Patchwork [043/270] pnfsblock: fix non-aligned DIO write

login
register
mail settings
Submitter Herton Ronaldo Krzesinski
Date Nov. 26, 2012, 4:55 p.m.
Message ID <1353949160-26803-44-git-send-email-herton.krzesinski@canonical.com>
Download mbox | patch
Permalink /patch/201776/
State New
Headers show

Comments

Herton Ronaldo Krzesinski - Nov. 26, 2012, 4:55 p.m.
3.5.7u1 -stable review patch.  If anyone has any objections, please let me know.

------------------

From: Peng Tao <bergwolf@gmail.com>

commit 96c9eae638765c2bf2ca4f5a6325484f9bb69aa7 upstream.

For DIO writes, if it is not blocksize aligned, we need to do
internal serialization. It may slow down writers anyway. So we
just bail them out and resend to MDS.

Signed-off-by: Peng Tao <tao.peng@emc.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Herton Ronaldo Krzesinski <herton.krzesinski@canonical.com>
---
 fs/nfs/blocklayout/blocklayout.c |   34 +++++++++++++++++++++++++++++++---
 1 file changed, 31 insertions(+), 3 deletions(-)

Patch

diff --git a/fs/nfs/blocklayout/blocklayout.c b/fs/nfs/blocklayout/blocklayout.c
index e5dfef5..1093968 100644
--- a/fs/nfs/blocklayout/blocklayout.c
+++ b/fs/nfs/blocklayout/blocklayout.c
@@ -685,7 +685,7 @@  bl_write_pagelist(struct nfs_write_data *wdata, int sync)
 	struct bio *bio = NULL;
 	struct pnfs_block_extent *be = NULL, *cow_read = NULL;
 	sector_t isect, last_isect = 0, extent_length = 0;
-	struct parallel_io *par;
+	struct parallel_io *par = NULL;
 	loff_t offset = wdata->args.offset;
 	size_t count = wdata->args.count;
 	unsigned int pg_offset, pg_len, saved_len;
@@ -697,6 +697,13 @@  bl_write_pagelist(struct nfs_write_data *wdata, int sync)
 	    NFS_SERVER(header->inode)->pnfs_blksize >> PAGE_CACHE_SHIFT;
 
 	dprintk("%s enter, %Zu@%lld\n", __func__, count, offset);
+
+	if (header->dreq != NULL &&
+	    (!IS_ALIGNED(offset, NFS_SERVER(header->inode)->pnfs_blksize) ||
+	     !IS_ALIGNED(count, NFS_SERVER(header->inode)->pnfs_blksize))) {
+		dprintk("pnfsblock nonblock aligned DIO writes. Resend MDS\n");
+		goto out_mds;
+	}
 	/* At this point, wdata->pages is a (sequential) list of nfs_pages.
 	 * We want to write each, and if there is an error set pnfs_error
 	 * to have it redone using nfs.
@@ -1197,6 +1204,27 @@  bl_pg_test_read(struct nfs_pageio_descriptor *pgio, struct nfs_page *prev,
 	return pnfs_generic_pg_test(pgio, prev, req);
 }
 
+void
+bl_pg_init_write(struct nfs_pageio_descriptor *pgio, struct nfs_page *req)
+{
+	if (pgio->pg_dreq != NULL &&
+	    !is_aligned_req(req, PAGE_CACHE_SIZE))
+		nfs_pageio_reset_write_mds(pgio);
+	else
+		pnfs_generic_pg_init_write(pgio, req);
+}
+
+static bool
+bl_pg_test_write(struct nfs_pageio_descriptor *pgio, struct nfs_page *prev,
+		 struct nfs_page *req)
+{
+	if (pgio->pg_dreq != NULL &&
+	    !is_aligned_req(req, PAGE_CACHE_SIZE))
+		return false;
+
+	return pnfs_generic_pg_test(pgio, prev, req);
+}
+
 static const struct nfs_pageio_ops bl_pg_read_ops = {
 	.pg_init = bl_pg_init_read,
 	.pg_test = bl_pg_test_read,
@@ -1204,8 +1232,8 @@  static const struct nfs_pageio_ops bl_pg_read_ops = {
 };
 
 static const struct nfs_pageio_ops bl_pg_write_ops = {
-	.pg_init = pnfs_generic_pg_init_write,
-	.pg_test = pnfs_generic_pg_test,
+	.pg_init = bl_pg_init_write,
+	.pg_test = bl_pg_test_write,
 	.pg_doio = pnfs_generic_pg_writepages,
 };