On 2/3/20 5:23 AM, Kirill A. Shutemov wrote:

On Fri, Jan 31, 2020 at 07:40:21PM -0800, John Hubbard wrote:

quoted

An upcoming patch requires subtracting a large chunk of refcounts from
a page, and checking what the resulting refcount is. This is a little
different than the usual "check for zero refcount" that many of the
page ref functions already do. However, it is similar to a few other
routines that (like this one) are generally useful for things such as
1-based refcounting.

Add page_ref_sub_return(), that subtracts a chunk of refcounts
atomically, and returns an atomic snapshot of the result.

Signed-off-by: John Hubbard <jhubbard@nvidia.com>
---
 include/linux/page_ref.h | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/include/linux/page_ref.h b/include/linux/page_ref.h
index 14d14beb1f7f..b9cbe553d1e7 100644
--- a/include/linux/page_ref.h
+++ b/include/linux/page_ref.h

@@ -102,6 +102,16 @@ static inline void page_ref_sub(struct page *page, int nr)
 		__page_ref_mod(page, -nr);
 }
 
+static inline int page_ref_sub_return(struct page *page, int nr)
+{
+	int ret = atomic_sub_return(nr, &page->_refcount);
+
+	if (page_ref_tracepoint_active(__tracepoint_page_ref_mod))
+		__page_ref_mod(page, -nr);

Shouldn't it be __page_ref_mod_and_return() and relevant tracepoint?

Why yes, it should. I didn't even notice that that more precise function existed,
thanks for catching that. I've changed it to this for the next version of the
patchset:

static inline int page_ref_sub_return(struct page *page, int nr)
{
	int ret = atomic_sub_return(nr, &page->_refcount);

	if (page_ref_tracepoint_active(__tracepoint_page_ref_mod))
		__page_ref_mod_and_return(page, -nr, ret);
	return ret;
}



thanks,
-- 
John Hubbard
NVIDIA

quoted

+
+	return ret;
+}
+
 static inline void page_ref_inc(struct page *page)
 {
 	atomic_inc(&page->_refcount);
-- 
2.25.0