[Drbd-dev] [PATCH] drbd: fix regression 'out of mem, failed to invoke fence-peer helper'

[Philipp Reisner]

From: Lars Ellenberg <lars.ellenberg at linbit.com>

Stable info:
  This patch landed in upstream with v3.16 as commit
  bbc1c5e8ad6dfebf9d13b8a4ccdf66c92913eac9
  it should go into v3.14+

Since linux kernel 3.13, kthread_run() internally uses
wait_for_completion_killable().  We sometimes may use kthread_run()
while we still have a signal pending, which we used to kick our threads
out of potentially blocking network functions, causing kthread_run() to
mistake that as a new fatal signal and fail.

Fix: flush_signals() before kthread_run().

Signed-off-by: Philipp Reisner <philipp.reisner at linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg at linbit.com>
Signed-off-by: Jens Axboe <axboe at fb.com>
---
 drivers/block/drbd/drbd_nl.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/drivers/block/drbd/drbd_nl.c b/drivers/block/drbd/drbd_nl.c
index 1b35c45..3f2e167 100644
--- a/drivers/block/drbd/drbd_nl.c
+++ b/drivers/block/drbd/drbd_nl.c
@@ -544,6 +544,12 @@ void conn_try_outdate_peer_async(struct drbd_connection *connection)
        struct task_struct *opa;

        kref_get(&connection->kref);
+       /* We may just have force_sig()'ed this thread
+        * to get it out of some blocking network function.
+        * Clear signals; otherwise kthread_run(), which internally uses
+        * wait_on_completion_killable(), will mistake our pending signal
+        * for a new fatal signal and fail. */
+       flush_signals(current);
        opa = kthread_run(_try_outdate_peer_async, connection, "drbd_async_h");
        if (IS_ERR(opa)) {
                drbd_err(connection, "out of mem, failed to invoke fence-peer helper\n");
-- 
1.9.1