dlm: detect available userspace daemon

If dlm_controld (the userspace daemon that controls the setup and
recovery of the dlm) fails, the kernel should shut down the lockspaces
in the kernel rather than leaving them running.  This is detected by
having dlm_controld hold a misc device open while running, and if
the kernel detects a close while the daemon is still needed, it stops
the lockspaces in the kernel.

Knowing that the userspace daemon isn't running also allows the
lockspace create/remove routines to avoid waiting on the daemon
for join/leave operations.

Signed-off-by: David Teigland <teigland@redhat.com>

4 files changed, 85 insertions, 2 deletions

diff --git a/fs/dlm/lockspace.c b/fs/dlm/lockspace.c
index 56eae4e4a954..ba672fe0a601 100644
--- a/fs/dlm/lockspace.c
+++ b/fs/dlm/lockspace.c
@@ -378,6 +378,11 @@ static int new_lockspace(char *name, int namelen, void **lockspace,
         if (!try_module_get(THIS_MODULE))
                 return -EINVAL;
 
+        if (!dlm_user_daemon_available()) {
+                module_put(THIS_MODULE);
+                return -EUNATCH;
+        }
+
         error = 0;
 
         spin_lock(&lslist_lock);
@@ -669,7 +674,7 @@ static int release_lockspace(struct dlm_ls *ls, int force)
 
         dlm_device_deregister(ls);
 
-        if (force < 3)
+        if (force < 3 && dlm_user_daemon_available())
                 do_uevent(ls, 0);
 
         dlm_recoverd_stop(ls);
@@ -791,3 +796,20 @@ int dlm_release_lockspace(void *lockspace, int force)
         return error;
 }
 
+void dlm_stop_lockspaces(void)
+{
+        struct dlm_ls *ls;
+
+ restart:
+        spin_lock(&lslist_lock);
+        list_for_each_entry(ls, &lslist, ls_list) {
+                if (!test_bit(LSFL_RUNNING, &ls->ls_flags))
+                        continue;
+                spin_unlock(&lslist_lock);
+                log_error(ls, "no userland control daemon, stopping lockspace");
+                dlm_ls_stop(ls);
+                goto restart;
+        }
+        spin_unlock(&lslist_lock);
+}
+
diff --git a/fs/dlm/lockspace.h b/fs/dlm/lockspace.h
index 891eabbdd021..f879f87901f8 100644
--- a/fs/dlm/lockspace.h
+++ b/fs/dlm/lockspace.h
@@ -20,6 +20,7 @@ struct dlm_ls *dlm_find_lockspace_global(uint32_t id);
 struct dlm_ls *dlm_find_lockspace_local(void *id);
 struct dlm_ls *dlm_find_lockspace_device(int minor);
 void dlm_put_lockspace(struct dlm_ls *ls);
+void dlm_stop_lockspaces(void);
 
 #endif                          /* __LOCKSPACE_DOT_H__ */
 
diff --git a/fs/dlm/user.c b/fs/dlm/user.c
index 6542110c0da4..81627b502a56 100644
--- a/fs/dlm/user.c
+++ b/fs/dlm/user.c
@@ -27,6 +27,8 @@
 
 static const char name_prefix[] = "dlm";
 static const struct file_operations device_fops;
+static atomic_t dlm_monitor_opened;
+static int dlm_monitor_unused = 1;
 
 #ifdef CONFIG_COMPAT
 
@@ -890,6 +892,26 @@ static unsigned int device_poll(struct file *file, poll_table *wait)
         return 0;
 }
 
+int dlm_user_daemon_available(void)
+{
+        /* dlm_controld hasn't started (or, has started, but not
+           properly populated configfs) */
+
+        if (!dlm_our_nodeid())
+                return 0;
+
+        /* This is to deal with versions of dlm_controld that don't
+           know about the monitor device.  We assume that if the
+           dlm_controld was started (above), but the monitor device
+           was never opened, that it's an old version.  dlm_controld
+           should open the monitor device before populating configfs. */
+
+        if (dlm_monitor_unused)
+                return 1;
+
+        return atomic_read(&dlm_monitor_opened) ? 1 : 0;
+}
+
 static int ctl_device_open(struct inode *inode, struct file *file)
 {
         cycle_kernel_lock();
@@ -902,6 +924,20 @@ static int ctl_device_close(struct inode *inode, struct file *file)
         return 0;
 }
 
+static int monitor_device_open(struct inode *inode, struct file *file)
+{
+        atomic_inc(&dlm_monitor_opened);
+        dlm_monitor_unused = 0;
+        return 0;
+}
+
+static int monitor_device_close(struct inode *inode, struct file *file)
+{
+        if (atomic_dec_and_test(&dlm_monitor_opened))
+                dlm_stop_lockspaces();
+        return 0;
+}
+
 static const struct file_operations device_fops = {
         .open    = device_open,
         .release = device_close,
@@ -925,19 +961,42 @@ static struct miscdevice ctl_device = {
         .minor = MISC_DYNAMIC_MINOR,
 };
 
+static const struct file_operations monitor_device_fops = {
+        .open    = monitor_device_open,
+        .release = monitor_device_close,
+        .owner   = THIS_MODULE,
+};
+
+static struct miscdevice monitor_device = {
+        .name  = "dlm-monitor",
+        .fops  = &monitor_device_fops,
+        .minor = MISC_DYNAMIC_MINOR,
+};
+
 int __init dlm_user_init(void)
 {
         int error;
 
+        atomic_set(&dlm_monitor_opened, 0);
+
         error = misc_register(&ctl_device);
-        if (error)
+        if (error) {
                 log_print("misc_register failed for control device");
+                goto out;
+        }
 
+        error = misc_register(&monitor_device);
+        if (error) {
+                log_print("misc_register failed for monitor device");
+                misc_deregister(&ctl_device);
+        }
+ out:
         return error;
 }
 
 void dlm_user_exit(void)
 {
         misc_deregister(&ctl_device);
+        misc_deregister(&monitor_device);
 }
 
diff --git a/fs/dlm/user.h b/fs/dlm/user.h
index c528b6b2991b..35eb6a13d616 100644
--- a/fs/dlm/user.h
+++ b/fs/dlm/user.h
@@ -13,5 +13,6 @@ void dlm_user_add_ast(struct dlm_lkb *lkb, int type);
 int dlm_user_init(void);
 void dlm_user_exit(void);
 int dlm_device_deregister(struct dlm_ls *ls);
+int dlm_user_daemon_available(void);
 
 #endif