staging: lustre: ldlm: cancel aged locks for LRUR
authorNiu Yawei <yawei.niu@intel.com>
Mon, 3 Oct 2016 02:28:30 +0000 (22:28 -0400)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sun, 16 Oct 2016 08:24:38 +0000 (10:24 +0200)
It doesn't make sense to keep the very aged lock even with the
LRUR policy. This patch decreased the default ns_max_age from 10
hours to 65 minutes and changed LRUR policy to cancel very aged
locks.

Signed-off-by: Niu Yawei <yawei.niu@intel.com>
Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-6529
Reviewed-on: http://review.whamcloud.com/14856
Reviewed-by: Jinshan Xiong <jinshan.xiong@intel.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>
Signed-off-by: James Simmons <jsimmons@infradead.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/staging/lustre/lustre/include/lustre_dlm.h
drivers/staging/lustre/lustre/ldlm/ldlm_request.c

index d035344..1c6b7b8 100644 (file)
@@ -59,7 +59,7 @@ struct obd_device;
 #define OBD_LDLM_DEVICENAME  "ldlm"
 
 #define LDLM_DEFAULT_LRU_SIZE (100 * num_online_cpus())
-#define LDLM_DEFAULT_MAX_ALIVE (cfs_time_seconds(36000))
+#define LDLM_DEFAULT_MAX_ALIVE (cfs_time_seconds(3900)) /* 65 min */
 #define LDLM_DEFAULT_PARALLEL_AST_LIMIT 1024
 
 /**
index 98730a3..ac1927c 100644 (file)
@@ -1182,6 +1182,14 @@ static enum ldlm_policy_res ldlm_cancel_lrur_policy(struct ldlm_namespace *ns,
        if (count && added >= count)
                return LDLM_POLICY_KEEP_LOCK;
 
+       /*
+        * Despite of the LV, It doesn't make sense to keep the lock which
+        * is unused for ns_max_age time.
+        */
+       if (cfs_time_after(cfs_time_current(),
+                          cfs_time_add(lock->l_last_used, ns->ns_max_age)))
+               return LDLM_POLICY_CANCEL_LOCK;
+
        slv = ldlm_pool_get_slv(pl);
        lvf = ldlm_pool_get_lvf(pl);
        la = cfs_duration_sec(cfs_time_sub(cur, lock->l_last_used));