@@ -1601,7 +1601,15 @@ static void o2net_start_connect(struct work_struct *work)
struct sockaddr_in myaddr = {0, }, remoteaddr = {0, };
int ret = 0, stop;
unsigned int timeout;
+ unsigned int noio_flag;
+ /*
+ * sock_create allocates the sock with GFP_KERNEL. We must set
+ * per-process flag PF_MEMALLOC_NOIO so that all allocations done
+ * by this process are done as if GFP_NOIO was specified. So we
+ * are not reentering filesystem while doing memory reclaim.
+ */
+ noio_flag = memalloc_noio_save();
/* if we're greater we initiate tx, otherwise we accept */
if (o2nm_this_node() <= o2net_num_from_nn(nn))
goto out;
@@ -1710,6 +1718,7 @@ out:
if (mynode)
o2nm_node_put(mynode);
+ memalloc_noio_restore(noio_flag);
return;
}
@@ -1835,6 +1844,15 @@ static int o2net_accept_one(struct socket *sock, int *more)
struct o2nm_node *local_node = NULL;
struct o2net_sock_container *sc = NULL;
struct o2net_node *nn;
+ unsigned int noio_flag;
+
+ /*
+ * sock_create_lite allocates the sock with GFP_KERNEL. We must set
+ * per-process flag PF_MEMALLOC_NOIO so that all allocations done
+ * by this process are done as if GFP_NOIO was specified. So we
+ * are not reentering filesystem while doing memory reclaim.
+ */
+ noio_flag = memalloc_noio_save();
BUG_ON(sock == NULL);
*more = 0;
@@ -1951,6 +1969,8 @@ out:
o2nm_node_put(local_node);
if (sc)
sc_put(sc);
+
+ memalloc_noio_restore(noio_flag);
return ret;
}
The patch fix a deadlock problem caused by direct memory reclaim in o2net_wq. The situation is as follows: 1) Receiving a connect message from another node, node queues a work_struct o2net_listen_work. 2) o2net_wq processes this work and call the following functions: o2net_wq -> o2net_accept_one -> sock_create_lite -> sock_alloc() -> kmem_cache_alloc with GFP_KERNEL -> ____cache_alloc_node ->__alloc_pages_nodemask -> do_try_to_free_pages -> shrink_slab -> evict -> ocfs2_evict_inode -> ocfs2_drop_lock -> dlmunlock -> o2net_send_message_vec then o2net_wq wait for the unlock reply from master. 3)tcp layer received the reply, call o2net_data_ready() and queue sc_rx_work, waiting o2net_wq to process this work. 4)o2net_wq is a single thread workqueue, it process the work one by one. Right now it is still doing o2net_listen_work and cannot handle sc_rx_work. so we deadlock. Junxiao Bi's patch (http://ozlabs.org/~akpm/mmots/broken-out/mm-clear-__gfp_fs-when-pf_memalloc_noio-is-set.patch) clearing __GFP_FS in memalloc_noio_flags() besides __GFP_IO. We use memalloc_noio_save() to set process flag PF_MEMALLOC_NOIO so that all allocations done by this process are done as if GFP_NOIO was specified. We are not reentering filesystem while doing memory reclaim. Signed-off-by: joyce.xue <xuejiufei@huawei.com> --- fs/ocfs2/cluster/tcp.c | 20 ++++++++++++++++++++ 1 file changed, 20 insertions(+)