *pud)
> +{
> +	pmd_t *pmd, *pmd_start;
> +	int i;
> +
> +	pmd_start = (pmd_t *)pud_page_vaddr(*pud);
> +	for (i = 0; i < PTRS_PER_PMD; i++) {
> +		pmd = pmd_start + i;
> +		if (pmd_val(*pmd))
> +			return;
> +	}
> +
> +	/* free a pmd talbe */
> +	vmemmap_free_pages(pud_page(*pud), 0);
> +	spin_lock(&init_mm.page_table_lock);
> +	pud_clear(pud);
> +	spin_unlock(&init_mm.page_table_lock);
> +}
> +
> +static void free_pud_table(pgd_t *pgd)
> +{
> +	pud_t *pud, *pud_start;
> +	int i;
> +
> +	pud_start = (pud_t *)pgd_page_vaddr(*pgd);
> +	for (i = 0; i < PTRS_PER_PUD; i++) {
> +		pud = pud_start + i;
> +		if (pud_val(*pud))
> +			return;
> +	}
> +
> +	/* free a pud table */
> +	vmemmap_free_pages(pgd_page(*pgd), 0);
> +	spin_lock(&init_mm.page_table_lock);
> +	pgd_clear(pgd);
> +	spin_unlock(&init_mm.page_table_lock);
> +}
> +
> +static int split_large_page(pte_t *kpte, unsigned long address, pte_t *pbase)
> +{
> +	struct page *page = pmd_page(*(pmd_t *)kpte);
> +	int i = 0;
> +	unsigned long magic;
> +	unsigned long section_nr;
> +
> +	__split_large_page(kpte, address, pbase);
> +	__flush_tlb_all();
> +
> +	magic = (unsigned long) page->lru.next;
> +	if (magic == SECTION_INFO) {
> +		section_nr = pfn_to_section_nr(page_to_pfn(page));
> +		while (i < PTRS_PER_PMD) {
> +			page++;
> +			i++;
> +			get_page_bootmem(section_nr, page, SECTION_INFO);
> +		}
> +	}
> +
> +	return 0;
> +}
> +
> +static void vmemmap_pte_remove(pmd_t *pmd, unsigned long addr, unsigned long end)
> +{
> +	pte_t *pte;
> +	unsigned long next;
> +
> +	pte = pte_offset_kernel(pmd, addr);
> +	for (; addr < end; pte++, addr += PAGE_SIZE) {
> +		next = (addr + PAGE_SIZE) & PAGE_MASK;
> +		if (next > end)
> +			next = end;
> +
> +		if (pte_none(*pte))
> +			continue;
> +		if (IS_ALIGNED(addr, PAGE_SIZE) &&
> +		    IS_ALIGNED(end, PAGE_SIZE)) {
> +			vmemmap_free_pages(pte_page(*pte), 0);
> +			spin_lock(&init_mm.page_table_lock);
> +			pte_clear(&init_mm, addr, pte);
> +			spin_unlock(&init_mm.page_table_lock);

If addr or end is not alianed with PAGE_SIZE, you may leak some
memory.

> +		}
> +	}
> +
> +	free_pte_table(pmd);
> +	__flush_tlb_all();
> +}
> +
> +static void vmemmap_pmd_remove(pud_t *pud, unsigned long addr, unsigned long end)
> +{
> +	unsigned long next;
> +	pmd_t *pmd;
> +
> +	pmd = pmd_offset(pud, addr);
> +	for (; addr < end; addr = next, pmd++) {
> +		next = pmd_addr_end(addr, end);
> +		if (pmd_none(*pmd))
> +			continue;
> +
> +		if (cpu_has_pse) {
> +			unsigned long pte_base;
> +
> +			if (IS_ALIGNED(addr, PMD_SIZE) &&
> +			    IS_ALIGNED(next, PMD_SIZE)) {
> +				vmemmap_free_pages(pmd_page(*pmd),
> +						   get_order(PMD_SIZE));
> +				spin_lock(&init_mm.page_table_lock);
> +				pmd_clear(pmd);
> +				spin_unlock(&init_mm.page_table_lock);
> +				continue;
> +			}
> +
> +			/*
> +			 * We use 2M page, but we need to remove part of them,
> +			 * so split 2M page to 4K page.
> +			 */
> +			pte_base = get_zeroed_page(GFP_ATOMIC | __GFP_NOTRACK);

get_zeored_page() may fail. You should handle this error.

> +			split_large_page((pte_t *)pmd, addr, (pte_t *)pte_base);
> +			__flush_tlb_all();
> +
> +			spin_lock(&init_mm.page_table_lock);
> +			pmd_populate_kernel(&init_mm, pmd, (pte_t *)pte_base);
> +			spin_unlock(&init_mm.page_table_lock);
> +		}
> +
> +		vmemmap_pte_remove(pmd, addr, next);
> +	}
> +
> +	free_pmd_table(pud);
> +	__flush_tlb_all();
> +}
> +
> +static void vmemmap_pud_remove(pgd_t *pgd, unsigned long addr, unsigned long end)
> +{
> +	unsigned long next;
> +	pud_t *pud;
> +
> +	pud = pud_offset(pgd, addr);
> +	for (; addr < end; addr = next, pud++) {
> +		next = pud_addr_end(addr, end);
> +		if (pud_none(*pud))
> +			continue;
> +
> +		vmemmap_pmd_remove(pud, addr, next);
> +	}
> +
> +	free_pud_table(pgd);
> +	__flush_tlb_all();
> +}
> +
> +void vmemmap_free(struct page *memmap, unsigned long nr_pages)
> +{
> +	unsigned long addr = (unsigned long)memmap;
> +	unsigned long end = (unsigned long)(memmap + nr_pages);
> +	unsigned long next;
> +
> +	for (; addr < end; addr = next) {
> +		pgd_t *pgd = pgd_offset_k(addr);
> +
> +		next = pgd_addr_end(addr, end);
> +		if (!pgd_present(*pgd))
> +			continue;
> +
> +		vmemmap_pud_remove(pgd, addr, next);
> +		sync_global_pgds(addr, next);

The parameter for sync_global_pgds() is [start, end], not
[start, end)

> +	}
> +}
> +#endif
> diff --git a/mm/sparse.c b/mm/sparse.c
> index fac95f2..3a16d68 100644
> --- a/mm/sparse.c
> +++ b/mm/sparse.c
> @@ -613,12 +613,13 @@ static inline struct page *kmalloc_section_memmap(unsigned long pnum, int nid,
>  	/* This will make the necessary allocations eventually. */
>  	return sparse_mem_map_populate(pnum, nid);
>  }
> -static void __kfree_section_memmap(struct page *memmap, unsigned long nr_pages)
> +static void __kfree_section_memmap(struct page *page, unsigned long nr_pages)
Why do you change this line?

>  {
> -	return; /* XXX: Not implemented yet */
> +	vmemmap_free(page, nr_pages);
>  }
>  static void free_map_bootmem(struct page *page, unsigned long nr_pages)
>  {
> +	vmemmap_free(page, nr_pages);
>  }
>  #else
>  static struct page *__kmalloc_section_memmap(unsigned long nr_pages)