Commit 2b832564 authored by Sunil Mushran's avatar Sunil Mushran Committed by Mark Fasheh

ocfs2/dlm: Fix a race between migrate request and exit domain

Patch address a racing migrate request message and an exit domain message.
Instead of blocking exit domains for the duration of the migrate, we ignore
failure to deliver that message. This is because an exiting domain should
not have any active locks and thus has no role to play in the migration.
Signed-off-by: default avatarSunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: default avatarMark Fasheh <mfasheh@suse.com>
parent 58896c4d
...@@ -2949,7 +2949,7 @@ static int dlm_do_migrate_request(struct dlm_ctxt *dlm, ...@@ -2949,7 +2949,7 @@ static int dlm_do_migrate_request(struct dlm_ctxt *dlm,
struct dlm_node_iter *iter) struct dlm_node_iter *iter)
{ {
struct dlm_migrate_request migrate; struct dlm_migrate_request migrate;
int ret, status = 0; int ret, skip, status = 0;
int nodenum; int nodenum;
memset(&migrate, 0, sizeof(migrate)); memset(&migrate, 0, sizeof(migrate));
...@@ -2966,12 +2966,27 @@ static int dlm_do_migrate_request(struct dlm_ctxt *dlm, ...@@ -2966,12 +2966,27 @@ static int dlm_do_migrate_request(struct dlm_ctxt *dlm,
nodenum == new_master) nodenum == new_master)
continue; continue;
/* We could race exit domain. If exited, skip. */
spin_lock(&dlm->spinlock);
skip = (!test_bit(nodenum, dlm->domain_map));
spin_unlock(&dlm->spinlock);
if (skip) {
clear_bit(nodenum, iter->node_map);
continue;
}
ret = o2net_send_message(DLM_MIGRATE_REQUEST_MSG, dlm->key, ret = o2net_send_message(DLM_MIGRATE_REQUEST_MSG, dlm->key,
&migrate, sizeof(migrate), nodenum, &migrate, sizeof(migrate), nodenum,
&status); &status);
if (ret < 0) if (ret < 0) {
mlog_errno(ret); mlog(0, "migrate_request returned %d!\n", ret);
else if (status < 0) { if (!dlm_is_host_down(ret)) {
mlog(ML_ERROR, "unhandled error=%d!\n", ret);
BUG();
}
clear_bit(nodenum, iter->node_map);
ret = 0;
} else if (status < 0) {
mlog(0, "migrate request (node %u) returned %d!\n", mlog(0, "migrate request (node %u) returned %d!\n",
nodenum, status); nodenum, status);
ret = status; ret = status;
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment