* [PATCH 1/2] mm: move mm counter updating out of set_pte_range()
2024-04-11 13:09 [PATCH 0/2] mm: batch mm counter updating in filemap_map_pages() Kefeng Wang
@ 2024-04-11 13:09 ` Kefeng Wang
2024-04-11 12:53 ` Matthew Wilcox
2024-04-11 13:09 ` [PATCH 2/2] mm: filemap: batch mm counter updating in filemap_map_pages() Kefeng Wang
1 sibling, 1 reply; 5+ messages in thread
From: Kefeng Wang @ 2024-04-11 13:09 UTC (permalink / raw
To: Andrew Morton
Cc: Matthew Wilcox (Oracle), linux-mm, linux-fsdevel, Kefeng Wang
In order to support batch mm counter updating in filemap_map_pages(),
make set_pte_range() return the type of MM_COUNTERS and move mm counter
updating out of set_pte_range().
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
---
include/linux/mm.h | 4 ++--
mm/filemap.c | 10 +++++++---
mm/memory.c | 16 +++++++++++-----
3 files changed, 20 insertions(+), 10 deletions(-)
diff --git a/include/linux/mm.h b/include/linux/mm.h
index 0b4046b1e63d..6ad440ac3706 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -1366,8 +1366,8 @@ static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
}
vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
-void set_pte_range(struct vm_fault *vmf, struct folio *folio,
- struct page *page, unsigned int nr, unsigned long addr);
+int set_pte_range(struct vm_fault *vmf, struct folio *folio,
+ struct page *page, unsigned int nr, unsigned long addr);
vm_fault_t finish_fault(struct vm_fault *vmf);
#endif
diff --git a/mm/filemap.c b/mm/filemap.c
index 92e2d43e4c9d..2274e590bab4 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -3512,6 +3512,7 @@ static vm_fault_t filemap_map_folio_range(struct vm_fault *vmf,
struct page *page = folio_page(folio, start);
unsigned int count = 0;
pte_t *old_ptep = vmf->pte;
+ int type;
do {
if (PageHWPoison(page + count))
@@ -3539,7 +3540,8 @@ static vm_fault_t filemap_map_folio_range(struct vm_fault *vmf,
continue;
skip:
if (count) {
- set_pte_range(vmf, folio, page, count, addr);
+ type = set_pte_range(vmf, folio, page, count, addr);
+ add_mm_counter(vmf->vma->vm_mm, type, count);
folio_ref_add(folio, count);
if (in_range(vmf->address, addr, count * PAGE_SIZE))
ret = VM_FAULT_NOPAGE;
@@ -3553,7 +3555,8 @@ static vm_fault_t filemap_map_folio_range(struct vm_fault *vmf,
} while (--nr_pages > 0);
if (count) {
- set_pte_range(vmf, folio, page, count, addr);
+ type = set_pte_range(vmf, folio, page, count, addr);
+ add_mm_counter(vmf->vma->vm_mm, type, count);
folio_ref_add(folio, count);
if (in_range(vmf->address, addr, count * PAGE_SIZE))
ret = VM_FAULT_NOPAGE;
@@ -3589,7 +3592,8 @@ static vm_fault_t filemap_map_order0_folio(struct vm_fault *vmf,
if (vmf->address == addr)
ret = VM_FAULT_NOPAGE;
- set_pte_range(vmf, folio, page, 1, addr);
+ add_mm_counter(vmf->vma->vm_mm,
+ set_pte_range(vmf, folio, page, 1, addr), 1);
folio_ref_inc(folio);
return ret;
diff --git a/mm/memory.c b/mm/memory.c
index 78422d1c7381..485ffec9d4c7 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -4661,15 +4661,18 @@ vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page)
* @page: The first page to create a PTE for.
* @nr: The number of PTEs to create.
* @addr: The first address to create a PTE for.
+ *
+ * Return: type of MM_COUNTERS to be updated
*/
-void set_pte_range(struct vm_fault *vmf, struct folio *folio,
- struct page *page, unsigned int nr, unsigned long addr)
+int set_pte_range(struct vm_fault *vmf, struct folio *folio,
+ struct page *page, unsigned int nr, unsigned long addr)
{
struct vm_area_struct *vma = vmf->vma;
bool uffd_wp = vmf_orig_pte_uffd_wp(vmf);
bool write = vmf->flags & FAULT_FLAG_WRITE;
bool prefault = in_range(vmf->address, addr, nr * PAGE_SIZE);
pte_t entry;
+ int type;
flush_icache_pages(vma, page, nr);
entry = mk_pte(page, vma->vm_page_prot);
@@ -4685,18 +4688,20 @@ void set_pte_range(struct vm_fault *vmf, struct folio *folio,
entry = pte_mkuffd_wp(entry);
/* copy-on-write page */
if (write && !(vma->vm_flags & VM_SHARED)) {
- add_mm_counter(vma->vm_mm, MM_ANONPAGES, nr);
+ type = MM_ANONPAGES;
VM_BUG_ON_FOLIO(nr != 1, folio);
folio_add_new_anon_rmap(folio, vma, addr);
folio_add_lru_vma(folio, vma);
} else {
- add_mm_counter(vma->vm_mm, mm_counter_file(folio), nr);
+ type = mm_counter_file(folio);
folio_add_file_rmap_ptes(folio, page, nr, vma);
}
set_ptes(vma->vm_mm, addr, vmf->pte, entry, nr);
/* no need to invalidate: a not-present page won't be cached */
update_mmu_cache_range(vmf, vma, addr, vmf->pte, nr);
+
+ return type;
}
static bool vmf_pte_changed(struct vm_fault *vmf)
@@ -4765,8 +4770,9 @@ vm_fault_t finish_fault(struct vm_fault *vmf)
/* Re-check under ptl */
if (likely(!vmf_pte_changed(vmf))) {
struct folio *folio = page_folio(page);
+ int type = set_pte_range(vmf, folio, page, 1, vmf->address);
- set_pte_range(vmf, folio, page, 1, vmf->address);
+ add_mm_counter(vmf->vma->vm_mm, type, 1);
ret = 0;
} else {
update_mmu_tlb(vma, vmf->address, vmf->pte);
--
2.41.0
^ permalink raw reply related [flat|nested] 5+ messages in thread* [PATCH 2/2] mm: filemap: batch mm counter updating in filemap_map_pages()
2024-04-11 13:09 [PATCH 0/2] mm: batch mm counter updating in filemap_map_pages() Kefeng Wang
2024-04-11 13:09 ` [PATCH 1/2] mm: move mm counter updating out of set_pte_range() Kefeng Wang
@ 2024-04-11 13:09 ` Kefeng Wang
1 sibling, 0 replies; 5+ messages in thread
From: Kefeng Wang @ 2024-04-11 13:09 UTC (permalink / raw
To: Andrew Morton
Cc: Matthew Wilcox (Oracle), linux-mm, linux-fsdevel, Kefeng Wang
Like copy_pte_range()/zap_pte_range(), make mm counter batch updating
in filemap_map_pages(), the 'lat_pagefault -P 1 file' test from lmbench
shows 12% improve, and the percpu_counter_add_batch() is gone from
perf flame graph.
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
---
include/linux/mm.h | 14 ++++++++++++++
mm/filemap.c | 19 +++++++++++--------
mm/memory.c | 14 --------------
3 files changed, 25 insertions(+), 22 deletions(-)
diff --git a/include/linux/mm.h b/include/linux/mm.h
index 6ad440ac3706..c7dffd358088 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -2655,6 +2655,20 @@ static inline int mm_counter(struct folio *folio)
return mm_counter_file(folio);
}
+static inline void init_rss_vec(int *rss)
+{
+ memset(rss, 0, sizeof(int) * NR_MM_COUNTERS);
+}
+
+static inline void add_mm_rss_vec(struct mm_struct *mm, int *rss)
+{
+ int i;
+
+ for (i = 0; i < NR_MM_COUNTERS; i++)
+ if (rss[i])
+ add_mm_counter(mm, i, rss[i]);
+}
+
static inline unsigned long get_mm_rss(struct mm_struct *mm)
{
return get_mm_counter(mm, MM_FILEPAGES) +
diff --git a/mm/filemap.c b/mm/filemap.c
index 2274e590bab4..d8b23e976a43 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -3506,7 +3506,7 @@ static struct folio *next_uptodate_folio(struct xa_state *xas,
static vm_fault_t filemap_map_folio_range(struct vm_fault *vmf,
struct folio *folio, unsigned long start,
unsigned long addr, unsigned int nr_pages,
- unsigned int *mmap_miss)
+ int *rss, unsigned int *mmap_miss)
{
vm_fault_t ret = 0;
struct page *page = folio_page(folio, start);
@@ -3541,7 +3541,7 @@ static vm_fault_t filemap_map_folio_range(struct vm_fault *vmf,
skip:
if (count) {
type = set_pte_range(vmf, folio, page, count, addr);
- add_mm_counter(vmf->vma->vm_mm, type, count);
+ rss[type] += count;
folio_ref_add(folio, count);
if (in_range(vmf->address, addr, count * PAGE_SIZE))
ret = VM_FAULT_NOPAGE;
@@ -3556,7 +3556,7 @@ static vm_fault_t filemap_map_folio_range(struct vm_fault *vmf,
if (count) {
type = set_pte_range(vmf, folio, page, count, addr);
- add_mm_counter(vmf->vma->vm_mm, type, count);
+ rss[type] += count;
folio_ref_add(folio, count);
if (in_range(vmf->address, addr, count * PAGE_SIZE))
ret = VM_FAULT_NOPAGE;
@@ -3569,7 +3569,7 @@ static vm_fault_t filemap_map_folio_range(struct vm_fault *vmf,
static vm_fault_t filemap_map_order0_folio(struct vm_fault *vmf,
struct folio *folio, unsigned long addr,
- unsigned int *mmap_miss)
+ int *rss, unsigned int *mmap_miss)
{
vm_fault_t ret = 0;
struct page *page = &folio->page;
@@ -3592,8 +3592,7 @@ static vm_fault_t filemap_map_order0_folio(struct vm_fault *vmf,
if (vmf->address == addr)
ret = VM_FAULT_NOPAGE;
- add_mm_counter(vmf->vma->vm_mm,
- set_pte_range(vmf, folio, page, 1, addr), 1);
+ rss[set_pte_range(vmf, folio, page, 1, addr)]++;
folio_ref_inc(folio);
return ret;
@@ -3610,6 +3609,7 @@ vm_fault_t filemap_map_pages(struct vm_fault *vmf,
XA_STATE(xas, &mapping->i_pages, start_pgoff);
struct folio *folio;
vm_fault_t ret = 0;
+ int rss[NR_MM_COUNTERS];
unsigned int nr_pages = 0, mmap_miss = 0, mmap_miss_saved;
rcu_read_lock();
@@ -3629,6 +3629,8 @@ vm_fault_t filemap_map_pages(struct vm_fault *vmf,
folio_put(folio);
goto out;
}
+
+ init_rss_vec(rss);
do {
unsigned long end;
@@ -3640,15 +3642,16 @@ vm_fault_t filemap_map_pages(struct vm_fault *vmf,
if (!folio_test_large(folio))
ret |= filemap_map_order0_folio(vmf,
- folio, addr, &mmap_miss);
+ folio, addr, rss, &mmap_miss);
else
ret |= filemap_map_folio_range(vmf, folio,
xas.xa_index - folio->index, addr,
- nr_pages, &mmap_miss);
+ nr_pages, rss, &mmap_miss);
folio_unlock(folio);
folio_put(folio);
} while ((folio = next_uptodate_folio(&xas, mapping, end_pgoff)) != NULL);
+ add_mm_rss_vec(vma->vm_mm, rss);
pte_unmap_unlock(vmf->pte, vmf->ptl);
out:
rcu_read_unlock();
diff --git a/mm/memory.c b/mm/memory.c
index 485ffec9d4c7..149208da1652 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -465,20 +465,6 @@ int __pte_alloc_kernel(pmd_t *pmd)
return 0;
}
-static inline void init_rss_vec(int *rss)
-{
- memset(rss, 0, sizeof(int) * NR_MM_COUNTERS);
-}
-
-static inline void add_mm_rss_vec(struct mm_struct *mm, int *rss)
-{
- int i;
-
- for (i = 0; i < NR_MM_COUNTERS; i++)
- if (rss[i])
- add_mm_counter(mm, i, rss[i]);
-}
-
/*
* This function is called to print an error when a bad pte
* is found. For example, we might have a PFN-mapped pte in
--
2.41.0
^ permalink raw reply related [flat|nested] 5+ messages in thread