Re: [Pv-ops][PATCH 2/4 v2] Netback: Introduce a new struct type page_ext.

[Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>]

On Thu, Apr 29, 2010 at 10:28:36PM +0800, Xu, Dongxiao wrote:
> Netback: Introduce a new struct type page_ext.
> 
> struct page_ext is used to store the group and idx information by
> which a specified page could be identified.
> 
> Signed-off-by: Dongxiao Xu <dongxiao.xu@intel.com>

.. snip.. 


-static inline int netif_page_index(struct page *pg)
-{
-	unsigned long idx = (unsigned long)pg->mapping - 1;
-
-	if (!PageForeign(pg))
-		return -1;

You are taking the check to see if the page is foreign. Is that OK?
-
-	if ((idx >= MAX_PENDING_REQS) || (netbk->mmap_pages[idx] != pg))
-		return -1;
-
-	return idx;
+	pg->mapping = (struct address_space *)page_extinfo;;
 }
 
 /*
@@ -1411,9 +1399,10 @@ static void netif_idx_release(u16 pending_idx)
 
 static void netif_page_release(struct page *page, unsigned int order)
 {
-	int idx = netif_page_index(page);

This did this nice little check to see if the page had the PageForeign
bit set. Is it OK to remove that check?

-	BUG_ON(order);
-	BUG_ON(idx < 0);
+	int idx = ((struct page_ext *)(page->mapping))->idx;
+	BUG_ON(order ||
+	       idx < 0 || idx >= MAX_PENDING_REQS ||
+	       netbk->mmap_pages[idx] != page);
 	netif_idx_release(idx);
 }
 
@@ -1565,7 +1554,9 @@ static int __init netback_init(void)
 	for (i = 0; i < MAX_PENDING_REQS; i++) {
 		page = netbk->mmap_pages[i];
 		SetPageForeign(page, netif_page_release);
-		netif_set_page_index(page, i);
+		netbk->page_extinfo[i].group = 0;
+		netbk->page_extinfo[i].idx = i;
+		netif_set_page_index(page, &netbk->page_extinfo[i]);
 		INIT_LIST_HEAD(&netbk->pending_inuse[i].list);
 	}
 
-- 
1.6.0

> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel