@@ -2020,7 +2020,7 @@ ok:
spin_lock(&mle->spinlock);
if (mle->type == DLM_MLE_BLOCK || mle->type == DLM_MLE_MIGRATION)
- extra_ref = 1;
+ extra_ref = test_bit(assert->node_idx, mle->maybe_map) ? 1 : 0;
else {
/* MASTER mle: if any bits set in the response map
* then the calling node needs to re-assert to clear
@@ -3465,12 +3465,18 @@ static void dlm_clean_block_mle(struct dlm_ctxt *dlm,
mlog(0, "mle found, but dead node %u would not have been "
"master\n", dead_node);
spin_unlock(&mle->spinlock);
+ } else if(mle->master != O2NM_MAX_NODES){
+ mlog(ML_NOTICE, "mle found, master assert received, master has "
+ "already set to %d.\n ", mle->master);
+ spin_unlock(&mle->spinlock);
} else {
/* Must drop the refcount by one since the assert_master will
* never arrive. This may result in the mle being unlinked and
* freed, but there may still be a process waiting in the
* dlmlock path which is fine. */
mlog(0, "node %u was expected master\n", dead_node);
+ clear_bit(bit, mle->maybe_map);
atomic_set(&mle->woken, 1);
spin_unlock(&mle->spinlock);
wake_up(&mle->wq);
--
Hi, During my test on OCFS2 suffering a storage failure, a crash issue was found. Below was the call trace when crashed. From the call trace, we can see a MLE's reference count is going to be negative, which aroused a BUG_ON() [143355.593258] Call Trace: [143355.593268] [<ffffffffc0328447>] dlm_put_mle_inuse+0x47/0x70 [ocfs2_dlm] [143355.593276] [<ffffffffc032bee5>] dlm_get_lock_resource+0xac5/0x10d0 [ocfs2_dlm] [143355.593286] [<ffffffff81724a7a>] ? ip_queue_xmit+0x14a/0x3d0 [143355.593292] [<ffffffff811e50b4>] ? kmem_cache_alloc+0x1e4/0x220 [143355.593300] [<ffffffffc03215cc>] ? dlm_wait_for_recovery+0x6c/0x190 [ocfs2_dlm] [143355.593311] [<ffffffffc0335c4d>] dlmlock+0x62d/0x16e0 [ocfs2_dlm] [143355.593316] [<ffffffff816cfbab>] ? __alloc_skb+0x9b/0x2b0 [143355.593323] [<ffffffffc01f6000>] ? 0xffffffffc01f6000 I think I probably have found the root cause of this issue. Please **Node 1** **Node 2** Storage failure An assert master message is sent to Node 1 Treat Node2 as down Assert master handler Decrease MLE reference count Clean blocked MLE Decrease MLE reference count In the above scenario, both dlm_assert_master_handler and dlm_clean_block_mle will decease MLE reference count, thus, in the following get_resouce procedure, the reference count is going to be negative. I propose a patch to solve this, please take review if you have any time. Signed-off-by: gechangwei <ge.changwei@h3c.com> --- dlm/dlmmaster.c | 8 +++++++- 1 file changed, 6 insertions(+), 1 deletion(-) BR. Changwei ------------------------------------------------------------------------------------------------------------------------------------- 本邮件及其附件含有杭州华三通信技术有限公司的保密信息，仅限于发送给上面地址中列出 的个人或群组。禁止任何其他人以任何形式使用（包括但不限于全部或部分地泄露、复制、 或散发）本邮件中的信息。如果您错收了本邮件，请您立即电话或邮件通知发件人并删除本 邮件！ This e-mail and its attachments contain confidential information from H3C, which is intended only for the person or entity whose address is listed above. Any use of the information contained herein in any way (including, but not limited to, total or partial disclosure, reproduction, or dissemination) by persons other than the intended recipient(s) is prohibited. If you receive this e-mail in error, please notify the sender by phone or email immediately and delete it!