From: Joao Martins <joao.m.martins@xxxxxxxxxx> Subject: mm/gup: decrement head page once for group of subpages Rather than decrementing the head page refcount one by one, we walk the page array and checking which belong to the same compound_head. Later on we decrement the calculated amount of references in a single write to the head page. To that end switch to for_each_compound_head() does most of the work. set_page_dirty() needs no adjustment as it's a nop for non-dirty head pages and it doesn't operate on tail pages. This considerably improves unpinning of pages with THP and hugetlbfs: - THP gup_test -t -m 16384 -r 10 [-L|-a] -S -n 512 -w PIN_LONGTERM_BENCHMARK (put values): ~87.6k us -> ~23.2k us - 16G with 1G huge page size gup_test -f /mnt/huge/file -m 16384 -r 10 [-L|-a] -S -n 512 -w PIN_LONGTERM_BENCHMARK: (put values): ~87.6k us -> ~27.5k us Link: https://lkml.kernel.org/r/20210212130843.13865-3-joao.m.martins@xxxxxxxxxx Signed-off-by: Joao Martins <joao.m.martins@xxxxxxxxxx> Reviewed-by: John Hubbard <jhubbard@xxxxxxxxxx> Reviewed-by: Jason Gunthorpe <jgg@xxxxxxxxxx> Cc: Christoph Hellwig <hch@xxxxxxxxxxxxx> Cc: Doug Ledford <dledford@xxxxxxxxxx> Cc: Matthew Wilcox <willy@xxxxxxxxxxxxx> Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> --- mm/gup.c | 29 +++++++++++------------------ 1 file changed, 11 insertions(+), 18 deletions(-) --- a/mm/gup.c~mm-gup-decrement-head-page-once-for-group-of-subpages +++ a/mm/gup.c @@ -265,20 +265,15 @@ void unpin_user_pages_dirty_lock(struct bool make_dirty) { unsigned long index; - - /* - * TODO: this can be optimized for huge pages: if a series of pages is - * physically contiguous and part of the same compound page, then a - * single operation to the head page should suffice. - */ + struct page *head; + unsigned int ntails; if (!make_dirty) { unpin_user_pages(pages, npages); return; } - for (index = 0; index < npages; index++) { - struct page *page = compound_head(pages[index]); + for_each_compound_head(index, pages, npages, head, ntails) { /* * Checking PageDirty at this point may race with * clear_page_dirty_for_io(), but that's OK. Two key @@ -299,9 +294,9 @@ void unpin_user_pages_dirty_lock(struct * written back, so it gets written back again in the * next writeback cycle. This is harmless. */ - if (!PageDirty(page)) - set_page_dirty_lock(page); - unpin_user_page(page); + if (!PageDirty(head)) + set_page_dirty_lock(head); + put_compound_head(head, ntails, FOLL_PIN); } } EXPORT_SYMBOL(unpin_user_pages_dirty_lock); @@ -318,6 +313,8 @@ EXPORT_SYMBOL(unpin_user_pages_dirty_loc void unpin_user_pages(struct page **pages, unsigned long npages) { unsigned long index; + struct page *head; + unsigned int ntails; /* * If this WARN_ON() fires, then the system *might* be leaking pages (by @@ -326,13 +323,9 @@ void unpin_user_pages(struct page **page */ if (WARN_ON(IS_ERR_VALUE(npages))) return; - /* - * TODO: this can be optimized for huge pages: if a series of pages is - * physically contiguous and part of the same compound page, then a - * single operation to the head page should suffice. - */ - for (index = 0; index < npages; index++) - unpin_user_page(pages[index]); + + for_each_compound_head(index, pages, npages, head, ntails) + put_compound_head(head, ntails, FOLL_PIN); } EXPORT_SYMBOL(unpin_user_pages); _