From: "J. Bruce Fields" <bfields@xxxxxxxxxx>
Date: Tue, 29 Oct 2019 16:02:18 -0400
The callback code relies on the fact that much of it is only ever called
from the ordered workqueue callback_wq, and this is worth documenting.
Reported-by: Trond Myklebust <trond.myklebust@xxxxxxxxxxxxxxx>
Signed-off-by: J. Bruce Fields <bfields@xxxxxxxxxx>
---
fs/nfsd/nfs4callback.c | 6 ++++++
1 file changed, 6 insertions(+)
Also adding a comment, since I know this was a source of confusion when
investigating these races.
diff --git a/fs/nfsd/nfs4callback.c b/fs/nfsd/nfs4callback.c
index c94768b096a3..24534db87e86 100644
--- a/fs/nfsd/nfs4callback.c
+++ b/fs/nfsd/nfs4callback.c
@@ -1243,6 +1243,12 @@ static struct nfsd4_conn * __nfsd4_find_backchannel(struct nfs4_client *clp)
return NULL;
}
+/*
+ * Note there isn't a lot of locking in this code; instead we depend on
+ * the fact that it is run from the callback_wq, which won't run two
+ * work items at once. So, for example, callback_wq handles all access
+ * of cl_cb_client and all calls to rpc_create or rpc_shutdown_client.
+ */
static void nfsd4_process_cb_update(struct nfsd4_callback *cb)
{
struct nfs4_cb_conn conn;
--
2.23.0
