From: ZhangPeng zhangpeng362@huawei.com
Like copy_pte_range()/zap_pte_range(), make mm counter batch updating in filemap_map_pages(), since folios type are same(MM_SHMEMPAGES or MM_FILEPAGES) in filemap_map_pages(), only check the first folio type is enough, the 'lat_pagefault -P 1 file' test from lmbench shows 12% improvement, and the percpu_counter_add_batch() is gone from perf flame graph.
Kefeng Wang (2): mm: move mm counter updating out of set_pte_range() mm: filemap: batch mm counter updating in filemap_map_pages()
mm/filemap.c | 17 ++++++++++++----- mm/memory.c | 8 +++++--- 2 files changed, 17 insertions(+), 8 deletions(-)
From: Kefeng Wang wangkefeng.wang@huawei.com
maillist inclusion category: performance bugzilla: https://gitee.com/openeuler/kernel/issues/I9GBVL CVE: NA
Reference: https://lore.kernel.org/all/20240412064751.119015-2-wangkefeng.wang@huawei.c...
--------------------------------
In order to support batch mm counter updating in filemap_map_pages(), move mm counter updating out of set_pte_range(), the folios are file from filemap, and distinguish folios by vmf->flags and vma->vm_flags from another caller finish_fault().
Signed-off-by: Kefeng Wang wangkefeng.wang@huawei.com Signed-off-by: ZhangPeng zhangpeng362@huawei.com --- mm/filemap.c | 4 ++++ mm/memory.c | 8 +++++--- 2 files changed, 9 insertions(+), 3 deletions(-)
diff --git a/mm/filemap.c b/mm/filemap.c index 94c9f36b17d8..57ed2e109707 100644 --- a/mm/filemap.c +++ b/mm/filemap.c @@ -3500,6 +3500,8 @@ static vm_fault_t filemap_map_folio_range(struct vm_fault *vmf, skip: if (count) { set_pte_range(vmf, folio, page, count, addr); + add_mm_counter(vmf->vma->vm_mm, mm_counter_file(folio), + count); folio_ref_add(folio, count); if (in_range(vmf->address, addr, count * PAGE_SIZE)) ret = VM_FAULT_NOPAGE; @@ -3514,6 +3516,7 @@ static vm_fault_t filemap_map_folio_range(struct vm_fault *vmf,
if (count) { set_pte_range(vmf, folio, page, count, addr); + add_mm_counter(vmf->vma->vm_mm, mm_counter_file(folio), count); folio_ref_add(folio, count); if (in_range(vmf->address, addr, count * PAGE_SIZE)) ret = VM_FAULT_NOPAGE; @@ -3548,6 +3551,7 @@ static vm_fault_t filemap_map_order0_folio(struct vm_fault *vmf, ret = VM_FAULT_NOPAGE;
set_pte_range(vmf, folio, page, 1, addr); + add_mm_counter(vmf->vma->vm_mm, mm_counter_file(folio), 1); folio_ref_inc(folio);
return ret; diff --git a/mm/memory.c b/mm/memory.c index 83f030d6818d..2c317f517d1d 100644 --- a/mm/memory.c +++ b/mm/memory.c @@ -4660,12 +4660,10 @@ void set_pte_range(struct vm_fault *vmf, struct folio *folio, /* copy-on-write page */ add_reliable_folio_counter(folio, vma->vm_mm, nr); if (write && !(vma->vm_flags & VM_SHARED)) { - add_mm_counter(vma->vm_mm, MM_ANONPAGES, nr); 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); folio_add_file_rmap_ptes(folio, page, nr, vma); } set_ptes(vma->vm_mm, addr, vmf->pte, entry, nr); @@ -4702,9 +4700,11 @@ vm_fault_t finish_fault(struct vm_fault *vmf) struct vm_area_struct *vma = vmf->vma; struct page *page; vm_fault_t ret; + bool is_cow = (vmf->flags & FAULT_FLAG_WRITE) && + !(vma->vm_flags & VM_SHARED);
/* Did we COW the page? */ - if ((vmf->flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) + if (is_cow) page = vmf->cow_page; else page = vmf->page; @@ -4740,8 +4740,10 @@ 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 = is_cow ? MM_ANONPAGES : mm_counter_file(folio);
set_pte_range(vmf, folio, page, 1, vmf->address); + add_mm_counter(vma->vm_mm, type, 1); ret = 0; } else { update_mmu_tlb(vma, vmf->address, vmf->pte);
From: Kefeng Wang wangkefeng.wang@huawei.com
maillist inclusion category: performance bugzilla: https://gitee.com/openeuler/kernel/issues/I9GBVL CVE: NA
Reference: https://lore.kernel.org/all/20240412064751.119015-3-wangkefeng.wang@huawei.c...
--------------------------------
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% improvement, and the percpu_counter_add_batch() is gone from perf flame graph.
Signed-off-by: Kefeng Wang wangkefeng.wang@huawei.com Signed-off-by: ZhangPeng zhangpeng362@huawei.com --- mm/filemap.c | 21 ++++++++++++--------- 1 file changed, 12 insertions(+), 9 deletions(-)
diff --git a/mm/filemap.c b/mm/filemap.c index 57ed2e109707..bf4c07bded63 100644 --- a/mm/filemap.c +++ b/mm/filemap.c @@ -3474,7 +3474,7 @@ 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) + unsigned long *rss, unsigned int *mmap_miss) { vm_fault_t ret = 0; struct page *page = folio_page(folio, start); @@ -3500,8 +3500,7 @@ static vm_fault_t filemap_map_folio_range(struct vm_fault *vmf, skip: if (count) { set_pte_range(vmf, folio, page, count, addr); - add_mm_counter(vmf->vma->vm_mm, mm_counter_file(folio), - count); + *rss += count; folio_ref_add(folio, count); if (in_range(vmf->address, addr, count * PAGE_SIZE)) ret = VM_FAULT_NOPAGE; @@ -3516,7 +3515,7 @@ static vm_fault_t filemap_map_folio_range(struct vm_fault *vmf,
if (count) { set_pte_range(vmf, folio, page, count, addr); - add_mm_counter(vmf->vma->vm_mm, mm_counter_file(folio), count); + *rss += count; folio_ref_add(folio, count); if (in_range(vmf->address, addr, count * PAGE_SIZE)) ret = VM_FAULT_NOPAGE; @@ -3529,7 +3528,7 @@ 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) + unsigned long *rss, unsigned int *mmap_miss) { vm_fault_t ret = 0; struct page *page = &folio->page; @@ -3551,7 +3550,7 @@ static vm_fault_t filemap_map_order0_folio(struct vm_fault *vmf, ret = VM_FAULT_NOPAGE;
set_pte_range(vmf, folio, page, 1, addr); - add_mm_counter(vmf->vma->vm_mm, mm_counter_file(folio), 1); + (*rss)++; folio_ref_inc(folio);
return ret; @@ -3568,7 +3567,8 @@ 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; - unsigned int nr_pages = 0, mmap_miss = 0, mmap_miss_saved; + unsigned long rss = 0; + unsigned int nr_pages = 0, mmap_miss = 0, mmap_miss_saved, folio_type;
rcu_read_lock(); folio = next_uptodate_folio(&xas, mapping, end_pgoff); @@ -3587,6 +3587,8 @@ vm_fault_t filemap_map_pages(struct vm_fault *vmf, folio_put(folio); goto out; } + + folio_type = mm_counter_file(folio); do { unsigned long end;
@@ -3598,15 +3600,16 @@ 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_counter(vma->vm_mm, folio_type, rss); pte_unmap_unlock(vmf->pte, vmf->ptl); out: rcu_read_unlock();
反馈： 您发送到kernel@openeuler.org的补丁/补丁集，已成功转换为PR！ PR链接地址： https://gitee.com/openeuler/kernel/pulls/6068 邮件列表地址：https://mailweb.openeuler.org/hyperkitty/list/kernel@openeuler.org/message/V...
FeedBack: The patch(es) which you have sent to kernel@openeuler.org mailing list has been converted to a pull request successfully! Pull request link: https://gitee.com/openeuler/kernel/pulls/6068 Mailing list address: https://mailweb.openeuler.org/hyperkitty/list/kernel@openeuler.org/message/V...
-
patchwork bot -
Peng Zhang