In ldlm_resource_add_lock(), call to ldlm_resource_dump()

starve other threads from the resource lock for a long time in
case of long waiting queue, so change the debug level from
D_OTHER to the less frequently used D_INFO.

b=15953
i=adilger
i=nathan.rutman

diff --git a/lustre/ChangeLog b/lustre/ChangeLog
index 5a6189e..5a9ea4d 100644 (file)
--- a/lustre/ChangeLog
+++ b/lustre/ChangeLog
@@ -184,6 +184,14 @@ Details    : Because the stride_start_offset are missing in stride read-ahead,
             it will cause clients read a lot of unused pages in read-ahead,
             then the read-performance drops.
 
+Severity   : normal
+Bugzilla   : 15953
+Description: more ldlm soft lockups
+Details    : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
+             starve other threads from the resource lock for a long time in
+             case of long waiting queue, so change the debug level from
+             D_OTHER to the less frequently used D_INFO.
+
 -------------------------------------------------------------------------------
 
 

diff --git a/lustre/ldlm/ldlm_resource.c b/lustre/ldlm/ldlm_resource.c
index 138743a..33a4bed 100644 (file)
--- a/lustre/ldlm/ldlm_resource.c
+++ b/lustre/ldlm/ldlm_resource.c
@@ -974,9 +974,9 @@ void ldlm_resource_add_lock(struct ldlm_resource *res, struct list_head *head,
 {
         check_res_locked(res);
 
-        ldlm_resource_dump(D_OTHER, res);
-        CDEBUG(D_OTHER, "About to add this lock:\n");
-        ldlm_lock_dump(D_OTHER, lock, 0);
+        ldlm_resource_dump(D_INFO, res);
+        CDEBUG(D_INFO, "About to add this lock:\n");
+        ldlm_lock_dump(D_INFO, lock, 0);
 
         if (lock->l_destroyed) {
                 CDEBUG(D_OTHER, "Lock destroyed, not adding to resource\n");