Re: 2.6.16-rc6-mm2

[Christoph Lameter <clameter@sgi.com>]

On Tue, 13 Jun 2006, Christoph Lameter wrote:

> On Tue, 13 Jun 2006, Cedric Le Goater wrote:
> 
> > thanks for the patch ! I gave it a try but req->wb_page seems bogus ?
> 
> It seems that req->wb_page is bogus after nfs_clear_page_writeback()
> has run. So we need to do the statistics before.

But then we miss the decrement of NR_UNSTABLE when a page is freed before 
we get to that function. NR_UNSTABLE grows beyond bounds and the write 
throttling stalls processes.

So we need an additional dec_zone_state..

Index: linux-2.6.17-rc6-mm2/fs/nfs/write.c
===================================================================
--- linux-2.6.17-rc6-mm2.orig/fs/nfs/write.c	2006-06-10 11:11:53.051397816 -0700
+++ linux-2.6.17-rc6-mm2/fs/nfs/write.c	2006-06-13 10:52:04.428456013 -0700
@@ -1418,8 +1418,9 @@ static void nfs_commit_done(struct rpc_t
 		dprintk(" mismatch\n");
 		nfs_mark_request_dirty(req);
 	next:
+		if (req->wb_page)
+			dec_zone_page_state(req->wb_page, NR_UNSTABLE);
 		nfs_clear_page_writeback(req);
-		dec_zone_page_state(req->wb_page, NR_UNSTABLE);
 	}
 }
 
Index: linux-2.6.17-rc6-mm2/fs/nfs/pagelist.c
===================================================================
--- linux-2.6.17-rc6-mm2.orig/fs/nfs/pagelist.c	2006-06-10 11:11:53.049444812 -0700
+++ linux-2.6.17-rc6-mm2/fs/nfs/pagelist.c	2006-06-13 12:33:39.545259204 -0700
@@ -154,6 +154,7 @@ void nfs_clear_request(struct nfs_page *
 {
 	struct page *page = req->wb_page;
 	if (page != NULL) {
+		dec_zone_page_state(page, NR_UNSTABLE);
 		page_cache_release(page);
 		req->wb_page = NULL;
 	}
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/