Merge tag 'dlm-4.15' of git://git.kernel.org/pub/scm/linux/kernel/git/teigland/linux-dlm

Pull dlm updates from David Teigland:
 "This set focuses, as usual, on fixes to the comms layer.

  New testing of the dlm with ocfs2 uncovered a number of bugs in the
  TCP connection handling during recovery, starting, and stopping"

* tag 'dlm-4.15' of git://git.kernel.org/pub/scm/linux/kernel/git/teigland/linux-dlm:
  dlm: remove dlm_send_rcom_lookup_dump
  dlm: recheck kthread_should_stop() before schedule()
  DLM: fix NULL pointer dereference in send_to_sock()
  DLM: fix to reschedule rwork
  DLM: fix to use sk_callback_lock correctly
  DLM: fix overflow dlm_cb_seq
  DLM: fix memory leak in tcp_accept_from_sock()
  DLM: fix conversion deadlock when DLM_LKF_NODLCKWT flag is set
  DLM: use CF_CLOSE flag to stop dlm_send correctly
  DLM: Reanimate CF_WRITE_PENDING flag
  DLM: fix race condition between dlm_recoverd_stop and dlm_recoverd
  DLM: close othercon at send/receive error
  DLM: retry rcom when dlm_wait_function is timed out.
  DLM: fix to use sock_mutex correctly in xxx_accept_from_sock
  DLM: fix race condition between dlm_send and dlm_recv
  DLM: fix double list_del()
  DLM: fix remove save_cb argument from add_sock()
  DLM: Fix saving of NULL callbacks
  DLM: Eliminate CF_WRITE_PENDING flag
  DLM: Eliminate CF_CONNECT_PENDING flag

fs/dlm/ast.c

@@ -181,6 +181,8 @@ void dlm_add_cb(struct dlm_lkb *lkb, uint32_t flags, int mode, int status,
 
 	spin_lock(&dlm_cb_seq_spin);
 	new_seq = ++dlm_cb_seq;
+	if (!dlm_cb_seq)
+		new_seq = ++dlm_cb_seq;
 	spin_unlock(&dlm_cb_seq_spin);
 
 	if (lkb->lkb_flags & DLM_IFL_USER) {

fs/dlm/lock.c

@@ -1003,7 +1003,6 @@ int dlm_master_lookup(struct dlm_ls *ls, int from_nodeid, char *name, int len,
 		if (r->res_master_nodeid == our_nodeid) {
 			log_error(ls, "from_master %d our_master", from_nodeid);
 			dlm_dump_rsb(r);
-			dlm_send_rcom_lookup_dump(r, from_nodeid);
 			goto out_found;
 		}
 
@@ -2465,14 +2464,12 @@ static int can_be_granted(struct dlm_rsb *r, struct dlm_lkb *lkb, int now,
 		if (lkb->lkb_exflags & DLM_LKF_CONVDEADLK) {
 			lkb->lkb_grmode = DLM_LOCK_NL;
 			lkb->lkb_sbflags |= DLM_SBF_DEMOTED;
-		} else if (!(lkb->lkb_exflags & DLM_LKF_NODLCKWT)) {
-			if (err)
-				*err = -EDEADLK;
-			else {
-				log_print("can_be_granted deadlock %x now %d",
-					  lkb->lkb_id, now);
-				dlm_dump_rsb(r);
-			}
+		} else if (err) {
+			*err = -EDEADLK;
+		} else {
+			log_print("can_be_granted deadlock %x now %d",
+				  lkb->lkb_id, now);
+			dlm_dump_rsb(r);
 		}
 		goto out;
 	}
@@ -2501,13 +2498,6 @@ static int can_be_granted(struct dlm_rsb *r, struct dlm_lkb *lkb, int now,
 	return rv;
 }
 
-/* FIXME: I don't think that can_be_granted() can/will demote or find deadlock
-   for locks pending on the convert list.  Once verified (watch for these
-   log_prints), we should be able to just call _can_be_granted() and not
-   bother with the demote/deadlk cases here (and there's no easy way to deal
-   with a deadlk here, we'd have to generate something like grant_lock with
-   the deadlk error.) */
-
 /* Returns the highest requested mode of all blocked conversions; sets
    cw if there's a blocked conversion to DLM_LOCK_CW. */
 
@@ -2545,9 +2535,22 @@ static int grant_pending_convert(struct dlm_rsb *r, int high, int *cw,
 		}
 
 		if (deadlk) {
-			log_print("WARN: pending deadlock %x node %d %s",
-				  lkb->lkb_id, lkb->lkb_nodeid, r->res_name);
-			dlm_dump_rsb(r);
+			/*
+			 * If DLM_LKB_NODLKWT flag is set and conversion
+			 * deadlock is detected, we request blocking AST and
+			 * down (or cancel) conversion.
+			 */
+			if (lkb->lkb_exflags & DLM_LKF_NODLCKWT) {
+				if (lkb->lkb_highbast < lkb->lkb_rqmode) {
+					queue_bast(r, lkb, lkb->lkb_rqmode);
+					lkb->lkb_highbast = lkb->lkb_rqmode;
+				}
+			} else {
+				log_print("WARN: pending deadlock %x node %d %s",
+					  lkb->lkb_id, lkb->lkb_nodeid,
+					  r->res_name);
+				dlm_dump_rsb(r);
+			}
 			continue;
 		}
 
@@ -3123,7 +3126,7 @@ static int do_convert(struct dlm_rsb *r, struct dlm_lkb *lkb)
 	   deadlock, so we leave it on the granted queue and return EDEADLK in
 	   the ast for the convert. */
 
-	if (deadlk) {
+	if (deadlk && !(lkb->lkb_exflags & DLM_LKF_NODLCKWT)) {
 		/* it's left on the granted queue */
 		revert_lock(r, lkb);
 		queue_cast(r, lkb, -EDEADLK);

fs/dlm/rcom.c

@@ -155,6 +155,7 @@ int dlm_rcom_status(struct dlm_ls *ls, int nodeid, uint32_t status_flags)
 		goto out;
 	}
 
+retry:
 	error = create_rcom(ls, nodeid, DLM_RCOM_STATUS,
 			    sizeof(struct rcom_status), &rc, &mh);
 	if (error)
@@ -169,6 +170,8 @@ int dlm_rcom_status(struct dlm_ls *ls, int nodeid, uint32_t status_flags)
 
 	error = dlm_wait_function(ls, &rcom_response);
 	disallow_sync_reply(ls);
+	if (error == -ETIMEDOUT)
+		goto retry;
 	if (error)
 		goto out;
 
@@ -276,6 +279,7 @@ int dlm_rcom_names(struct dlm_ls *ls, int nodeid, char *last_name, int last_len)
 
 	ls->ls_recover_nodeid = nodeid;
 
+retry:
 	error = create_rcom(ls, nodeid, DLM_RCOM_NAMES, last_len, &rc, &mh);
 	if (error)
 		goto out;
@@ -288,6 +292,8 @@ int dlm_rcom_names(struct dlm_ls *ls, int nodeid, char *last_name, int last_len)
 
 	error = dlm_wait_function(ls, &rcom_response);
 	disallow_sync_reply(ls);
+	if (error == -ETIMEDOUT)
+		goto retry;
  out:
 	return error;
 }
@@ -332,25 +338,6 @@ int dlm_send_rcom_lookup(struct dlm_rsb *r, int dir_nodeid)
 	return error;
 }
 
-int dlm_send_rcom_lookup_dump(struct dlm_rsb *r, int to_nodeid)
-{
-	struct dlm_rcom *rc;
-	struct dlm_mhandle *mh;
-	struct dlm_ls *ls = r->res_ls;
-	int error;
-
-	error = create_rcom(ls, to_nodeid, DLM_RCOM_LOOKUP, r->res_length,
-			    &rc, &mh);
-	if (error)
-		goto out;
-	memcpy(rc->rc_buf, r->res_name, r->res_length);
-	rc->rc_id = 0xFFFFFFFF;
-
-	send_rcom(ls, mh, rc);
- out:
-	return error;
-}
-
 static void receive_rcom_lookup(struct dlm_ls *ls, struct dlm_rcom *rc_in)
 {
 	struct dlm_rcom *rc;
@@ -362,6 +349,7 @@ static void receive_rcom_lookup(struct dlm_ls *ls, struct dlm_rcom *rc_in)
 	if (error)
 		return;
 
+	/* Old code would send this special id to trigger a debug dump. */
 	if (rc_in->rc_id == 0xFFFFFFFF) {
 		log_error(ls, "receive_rcom_lookup dump from %d", nodeid);
 		dlm_dump_rsb_name(ls, rc_in->rc_buf, len);

fs/dlm/rcom.h

@@ -17,7 +17,6 @@
 int dlm_rcom_status(struct dlm_ls *ls, int nodeid, uint32_t status_flags);
 int dlm_rcom_names(struct dlm_ls *ls, int nodeid, char *last_name,int last_len);
 int dlm_send_rcom_lookup(struct dlm_rsb *r, int dir_nodeid);
-int dlm_send_rcom_lookup_dump(struct dlm_rsb *r, int to_nodeid);
 int dlm_send_rcom_lock(struct dlm_rsb *r, struct dlm_lkb *lkb);
 void dlm_receive_rcom(struct dlm_ls *ls, struct dlm_rcom *rc, int nodeid);
 int dlm_send_ls_not_ready(int nodeid, struct dlm_rcom *rc_in);

fs/dlm/recover.c

@@ -52,6 +52,10 @@ int dlm_wait_function(struct dlm_ls *ls, int (*testfn) (struct dlm_ls *ls))
 					dlm_config.ci_recover_timer * HZ);
 		if (rv)
 			break;
+		if (test_bit(LSFL_RCOM_WAIT, &ls->ls_flags)) {
+			log_debug(ls, "dlm_wait_function timed out");
+			return -ETIMEDOUT;
+		}
 	}
 
 	if (dlm_recovery_stopped(ls)) {

fs/dlm/recoverd.c

@@ -287,11 +287,23 @@ static int dlm_recoverd(void *arg)
 	set_bit(LSFL_RECOVER_LOCK, &ls->ls_flags);
 	wake_up(&ls->ls_recover_lock_wait);
 
-	while (!kthread_should_stop()) {
+	while (1) {
+		/*
+		 * We call kthread_should_stop() after set_current_state().
+		 * This is because it works correctly if kthread_stop() is
+		 * called just before set_current_state().
+		 */
 		set_current_state(TASK_INTERRUPTIBLE);
+		if (kthread_should_stop()) {
+			set_current_state(TASK_RUNNING);
+			break;
+		}
 		if (!test_bit(LSFL_RECOVER_WORK, &ls->ls_flags) &&
-		    !test_bit(LSFL_RECOVER_DOWN, &ls->ls_flags))
+		    !test_bit(LSFL_RECOVER_DOWN, &ls->ls_flags)) {
+			if (kthread_should_stop())
+				break;
 			schedule();
+		}
 		set_current_state(TASK_RUNNING);
 
 		if (test_and_clear_bit(LSFL_RECOVER_DOWN, &ls->ls_flags)) {