[DLM] wait for config check during join [6/6]
Joining the lockspace should wait for the initial round of inter-node
config checks to complete before returning. This way, if there's a
configuration mismatch between the joining node and the existing nodes,
the join can fail and return an error to the application.
Signed-off-by: David Teigland <teigland@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
diff --git a/fs/dlm/dlm_internal.h b/fs/dlm/dlm_internal.h
index a8d6e99..03ba6c4 100644
--- a/fs/dlm/dlm_internal.h
+++ b/fs/dlm/dlm_internal.h
@@ -472,6 +472,8 @@
wait_queue_head_t ls_uevent_wait; /* user part of join/leave */
int ls_uevent_result;
+ struct completion ls_members_done;
+ int ls_members_result;
struct miscdevice ls_device;
diff --git a/fs/dlm/lockspace.c b/fs/dlm/lockspace.c
index a3a50e6..c8f0c15 100644
--- a/fs/dlm/lockspace.c
+++ b/fs/dlm/lockspace.c
@@ -197,13 +197,24 @@
else
kobject_uevent(&ls->ls_kobj, KOBJ_OFFLINE);
+ log_debug(ls, "%s the lockspace group...", in ? "joining" : "leaving");
+
+ /* dlm_controld will see the uevent, do the necessary group management
+ and then write to sysfs to wake us */
+
error = wait_event_interruptible(ls->ls_uevent_wait,
test_and_clear_bit(LSFL_UEVENT_WAIT, &ls->ls_flags));
+
+ log_debug(ls, "group event done %d %d", error, ls->ls_uevent_result);
+
if (error)
goto out;
error = ls->ls_uevent_result;
out:
+ if (error)
+ log_error(ls, "group %s failed %d %d", in ? "join" : "leave",
+ error, ls->ls_uevent_result);
return error;
}
@@ -490,6 +501,8 @@
init_waitqueue_head(&ls->ls_uevent_wait);
ls->ls_uevent_result = 0;
+ init_completion(&ls->ls_members_done);
+ ls->ls_members_result = -1;
ls->ls_recoverd_task = NULL;
mutex_init(&ls->ls_recoverd_active);
@@ -540,10 +553,21 @@
/* let kobject handle freeing of ls if there's an error */
do_unreg = 1;
+ /* This uevent triggers dlm_controld in userspace to add us to the
+ group of nodes that are members of this lockspace (managed by the
+ cluster infrastructure.) Once it's done that, it tells us who the
+ current lockspace members are (via configfs) and then tells the
+ lockspace to start running (via sysfs) in dlm_ls_start(). */
+
error = do_uevent(ls, 1);
if (error)
goto out_stop;
+ wait_for_completion(&ls->ls_members_done);
+ error = ls->ls_members_result;
+ if (error)
+ goto out_members;
+
dlm_create_debug_file(ls);
log_debug(ls, "join complete");
@@ -551,6 +575,10 @@
*lockspace = ls;
return 0;
+ out_members:
+ do_uevent(ls, 0);
+ dlm_clear_members(ls);
+ kfree(ls->ls_node_array);
out_stop:
dlm_recoverd_stop(ls);
out_delist:
@@ -588,6 +616,8 @@
error = new_lockspace(name, namelen, lockspace, flags, lvblen);
if (!error)
ls_count++;
+ else if (!ls_count)
+ threads_stop();
out:
mutex_unlock(&ls_lock);
return error;
diff --git a/fs/dlm/member.c b/fs/dlm/member.c
index f08faec..073599d 100644
--- a/fs/dlm/member.c
+++ b/fs/dlm/member.c
@@ -233,6 +233,12 @@
*neg_out = neg;
error = ping_members(ls);
+ if (!error || error == -EPROTO) {
+ /* new_lockspace() may be waiting to know if the config
+ is good or bad */
+ ls->ls_members_result = error;
+ complete(&ls->ls_members_done);
+ }
if (error)
goto out;
diff --git a/fs/dlm/rcom.c b/fs/dlm/rcom.c
index 6bfbd61..f71c235 100644
--- a/fs/dlm/rcom.c
+++ b/fs/dlm/rcom.c
@@ -90,7 +90,7 @@
log_error(ls, "version mismatch: %x nodeid %d: %x",
DLM_HEADER_MAJOR | DLM_HEADER_MINOR, nodeid,
rc->rc_header.h_version);
- return -EINVAL;
+ return -EPROTO;
}
if (rf->rf_lvblen != ls->ls_lvblen ||
@@ -98,7 +98,7 @@
log_error(ls, "config mismatch: %d,%x nodeid %d: %d,%x",
ls->ls_lvblen, ls->ls_exflags,
nodeid, rf->rf_lvblen, rf->rf_lsflags);
- return -EINVAL;
+ return -EPROTO;
}
return 0;
}