The patch titled
Subject: mm/hugetlb: use __GFP_COMP for gigantic folios
has been added to the -mm mm-unstable branch. Its filename is
mm-hugetlb-use-__gfp_comp-for-gigantic-folios.patch
This patch will shortly appear at
https://git.kernel.org/pub/scm/linux/kernel/git/akpm/25-new.git/tree/patches/mm-hugetlb-use-__gfp_comp-for-gigantic-folios.patch
This patch will later appear in the mm-unstable branch at
git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
Before you just go and hit "reply", please:
a) Consider who else should be cc'ed
b) Prefer to cc a suitable mailing list as well
c) Ideally: find the original patch on the mailing list and do a
reply-to-all to that, adding suitable additional cc's
*** Remember to use Documentation/process/submit-checklist.rst when testing your code ***
The -mm tree is included into linux-next via the mm-everything
branch at git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
and is updated there every 2-3 working days
------------------------------------------------------
From: Yu Zhao <yuzhao@xxxxxxxxxx>
Subject: mm/hugetlb: use __GFP_COMP for gigantic folios
Date: Sun, 11 Aug 2024 15:21:29 -0600
Use __GFP_COMP for gigantic folios to greatly reduce not only the code
but also the allocation and free time.
LOC (approximately): -200, +50
Allocate and free 500 1GB hugeTLB memory without HVO by:
time echo 500 >/sys/kernel/mm/hugepages/hugepages-1048576kB/nr_hugepages
time echo 0 >/sys/kernel/mm/hugepages/hugepages-1048576kB/nr_hugepages
Before After
Alloc ~13s ~10s
Free ~15s <1s
The above magnitude generally holds for multiple x86 and arm64 CPU
models.
Link: https://lkml.kernel.org/r/20240811212129.3074314-4-yuzhao@xxxxxxxxxx
Signed-off-by: Yu Zhao <yuzhao@xxxxxxxxxx>
Cc: Matthew Wilcox <willy@xxxxxxxxxxxxx>
Cc: Muchun Song <muchun.song@xxxxxxxxx>
Cc: Zi Yan <ziy@xxxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
include/linux/hugetlb.h | 9 -
mm/hugetlb.c | 239 ++++++--------------------------------
2 files changed, 48 insertions(+), 200 deletions(-)
--- a/include/linux/hugetlb.h~mm-hugetlb-use-__gfp_comp-for-gigantic-folios
+++ a/include/linux/hugetlb.h
@@ -896,10 +896,11 @@ static inline bool hugepage_movable_supp
/* Movability of hugepages depends on migration support. */
static inline gfp_t htlb_alloc_mask(struct hstate *h)
{
- if (hugepage_movable_supported(h))
- return GFP_HIGHUSER_MOVABLE;
- else
- return GFP_HIGHUSER;
+ gfp_t gfp = __GFP_COMP | __GFP_NOWARN;
+
+ gfp |= hugepage_movable_supported(h) ? GFP_HIGHUSER_MOVABLE : GFP_HIGHUSER;
+
+ return gfp;
}
static inline gfp_t htlb_modify_alloc_mask(struct hstate *h, gfp_t gfp_mask)
--- a/mm/hugetlb.c~mm-hugetlb-use-__gfp_comp-for-gigantic-folios
+++ a/mm/hugetlb.c
@@ -1512,43 +1512,7 @@ static int hstate_next_node_to_free(stru
((node = hstate_next_node_to_free(hs, mask)) || 1); \
nr_nodes--)
-/* used to demote non-gigantic_huge pages as well */
-static void __destroy_compound_gigantic_folio(struct folio *folio,
- unsigned int order, bool demote)
-{
- int i;
- int nr_pages = 1 << order;
- struct page *p;
-
- atomic_set(&folio->_entire_mapcount, 0);
- atomic_set(&folio->_large_mapcount, 0);
- atomic_set(&folio->_pincount, 0);
-
- for (i = 1; i < nr_pages; i++) {
- p = folio_page(folio, i);
- p->flags &= ~PAGE_FLAGS_CHECK_AT_FREE;
- p->mapping = NULL;
- clear_compound_head(p);
- if (!demote)
- set_page_refcounted(p);
- }
-
- __folio_clear_head(folio);
-}
-
-static void destroy_compound_hugetlb_folio_for_demote(struct folio *folio,
- unsigned int order)
-{
- __destroy_compound_gigantic_folio(folio, order, true);
-}
-
#ifdef CONFIG_ARCH_HAS_GIGANTIC_PAGE
-static void destroy_compound_gigantic_folio(struct folio *folio,
- unsigned int order)
-{
- __destroy_compound_gigantic_folio(folio, order, false);
-}
-
static void free_gigantic_folio(struct folio *folio, unsigned int order)
{
/*
@@ -1569,38 +1533,52 @@ static void free_gigantic_folio(struct f
static struct folio *alloc_gigantic_folio(struct hstate *h, gfp_t gfp_mask,
int nid, nodemask_t *nodemask)
{
- struct page *page;
- unsigned long nr_pages = pages_per_huge_page(h);
+ struct folio *folio;
+ int order = huge_page_order(h);
+ bool retry = false;
+
if (nid == NUMA_NO_NODE)
nid = numa_mem_id();
-
+retry:
+ folio = NULL;
#ifdef CONFIG_CMA
{
int node;
- if (hugetlb_cma[nid]) {
- page = cma_alloc(hugetlb_cma[nid], nr_pages,
- huge_page_order(h), true);
- if (page)
- return page_folio(page);
- }
+ if (hugetlb_cma[nid])
+ folio = cma_alloc_folio(hugetlb_cma[nid], order, gfp_mask);
- if (!(gfp_mask & __GFP_THISNODE)) {
+ if (!folio && !(gfp_mask & __GFP_THISNODE)) {
for_each_node_mask(node, *nodemask) {
if (node == nid || !hugetlb_cma[node])
continue;
- page = cma_alloc(hugetlb_cma[node], nr_pages,
- huge_page_order(h), true);
- if (page)
- return page_folio(page);
+ folio = cma_alloc_folio(hugetlb_cma[node], order, gfp_mask);
+ if (folio)
+ break;
}
}
}
#endif
+ if (!folio) {
+ struct page *page = alloc_contig_pages(1 << order, gfp_mask, nid, nodemask);
+
+ if (!page)
+ return NULL;
- page = alloc_contig_pages(nr_pages, gfp_mask, nid, nodemask);
- return page ? page_folio(page) : NULL;
+ folio = page_folio(page);
+ }
+
+ if (folio_ref_freeze(folio, 1))
+ return folio;
+
+ pr_warn("HugeTLB: unexpected refcount on PFN %lu\n", folio_pfn(folio));
+ free_gigantic_folio(folio, order);
+ if (!retry) {
+ retry = true;
+ goto retry;
+ }
+ return NULL;
}
#else /* !CONFIG_CONTIG_ALLOC */
@@ -1619,8 +1597,6 @@ static struct folio *alloc_gigantic_foli
}
static inline void free_gigantic_folio(struct folio *folio,
unsigned int order) { }
-static inline void destroy_compound_gigantic_folio(struct folio *folio,
- unsigned int order) { }
#endif
/*
@@ -1747,6 +1723,7 @@ static void __update_and_free_hugetlb_fo
folio_clear_hugetlb_hwpoison(folio);
folio_ref_unfreeze(folio, 1);
+ INIT_LIST_HEAD(&folio->_deferred_list);
/*
* Non-gigantic pages demoted from CMA allocated gigantic pages
@@ -1754,10 +1731,8 @@ static void __update_and_free_hugetlb_fo
*/
if (hstate_is_gigantic(h) ||
hugetlb_cma_folio(folio, huge_page_order(h))) {
- destroy_compound_gigantic_folio(folio, huge_page_order(h));
free_gigantic_folio(folio, huge_page_order(h));
} else {
- INIT_LIST_HEAD(&folio->_deferred_list);
folio_clear_partially_mapped(folio);
folio_put(folio);
}
@@ -2033,95 +2008,6 @@ static void prep_new_hugetlb_folio(struc
spin_unlock_irq(&hugetlb_lock);
}
-static bool __prep_compound_gigantic_folio(struct folio *folio,
- unsigned int order, bool demote)
-{
- int i, j;
- int nr_pages = 1 << order;
- struct page *p;
-
- __folio_clear_reserved(folio);
- for (i = 0; i < nr_pages; i++) {
- p = folio_page(folio, i);
-
- /*
- * For gigantic hugepages allocated through bootmem at
- * boot, it's safer to be consistent with the not-gigantic
- * hugepages and clear the PG_reserved bit from all tail pages
- * too. Otherwise drivers using get_user_pages() to access tail
- * pages may get the reference counting wrong if they see
- * PG_reserved set on a tail page (despite the head page not
- * having PG_reserved set). Enforcing this consistency between
- * head and tail pages allows drivers to optimize away a check
- * on the head page when they need know if put_page() is needed
- * after get_user_pages().
- */
- if (i != 0) /* head page cleared above */
- __ClearPageReserved(p);
- /*
- * Subtle and very unlikely
- *
- * Gigantic 'page allocators' such as memblock or cma will
- * return a set of pages with each page ref counted. We need
- * to turn this set of pages into a compound page with tail
- * page ref counts set to zero. Code such as speculative page
- * cache adding could take a ref on a 'to be' tail page.
- * We need to respect any increased ref count, and only set
- * the ref count to zero if count is currently 1. If count
- * is not 1, we return an error. An error return indicates
- * the set of pages can not be converted to a gigantic page.
- * The caller who allocated the pages should then discard the
- * pages using the appropriate free interface.
- *
- * In the case of demote, the ref count will be zero.
- */
- if (!demote) {
- if (!page_ref_freeze(p, 1)) {
- pr_warn("HugeTLB page can not be used due to unexpected inflated ref count\n");
- goto out_error;
- }
- } else {
- VM_BUG_ON_PAGE(page_count(p), p);
- }
- if (i != 0)
- set_compound_head(p, &folio->page);
- }
- __folio_set_head(folio);
- /* we rely on prep_new_hugetlb_folio to set the hugetlb flag */
- folio_set_order(folio, order);
- atomic_set(&folio->_entire_mapcount, -1);
- atomic_set(&folio->_large_mapcount, -1);
- atomic_set(&folio->_pincount, 0);
- return true;
-
-out_error:
- /* undo page modifications made above */
- for (j = 0; j < i; j++) {
- p = folio_page(folio, j);
- if (j != 0)
- clear_compound_head(p);
- set_page_refcounted(p);
- }
- /* need to clear PG_reserved on remaining tail pages */
- for (; j < nr_pages; j++) {
- p = folio_page(folio, j);
- __ClearPageReserved(p);
- }
- return false;
-}
-
-static bool prep_compound_gigantic_folio(struct folio *folio,
- unsigned int order)
-{
- return __prep_compound_gigantic_folio(folio, order, false);
-}
-
-static bool prep_compound_gigantic_folio_for_demote(struct folio *folio,
- unsigned int order)
-{
- return __prep_compound_gigantic_folio(folio, order, true);
-}
-
/*
* Find and lock address space (mapping) in write mode.
*
@@ -2160,7 +2046,6 @@ static struct folio *alloc_buddy_hugetlb
*/
if (node_alloc_noretry && node_isset(nid, *node_alloc_noretry))
alloc_try_hard = false;
- gfp_mask |= __GFP_COMP|__GFP_NOWARN;
if (alloc_try_hard)
gfp_mask |= __GFP_RETRY_MAYFAIL;
if (nid == NUMA_NO_NODE)
@@ -2207,48 +2092,14 @@ retry:
return folio;
}
-static struct folio *__alloc_fresh_hugetlb_folio(struct hstate *h,
- gfp_t gfp_mask, int nid, nodemask_t *nmask,
- nodemask_t *node_alloc_noretry)
-{
- struct folio *folio;
- bool retry = false;
-
-retry:
- if (hstate_is_gigantic(h))
- folio = alloc_gigantic_folio(h, gfp_mask, nid, nmask);
- else
- folio = alloc_buddy_hugetlb_folio(h, gfp_mask,
- nid, nmask, node_alloc_noretry);
- if (!folio)
- return NULL;
-
- if (hstate_is_gigantic(h)) {
- if (!prep_compound_gigantic_folio(folio, huge_page_order(h))) {
- /*
- * Rare failure to convert pages to compound page.
- * Free pages and try again - ONCE!
- */
- free_gigantic_folio(folio, huge_page_order(h));
- if (!retry) {
- retry = true;
- goto retry;
- }
- return NULL;
- }
- }
-
- return folio;
-}
-
static struct folio *only_alloc_fresh_hugetlb_folio(struct hstate *h,
gfp_t gfp_mask, int nid, nodemask_t *nmask,
nodemask_t *node_alloc_noretry)
{
struct folio *folio;
- folio = __alloc_fresh_hugetlb_folio(h, gfp_mask, nid, nmask,
- node_alloc_noretry);
+ folio = hstate_is_gigantic(h) ? alloc_gigantic_folio(h, gfp_mask, nid, nmask) :
+ alloc_buddy_hugetlb_folio(h, gfp_mask, nid, nmask, node_alloc_noretry);
if (folio)
init_new_hugetlb_folio(h, folio);
return folio;
@@ -2266,7 +2117,8 @@ static struct folio *alloc_fresh_hugetlb
{
struct folio *folio;
- folio = __alloc_fresh_hugetlb_folio(h, gfp_mask, nid, nmask, NULL);
+ folio = hstate_is_gigantic(h) ? alloc_gigantic_folio(h, gfp_mask, nid, nmask) :
+ alloc_buddy_hugetlb_folio(h, gfp_mask, nid, nmask, NULL);
if (!folio)
return NULL;
@@ -2550,9 +2402,8 @@ struct folio *alloc_buddy_hugetlb_folio_
nid = huge_node(vma, addr, gfp_mask, &mpol, &nodemask);
if (mpol_is_preferred_many(mpol)) {
- gfp_t gfp = gfp_mask | __GFP_NOWARN;
+ gfp_t gfp = gfp_mask & ~(__GFP_DIRECT_RECLAIM | __GFP_NOFAIL);
- gfp &= ~(__GFP_DIRECT_RECLAIM | __GFP_NOFAIL);
folio = alloc_surplus_hugetlb_folio(h, gfp, nid, nodemask);
/* Fallback to all nodes if page==NULL */
@@ -3334,6 +3185,7 @@ static void __init hugetlb_folio_init_ta
for (pfn = head_pfn + start_page_number; pfn < end_pfn; pfn++) {
struct page *page = pfn_to_page(pfn);
+ __ClearPageReserved(folio_page(folio, pfn - head_pfn));
__init_single_page(page, pfn, zone, nid);
prep_compound_tail((struct page *)folio, pfn - head_pfn);
ret = page_ref_freeze(page, 1);
@@ -3950,21 +3802,16 @@ static long demote_free_hugetlb_folios(s
continue;
list_del(&folio->lru);
- /*
- * Use destroy_compound_hugetlb_folio_for_demote for all huge page
- * sizes as it will not ref count folios.
- */
- destroy_compound_hugetlb_folio_for_demote(folio, huge_page_order(src));
+
+ split_page_owner(&folio->page, huge_page_order(src), huge_page_order(dst));
+ pgalloc_tag_split(&folio->page, 1 << huge_page_order(src));
for (i = 0; i < pages_per_huge_page(src); i += pages_per_huge_page(dst)) {
struct page *page = folio_page(folio, i);
- if (hstate_is_gigantic(dst))
- prep_compound_gigantic_folio_for_demote(page_folio(page),
- dst->order);
- else
- prep_compound_page(page, dst->order);
- set_page_private(page, 0);
+ page->mapping = NULL;
+ clear_compound_head(page);
+ prep_compound_page(page, dst->order);
init_new_hugetlb_folio(dst, page_folio(page));
list_add(&page->lru, &dst_list);
_
Patches currently in -mm which might be from yuzhao@xxxxxxxxxx are
mm-hugetlb_vmemmap-dont-synchronize_rcu-without-hvo.patch
mm-swap-reduce-indentation-level.patch
mm-swap-rename-cpu_fbatches-activate.patch
mm-swap-fold-lru_rotate-into-cpu_fbatches.patch
mm-swap-remove-remaining-_fn-suffix.patch
mm-swap-remove-boilerplate.patch
mm-swap-remove-boilerplate-fix.patch
mm-free-zapped-tail-pages-when-splitting-isolated-thp.patch
mm-remap-unused-subpages-to-shared-zeropage-when-splitting-isolated-thp.patch
mm-hugetlb_vmemmap-batch-hvo-work-when-demoting.patch
mm-contig_alloc-support-__gfp_comp.patch
mm-cma-add-cma_alloc_folio.patch
mm-hugetlb-use-__gfp_comp-for-gigantic-folios.patch
