On 11/19/20 12:45 AM, Zhao Heming wrote:

md-cluster uses MD_CLUSTER_SEND_LOCK to make node can exclusively send msg.
During sending msg, node can concurrently receive msg from another node.
[... ...]

Repro steps (I only triggered 3 times with hundreds tests):

sorry to send v4 patch so late.
I spent more than 2 days to run test script to trigger the deadlock.
The result is I failed.
So I wrote "I only triggered 3 times with hundreds tests" in commit log.

two nodes share 3 iSCSI luns: sdg/sdh/sdi. Each lun size is 1GB.

ssh root@node2 "mdadm -S --scan"
[... ...]

At last, thanks for Xiao's solution.

Signed-off-by: Zhao Heming <heming.zhao@suse.com>
Suggested-by: Xiao Ni <xni@redhat.com>
Reviewed-by: Xiao Ni <xni@redhat.com>
---
  drivers/md/md-cluster.c | 69 +++++++++++++++++++++++------------------
  drivers/md/md.c         |  6 ++--
  2 files changed, 43 insertions(+), 32 deletions(-)