NFSv4: Ensure the callback daemon flushes signals

If the callback daemon is signalled, but is unable to exit because it still
has users, then we need to flush signals. If not, then svc_recv() can
never sleep, and so we hang.
If we flush signals, then we also have to be prepared to resend them when
we want the thread to exit.

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>

1 files changed, 13 insertions, 6 deletions

diff --git a/fs/nfs/callback.c b/fs/nfs/callback.c
index 2c042f8d70b5..99d2cfbce863 100644
--- a/fs/nfs/callback.c
+++ b/fs/nfs/callback.c
@@ -55,7 +55,12 @@ static void nfs_callback_svc(struct svc_rqst *rqstp)
 
         complete(&nfs_callback_info.started);
 
-        while (nfs_callback_info.users != 0 || !signalled()) {
+        for(;;) {
+                if (signalled()) {
+                        if (nfs_callback_info.users == 0)
+                                break;
+                        flush_signals(current);
+                }
                 /*
                  * Listen for a request on the socket
                  */
@@ -135,11 +140,13 @@ int nfs_callback_down(void)
 
         lock_kernel();
         down(&nfs_callback_sema);
-        if (--nfs_callback_info.users || nfs_callback_info.pid == 0)
-                goto out;
-        kill_proc(nfs_callback_info.pid, SIGKILL, 1);
-        wait_for_completion(&nfs_callback_info.stopped);
-out:
+        nfs_callback_info.users--;
+        do {
+                if (nfs_callback_info.users != 0 || nfs_callback_info.pid == 0)
+                        break;
+                if (kill_proc(nfs_callback_info.pid, SIGKILL, 1) < 0)
+                        break;
+        } while (wait_for_completion_timeout(&nfs_callback_info.stopped, 5*HZ) == 0);
         up(&nfs_callback_sema);
         unlock_kernel();
         return ret;