[PATCH v12 04/22] mm: devmap: refactor 1-based refcounting for ZONE_DEVICE pages

[John Hubbard]

On 1/15/20 7:23 AM, Christoph Hellwig wrote:
...
> 
> I'm really not sold on this scheme.  Note that I think it is
> particularly bad, but it also doesn't seem any better than what
> we had before, and it introduced quite a bit more code.
> 

Hi Christoph,

All by itself, yes. But the very next patch (which needs a little 
rework for other reasons, so not included here) needs to reuse some of 
these functions within __unpin_devmap_managed_user_page():

    page_is_devmap_managed()
    free_devmap_managed_page()

That patch was posted as part of the v11 series [1], and it did this:

+#ifdef CONFIG_DEV_PAGEMAP_OPS
+static bool __unpin_devmap_managed_user_page(struct page *page)
+{
+	int count;
+
+	if (!page_is_devmap_managed(page))
+		return false;
+
+	count = page_ref_sub_return(page, GUP_PIN_COUNTING_BIAS);
+
+	__update_proc_vmstat(page, NR_FOLL_PIN_RETURNED, 1);
+	/*
+	 * devmap page refcounts are 1-based, rather than 0-based: if
+	 * refcount is 1, then the page is free and the refcount is
+	 * stable because nobody holds a reference on the page.
+	 */
+	if (count == 1)
+		free_devmap_managed_page(page);
+	else if (!count)
+		__put_page(page);
+
+	return true;
+}
+#else
+static bool __unpin_devmap_managed_user_page(struct page *page)
+{
+	return false;
+}
+#endif /* CONFIG_DEV_PAGEMAP_OPS */
+
+/**
+ * unpin_user_page() - release a dma-pinned page
+ * @page:            pointer to page to be released
+ *
+ * Pages that were pinned via pin_user_pages*() must be released via either
+ * unpin_user_page(), or one of the unpin_user_pages*() routines. This is so
+ * that such pages can be separately tracked and uniquely handled. In
+ * particular, interactions with RDMA and filesystems need special handling.
+ */
+void unpin_user_page(struct page *page)
+{
+	page = compound_head(page);
+
+	/*
+	 * For devmap managed pages we need to catch refcount transition from
+	 * GUP_PIN_COUNTING_BIAS to 1, when refcount reach one it means the
+	 * page is free and we need to inform the device driver through
+	 * callback. See include/linux/memremap.h and HMM for details.
+	 */
+	if (__unpin_devmap_managed_user_page(page))
+		return;
+
+	if (page_ref_sub_and_test(page, GUP_PIN_COUNTING_BIAS))
+		__put_page(page);
+
+	__update_proc_vmstat(page, NR_FOLL_PIN_RETURNED, 1);
+}
+EXPORT_SYMBOL(unpin_user_page);


[1] https://lore.kernel.org/r/20191216222537.491123-24-jhubbard@nvidia.com  
    [PATCH v11 23/25] mm/gup: track FOLL_PIN pages

thanks,
-- 
John Hubbard
NVIDIA