[v1,1/6] mm: hugetlb: Skip prep of tail pages when HVO is enabled
Commit Message
When vmemmap is optimizable, it will free all the
duplicated tail pages in hugetlb_vmemmap_optimize while
preparing the new hugepage. Hence, there is no need to
prepare them.
For 1G x86 hugepages, it avoids preparing
262144 - 64 = 262080 struct pages per hugepage.
Signed-off-by: Usama Arif <usama.arif@bytedance.com>
---
mm/hugetlb.c | 32 +++++++++++++++++++++++---------
mm/hugetlb_vmemmap.c | 2 +-
mm/hugetlb_vmemmap.h | 7 +++++++
3 files changed, 31 insertions(+), 10 deletions(-)
Comments
Hi Usama,
kernel test robot noticed the following build warnings:
[auto build test WARNING on akpm-mm/mm-everything]
url: https://github.com/intel-lab-lkp/linux/commits/Usama-Arif/mm-hugetlb-Skip-prep-of-tail-pages-when-HVO-is-enabled/20230728-044839
base: https://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm.git mm-everything
patch link: https://lore.kernel.org/r/20230727204624.1942372-2-usama.arif%40bytedance.com
patch subject: [v1 1/6] mm: hugetlb: Skip prep of tail pages when HVO is enabled
config: arm64-randconfig-r016-20230727 (https://download.01.org/0day-ci/archive/20230728/202307281636.rXw2Dlzr-lkp@intel.com/config)
compiler: aarch64-linux-gcc (GCC) 12.3.0
reproduce: (https://download.01.org/0day-ci/archive/20230728/202307281636.rXw2Dlzr-lkp@intel.com/reproduce)
If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202307281636.rXw2Dlzr-lkp@intel.com/
All warnings (new ones prefixed by >>):
In file included from mm/hugetlb.c:49:
>> mm/hugetlb_vmemmap.h:56:6: warning: no previous prototype for 'vmemmap_should_optimize' [-Wmissing-prototypes]
56 | bool vmemmap_should_optimize(const struct hstate *h, const struct page *head)
| ^~~~~~~~~~~~~~~~~~~~~~~
vim +/vmemmap_should_optimize +56 mm/hugetlb_vmemmap.h
55
> 56 bool vmemmap_should_optimize(const struct hstate *h, const struct page *head)
57 {
58 return false;
59 }
60
Hi Usama,
kernel test robot noticed the following build warnings:
[auto build test WARNING on akpm-mm/mm-everything]
url: https://github.com/intel-lab-lkp/linux/commits/Usama-Arif/mm-hugetlb-Skip-prep-of-tail-pages-when-HVO-is-enabled/20230728-044839
base: https://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm.git mm-everything
patch link: https://lore.kernel.org/r/20230727204624.1942372-2-usama.arif%40bytedance.com
patch subject: [v1 1/6] mm: hugetlb: Skip prep of tail pages when HVO is enabled
config: arm64-randconfig-r032-20230727 (https://download.01.org/0day-ci/archive/20230728/202307281922.cdKbT0tX-lkp@intel.com/config)
compiler: clang version 17.0.0 (https://github.com/llvm/llvm-project.git 4a5ac14ee968ff0ad5d2cc1ffa0299048db4c88a)
reproduce: (https://download.01.org/0day-ci/archive/20230728/202307281922.cdKbT0tX-lkp@intel.com/reproduce)
If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202307281922.cdKbT0tX-lkp@intel.com/
All warnings (new ones prefixed by >>):
In file included from mm/hugetlb.c:49:
>> mm/hugetlb_vmemmap.h:56:6: warning: no previous prototype for function 'vmemmap_should_optimize' [-Wmissing-prototypes]
56 | bool vmemmap_should_optimize(const struct hstate *h, const struct page *head)
| ^
mm/hugetlb_vmemmap.h:56:1: note: declare 'static' if the function is not intended to be used outside of this translation unit
56 | bool vmemmap_should_optimize(const struct hstate *h, const struct page *head)
| ^
| static
1 warning generated.
vim +/vmemmap_should_optimize +56 mm/hugetlb_vmemmap.h
55
> 56 bool vmemmap_should_optimize(const struct hstate *h, const struct page *head)
57 {
58 return false;
59 }
60
On Thu, Jul 27, 2023 at 09:46:19PM +0100, Usama Arif wrote:
> When vmemmap is optimizable, it will free all the
> duplicated tail pages in hugetlb_vmemmap_optimize while
> preparing the new hugepage. Hence, there is no need to
> prepare them.
>
> For 1G x86 hugepages, it avoids preparing
> 262144 - 64 = 262080 struct pages per hugepage.
>
> Signed-off-by: Usama Arif <usama.arif@bytedance.com>
> ---
> mm/hugetlb.c | 32 +++++++++++++++++++++++---------
> mm/hugetlb_vmemmap.c | 2 +-
> mm/hugetlb_vmemmap.h | 7 +++++++
> 3 files changed, 31 insertions(+), 10 deletions(-)
>
> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> index 64a3239b6407..58cf5978bee1 100644
> --- a/mm/hugetlb.c
> +++ b/mm/hugetlb.c
> @@ -1943,13 +1943,24 @@ static void prep_new_hugetlb_folio(struct hstate *h, struct folio *folio, int ni
> }
>
> static bool __prep_compound_gigantic_folio(struct folio *folio,
> - unsigned int order, bool demote)
> + unsigned int order, bool demote,
> + bool hvo)
I think it would be cleaner to pass struct hstate * instead of order here
so that order and hvo can be computed locally.
> {
> int i, j;
> int nr_pages = 1 << order;
> struct page *p;
>
> __folio_clear_reserved(folio);
> +
> +#ifdef CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP
> + /*
> + * No need to prep pages that will be freed later by hugetlb_vmemmap_optimize
> + * in prep_new_huge_page. Hence, reduce nr_pages to the pages that will be kept.
> + */
> + if (hvo)
if (IS_ENABLED(CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP) && hvo)
is better than ifdef IMO.
> + nr_pages = HUGETLB_VMEMMAP_RESERVE_SIZE / sizeof(struct page);
> +#endif
> +
> for (i = 0; i < nr_pages; i++) {
> p = folio_page(folio, i);
>
> @@ -2020,15 +2031,15 @@ static bool __prep_compound_gigantic_folio(struct folio *folio,
> }
>
> static bool prep_compound_gigantic_folio(struct folio *folio,
> - unsigned int order)
> + unsigned int order, bool hvo)
> {
> - return __prep_compound_gigantic_folio(folio, order, false);
> + return __prep_compound_gigantic_folio(folio, order, false, hvo);
> }
>
> static bool prep_compound_gigantic_folio_for_demote(struct folio *folio,
> - unsigned int order)
> + unsigned int order, bool hvo)
> {
> - return __prep_compound_gigantic_folio(folio, order, true);
> + return __prep_compound_gigantic_folio(folio, order, true, hvo);
> }
>
> /*
> @@ -2185,7 +2196,8 @@ static struct folio *alloc_fresh_hugetlb_folio(struct hstate *h,
> if (!folio)
> return NULL;
> if (hstate_is_gigantic(h)) {
> - if (!prep_compound_gigantic_folio(folio, huge_page_order(h))) {
> + if (!prep_compound_gigantic_folio(folio, huge_page_order(h),
> + vmemmap_should_optimize(h, &folio->page))) {
> /*
> * Rare failure to convert pages to compound page.
> * Free pages and try again - ONCE!
> @@ -3201,7 +3213,8 @@ static void __init gather_bootmem_prealloc(void)
>
> VM_BUG_ON(!hstate_is_gigantic(h));
> WARN_ON(folio_ref_count(folio) != 1);
> - if (prep_compound_gigantic_folio(folio, huge_page_order(h))) {
> + if (prep_compound_gigantic_folio(folio, huge_page_order(h),
> + vmemmap_should_optimize(h, page))) {
> WARN_ON(folio_test_reserved(folio));
> prep_new_hugetlb_folio(h, folio, folio_nid(folio));
> free_huge_page(page); /* add to the hugepage allocator */
> @@ -3624,8 +3637,9 @@ static int demote_free_hugetlb_folio(struct hstate *h, struct folio *folio)
> subpage = folio_page(folio, i);
> inner_folio = page_folio(subpage);
> if (hstate_is_gigantic(target_hstate))
> - prep_compound_gigantic_folio_for_demote(inner_folio,
> - target_hstate->order);
> + prep_compound_gigantic_folio_for_demote(folio,
> + target_hstate->order,
> + vmemmap_should_optimize(target_hstate, subpage));
> else
> prep_compound_page(subpage, target_hstate->order);
> folio_change_private(inner_folio, NULL);
> diff --git a/mm/hugetlb_vmemmap.c b/mm/hugetlb_vmemmap.c
> index c2007ef5e9b0..b721e87de2b3 100644
> --- a/mm/hugetlb_vmemmap.c
> +++ b/mm/hugetlb_vmemmap.c
> @@ -486,7 +486,7 @@ int hugetlb_vmemmap_restore(const struct hstate *h, struct page *head)
> }
>
> /* Return true iff a HugeTLB whose vmemmap should and can be optimized. */
> -static bool vmemmap_should_optimize(const struct hstate *h, const struct page *head)
> +bool vmemmap_should_optimize(const struct hstate *h, const struct page *head)
> {
> if (!READ_ONCE(vmemmap_optimize_enabled))
> return false;
> diff --git a/mm/hugetlb_vmemmap.h b/mm/hugetlb_vmemmap.h
> index 25bd0e002431..07555d2dc0cb 100644
> --- a/mm/hugetlb_vmemmap.h
> +++ b/mm/hugetlb_vmemmap.h
> @@ -13,6 +13,7 @@
> #ifdef CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP
> int hugetlb_vmemmap_restore(const struct hstate *h, struct page *head);
> void hugetlb_vmemmap_optimize(const struct hstate *h, struct page *head);
> +bool vmemmap_should_optimize(const struct hstate *h, const struct page *head);
>
> /*
> * Reserve one vmemmap page, all vmemmap addresses are mapped to it. See
> @@ -51,6 +52,12 @@ static inline unsigned int hugetlb_vmemmap_optimizable_size(const struct hstate
> {
> return 0;
> }
> +
> +bool vmemmap_should_optimize(const struct hstate *h, const struct page *head)
> +{
> + return false;
> +}
> +
> #endif /* CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP */
>
> static inline bool hugetlb_vmemmap_optimizable(const struct hstate *h)
> --
> 2.25.1
>
@@ -1943,13 +1943,24 @@ static void prep_new_hugetlb_folio(struct hstate *h, struct folio *folio, int ni
}
static bool __prep_compound_gigantic_folio(struct folio *folio,
- unsigned int order, bool demote)
+ unsigned int order, bool demote,
+ bool hvo)
{
int i, j;
int nr_pages = 1 << order;
struct page *p;
__folio_clear_reserved(folio);
+
+#ifdef CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP
+ /*
+ * No need to prep pages that will be freed later by hugetlb_vmemmap_optimize
+ * in prep_new_huge_page. Hence, reduce nr_pages to the pages that will be kept.
+ */
+ if (hvo)
+ nr_pages = HUGETLB_VMEMMAP_RESERVE_SIZE / sizeof(struct page);
+#endif
+
for (i = 0; i < nr_pages; i++) {
p = folio_page(folio, i);
@@ -2020,15 +2031,15 @@ static bool __prep_compound_gigantic_folio(struct folio *folio,
}
static bool prep_compound_gigantic_folio(struct folio *folio,
- unsigned int order)
+ unsigned int order, bool hvo)
{
- return __prep_compound_gigantic_folio(folio, order, false);
+ return __prep_compound_gigantic_folio(folio, order, false, hvo);
}
static bool prep_compound_gigantic_folio_for_demote(struct folio *folio,
- unsigned int order)
+ unsigned int order, bool hvo)
{
- return __prep_compound_gigantic_folio(folio, order, true);
+ return __prep_compound_gigantic_folio(folio, order, true, hvo);
}
/*
@@ -2185,7 +2196,8 @@ static struct folio *alloc_fresh_hugetlb_folio(struct hstate *h,
if (!folio)
return NULL;
if (hstate_is_gigantic(h)) {
- if (!prep_compound_gigantic_folio(folio, huge_page_order(h))) {
+ if (!prep_compound_gigantic_folio(folio, huge_page_order(h),
+ vmemmap_should_optimize(h, &folio->page))) {
/*
* Rare failure to convert pages to compound page.
* Free pages and try again - ONCE!
@@ -3201,7 +3213,8 @@ static void __init gather_bootmem_prealloc(void)
VM_BUG_ON(!hstate_is_gigantic(h));
WARN_ON(folio_ref_count(folio) != 1);
- if (prep_compound_gigantic_folio(folio, huge_page_order(h))) {
+ if (prep_compound_gigantic_folio(folio, huge_page_order(h),
+ vmemmap_should_optimize(h, page))) {
WARN_ON(folio_test_reserved(folio));
prep_new_hugetlb_folio(h, folio, folio_nid(folio));
free_huge_page(page); /* add to the hugepage allocator */
@@ -3624,8 +3637,9 @@ static int demote_free_hugetlb_folio(struct hstate *h, struct folio *folio)
subpage = folio_page(folio, i);
inner_folio = page_folio(subpage);
if (hstate_is_gigantic(target_hstate))
- prep_compound_gigantic_folio_for_demote(inner_folio,
- target_hstate->order);
+ prep_compound_gigantic_folio_for_demote(folio,
+ target_hstate->order,
+ vmemmap_should_optimize(target_hstate, subpage));
else
prep_compound_page(subpage, target_hstate->order);
folio_change_private(inner_folio, NULL);
@@ -486,7 +486,7 @@ int hugetlb_vmemmap_restore(const struct hstate *h, struct page *head)
}
/* Return true iff a HugeTLB whose vmemmap should and can be optimized. */
-static bool vmemmap_should_optimize(const struct hstate *h, const struct page *head)
+bool vmemmap_should_optimize(const struct hstate *h, const struct page *head)
{
if (!READ_ONCE(vmemmap_optimize_enabled))
return false;
@@ -13,6 +13,7 @@
#ifdef CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP
int hugetlb_vmemmap_restore(const struct hstate *h, struct page *head);
void hugetlb_vmemmap_optimize(const struct hstate *h, struct page *head);
+bool vmemmap_should_optimize(const struct hstate *h, const struct page *head);
/*
* Reserve one vmemmap page, all vmemmap addresses are mapped to it. See
@@ -51,6 +52,12 @@ static inline unsigned int hugetlb_vmemmap_optimizable_size(const struct hstate
{
return 0;
}
+
+bool vmemmap_should_optimize(const struct hstate *h, const struct page *head)
+{
+ return false;
+}
+
#endif /* CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP */
static inline bool hugetlb_vmemmap_optimizable(const struct hstate *h)