gfs2: buffered write prefaulting

In gfs2_file_buffered_write, to increase the likelihood that all the
user memory we're trying to write will be resident in memory, carry out
the write in chunks and fault in each chunk of user memory before trying
to write it.  Otherwise, some workloads will trigger frequent short
"internal" writes, causing filesystem blocks to be allocated and then
partially deallocated again when writing into holes, which is wasteful
and breaks reservations.

Neither the chunked writes nor any of the short "internal" writes are
user visible.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
diff --git a/fs/gfs2/file.c b/fs/gfs2/file.c
index 11c4640..5eda1bc 100644
--- a/fs/gfs2/file.c
+++ b/fs/gfs2/file.c
@@ -778,7 +778,7 @@
 	size_t count = iov_iter_count(i);
 	size_t size, offs;
 
-	if (likely(!count))
+	if (!count)
 		return false;
 	if (!iter_is_iovec(i))
 		return false;
@@ -1033,7 +1033,20 @@
 	ret = gfs2_glock_nq(gh);
 	if (ret)
 		goto out_uninit;
+	if (should_fault_in_pages(from, iocb, &prev_count, &window_size)) {
 retry_under_glock:
+		gfs2_holder_allow_demote(gh);
+		window_size -= fault_in_iov_iter_readable(from, window_size);
+		gfs2_holder_disallow_demote(gh);
+		if (!window_size) {
+			ret = -EFAULT;
+			goto out_unlock;
+		}
+		if (!gfs2_holder_queued(gh))
+			goto retry;
+		from->count = min(from->count, window_size);
+	}
+
 	if (inode == sdp->sd_rindex) {
 		struct gfs2_inode *m_ip = GFS2_I(sdp->sd_statfs_inode);
 
@@ -1060,17 +1073,8 @@
 		goto out_unlock;
 
 	from->count = orig_count - written;
-	if (should_fault_in_pages(from, iocb, &prev_count, &window_size)) {
-		gfs2_holder_allow_demote(gh);
-		window_size -= fault_in_iov_iter_readable(from, window_size);
-		gfs2_holder_disallow_demote(gh);
-		if (window_size) {
-			from->count = min(from->count, window_size);
-			if (gfs2_holder_queued(gh))
-				goto retry_under_glock;
-			goto retry;
-		}
-	}
+	if (should_fault_in_pages(from, iocb, &prev_count, &window_size))
+		goto retry_under_glock;
 out_unlock:
 	if (gfs2_holder_queued(gh))
 		gfs2_glock_dq(gh);