| Message ID | 1588279279-61908-2-git-send-email-yang.shi@linux.alibaba.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [v2,linux-next,1/2] mm: khugepaged: add exceed_max_ptes_* helpers | expand |
On Fri, May 01, 2020 at 04:41:19AM +0800, Yang Shi wrote: > When khugepaged successfully isolated and copied data from old page to > collapsed THP, the old page is about to be freed if its last mapcount > is gone. So putting the page back to lru sounds not that productive in > this case since the page might be isolated by vmscan but it can't be > reclaimed by vmscan since it can't be unmapped by try_to_unmap() at all. > > Actually if khugepaged is the last user of this page so it can be freed > directly. So, clearing active and unevictable flags, unlocking and > dropping refcount from isolate instead of calling putback_lru_page(). Any reason putback_lru_page() cannot do it internally? I mean if it is page_count() == 1, free the page. > > Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> > Cc: Hugh Dickins <hughd@google.com> > Cc: Andrea Arcangeli <aarcange@redhat.com> > Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com> > --- > v2: Check mapcount and skip putback lru if the last mapcount is gone > > mm/khugepaged.c | 20 ++++++++++++++------ > 1 file changed, 14 insertions(+), 6 deletions(-) > > diff --git a/mm/khugepaged.c b/mm/khugepaged.c > index 0c8d30b..1fdd677 100644 > --- a/mm/khugepaged.c > +++ b/mm/khugepaged.c > @@ -559,10 +559,18 @@ void __khugepaged_exit(struct mm_struct *mm) > static void release_pte_page(struct page *page) > { > mod_node_page_state(page_pgdat(page), > - NR_ISOLATED_ANON + page_is_file_lru(page), > - -compound_nr(page)); > - unlock_page(page); > - putback_lru_page(page); > + NR_ISOLATED_ANON + page_is_file_lru(page), -compound_nr(page)); > + > + if (total_mapcount(page)) { > + unlock_page(page); > + putback_lru_page(page); > + } else { > + ClearPageActive(page); > + ClearPageUnevictable(page); > + unlock_page(page); > + /* Drop refcount from isolate */ > + put_page(page); > + } > } > > static void release_pte_pages(pte_t *pte, pte_t *_pte, > @@ -771,8 +779,6 @@ static void __collapse_huge_page_copy(pte_t *pte, struct page *page, > } else { > src_page = pte_page(pteval); > copy_user_highpage(page, src_page, address, vma); > - if (!PageCompound(src_page)) > - release_pte_page(src_page); > /* > * ptl mostly unnecessary, but preempt has to > * be disabled to update the per-cpu stats > @@ -786,6 +792,8 @@ static void __collapse_huge_page_copy(pte_t *pte, struct page *page, > pte_clear(vma->vm_mm, address, _pte); > page_remove_rmap(src_page, false); > spin_unlock(ptl); > + if (!PageCompound(src_page)) > + release_pte_page(src_page); > free_page_and_swap_cache(src_page); > } > } > -- > 1.8.3.1 > >
On 5/1/20 12:04 AM, Kirill A. Shutemov wrote: > On Fri, May 01, 2020 at 04:41:19AM +0800, Yang Shi wrote: >> When khugepaged successfully isolated and copied data from old page to >> collapsed THP, the old page is about to be freed if its last mapcount >> is gone. So putting the page back to lru sounds not that productive in >> this case since the page might be isolated by vmscan but it can't be >> reclaimed by vmscan since it can't be unmapped by try_to_unmap() at all. >> >> Actually if khugepaged is the last user of this page so it can be freed >> directly. So, clearing active and unevictable flags, unlocking and >> dropping refcount from isolate instead of calling putback_lru_page(). > Any reason putback_lru_page() cannot do it internally? I mean if it is > page_count() == 1, free the page. Because it seems putback_lru_page() is just called when the page is *not* about to be freed, for example compaction is aborted, migration is failed on some pages, etc. I checked a couple of places, i.e. compaction, migration, vmscan. All these places free the page (i.e. old page is migrated to new page successfully) via put_page() without putting back to lru. So, it sounds free page in putback_lru_page() may never happen. We could add this, but it sounds khugepaged would be the only path that may free page in putback_lru_page(). So, it sounds less confusing to follow other users IMHO. >> Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> >> Cc: Hugh Dickins <hughd@google.com> >> Cc: Andrea Arcangeli <aarcange@redhat.com> >> Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com> >> --- >> v2: Check mapcount and skip putback lru if the last mapcount is gone >> >> mm/khugepaged.c | 20 ++++++++++++++------ >> 1 file changed, 14 insertions(+), 6 deletions(-) >> >> diff --git a/mm/khugepaged.c b/mm/khugepaged.c >> index 0c8d30b..1fdd677 100644 >> --- a/mm/khugepaged.c >> +++ b/mm/khugepaged.c >> @@ -559,10 +559,18 @@ void __khugepaged_exit(struct mm_struct *mm) >> static void release_pte_page(struct page *page) >> { >> mod_node_page_state(page_pgdat(page), >> - NR_ISOLATED_ANON + page_is_file_lru(page), >> - -compound_nr(page)); >> - unlock_page(page); >> - putback_lru_page(page); >> + NR_ISOLATED_ANON + page_is_file_lru(page), -compound_nr(page)); >> + >> + if (total_mapcount(page)) { >> + unlock_page(page); >> + putback_lru_page(page); >> + } else { >> + ClearPageActive(page); >> + ClearPageUnevictable(page); >> + unlock_page(page); >> + /* Drop refcount from isolate */ >> + put_page(page); >> + } >> } >> >> static void release_pte_pages(pte_t *pte, pte_t *_pte, >> @@ -771,8 +779,6 @@ static void __collapse_huge_page_copy(pte_t *pte, struct page *page, >> } else { >> src_page = pte_page(pteval); >> copy_user_highpage(page, src_page, address, vma); >> - if (!PageCompound(src_page)) >> - release_pte_page(src_page); >> /* >> * ptl mostly unnecessary, but preempt has to >> * be disabled to update the per-cpu stats >> @@ -786,6 +792,8 @@ static void __collapse_huge_page_copy(pte_t *pte, struct page *page, >> pte_clear(vma->vm_mm, address, _pte); >> page_remove_rmap(src_page, false); >> spin_unlock(ptl); >> + if (!PageCompound(src_page)) >> + release_pte_page(src_page); >> free_page_and_swap_cache(src_page); >> } >> } >> -- >> 1.8.3.1 >> >>
diff --git a/mm/khugepaged.c b/mm/khugepaged.c index 0c8d30b..1fdd677 100644 --- a/mm/khugepaged.c +++ b/mm/khugepaged.c @@ -559,10 +559,18 @@ void __khugepaged_exit(struct mm_struct *mm) static void release_pte_page(struct page *page) { mod_node_page_state(page_pgdat(page), - NR_ISOLATED_ANON + page_is_file_lru(page), - -compound_nr(page)); - unlock_page(page); - putback_lru_page(page); + NR_ISOLATED_ANON + page_is_file_lru(page), -compound_nr(page)); + + if (total_mapcount(page)) { + unlock_page(page); + putback_lru_page(page); + } else { + ClearPageActive(page); + ClearPageUnevictable(page); + unlock_page(page); + /* Drop refcount from isolate */ + put_page(page); + } } static void release_pte_pages(pte_t *pte, pte_t *_pte, @@ -771,8 +779,6 @@ static void __collapse_huge_page_copy(pte_t *pte, struct page *page, } else { src_page = pte_page(pteval); copy_user_highpage(page, src_page, address, vma); - if (!PageCompound(src_page)) - release_pte_page(src_page); /* * ptl mostly unnecessary, but preempt has to * be disabled to update the per-cpu stats @@ -786,6 +792,8 @@ static void __collapse_huge_page_copy(pte_t *pte, struct page *page, pte_clear(vma->vm_mm, address, _pte); page_remove_rmap(src_page, false); spin_unlock(ptl); + if (!PageCompound(src_page)) + release_pte_page(src_page); free_page_and_swap_cache(src_page); } }
When khugepaged successfully isolated and copied data from old page to collapsed THP, the old page is about to be freed if its last mapcount is gone. So putting the page back to lru sounds not that productive in this case since the page might be isolated by vmscan but it can't be reclaimed by vmscan since it can't be unmapped by try_to_unmap() at all. Actually if khugepaged is the last user of this page so it can be freed directly. So, clearing active and unevictable flags, unlocking and dropping refcount from isolate instead of calling putback_lru_page(). Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Hugh Dickins <hughd@google.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com> --- v2: Check mapcount and skip putback lru if the last mapcount is gone mm/khugepaged.c | 20 ++++++++++++++------ 1 file changed, 14 insertions(+), 6 deletions(-)