b=10894

r=adilger
Clarify watchdog timer units.
Add ability to change MDT ldlm timeout.

diff --git a/lnet/libcfs/watchdog.c b/lnet/libcfs/watchdog.c
index 3000e8f..22754af 100644 (file)
--- a/lnet/libcfs/watchdog.c
+++ b/lnet/libcfs/watchdog.c
@@ -36,7 +36,7 @@ struct lc_watchdog {
         void             *lcw_data;
 
         pid_t             lcw_pid;
-        int               lcw_time; /* time until watchdog fires, in ms */
+        cfs_duration_t    lcw_time; /* time until watchdog fires, jiffies */
 
         enum {
                 LC_WATCHDOG_DISABLED,
@@ -126,8 +126,8 @@ static void lcw_cb(unsigned long data)
         /* NB this warning should appear on the console, but may not get into
          * the logs since we're running in a softirq handler */
 
-        CWARN("Watchdog triggered for pid %d: it was inactive for %ldms\n",
-              (int)lcw->lcw_pid, cfs_duration_sec(lcw->lcw_time) * 1000);
+        CWARN("Watchdog triggered for pid %d: it was inactive for %lds\n",
+              (int)lcw->lcw_pid, cfs_duration_sec(lcw->lcw_time));
         lcw_dump(lcw);
 
         spin_lock_bh(&lcw_pending_timers_lock);
@@ -197,8 +197,9 @@ static int lcw_dispatch_main(void *data)
                         list_del_init(&lcw->lcw_list);
                         spin_unlock_bh(&lcw_pending_timers_lock);
 
-                        CDEBUG(D_INFO, "found lcw for pid %d: inactive for %ldms\n", 
-                               (int)lcw->lcw_pid, cfs_duration_sec(lcw->lcw_time) * 1000);
+                        CDEBUG(D_INFO, "found lcw for pid %d: inactive for "
+                               "%lds\n", (int)lcw->lcw_pid,
+                               cfs_duration_sec(lcw->lcw_time));
 
                         if (lcw->lcw_state != LC_WATCHDOG_DISABLED)
                                 lcw->lcw_callback(lcw->lcw_pid, lcw->lcw_data);

diff --git a/lustre/include/lustre_net.h b/lustre/include/lustre_net.h
index 5835442..e1fa8e8 100644 (file)
--- a/lustre/include/lustre_net.h
+++ b/lustre/include/lustre_net.h
@@ -120,7 +120,7 @@
 #define MDS_MAXREPSIZE  max(9 * 1024, 280 + LOV_MAX_STRIPE_COUNT * 56)
 
 #define MGS_THREADS_AUTO_MIN 2
-#define MGS_THREADS_AUTO_MAX 128
+#define MGS_THREADS_AUTO_MAX 32
 #define MGS_NBUFS       (64 * smp_num_cpus)
 #define MGS_BUFSIZE     (8 * 1024)
 #define MGS_MAXREQSIZE  (8 * 1024)

diff --git a/lustre/include/obd_support.h b/lustre/include/obd_support.h
index 6cdcc9a..1b31dd4 100644 (file)
--- a/lustre/include/obd_support.h
+++ b/lustre/include/obd_support.h
@@ -35,6 +35,7 @@ extern unsigned int obd_dump_on_eviction;
 extern unsigned int obd_timeout;          /* seconds */
 #define PING_INTERVAL max(obd_timeout / 4, 1U)
 #define RECONNECT_INTERVAL max(obd_timeout / 10, 10U)
+#define LDLM_TIMEOUT_DEFAULT 20
 extern unsigned int ldlm_timeout;
 extern unsigned int obd_health_check_timeout;
 extern unsigned int obd_sync_filter;
@@ -188,6 +189,7 @@ extern int obd_race_state;
 #define OBD_FAIL_MGS_ALL_REPLY_NET       0x902
 #define OBD_FAIL_MGC_PROCESS_LOG         0x903
 #define OBD_FAIL_MGS_SLOW_REQUEST_NET    0x904
+#define OBD_FAIL_MGS_SLOW_TARGET_REG     0x905
 
 #define OBD_FAIL_QUOTA_QD_COUNT_32BIT    0xA00
 

diff --git a/lustre/mds/handler.c b/lustre/mds/handler.c
index 5f59ab3..8cc1aba 100644 (file)
--- a/lustre/mds/handler.c
+++ b/lustre/mds/handler.c
@@ -2053,7 +2053,8 @@ static int mds_setup(struct obd_device *obd, obd_count len, void *buf)
                               obd->obd_replayable ? "enabled" : "disabled");
         }
 
-        ldlm_timeout = 6;
+        if (ldlm_timeout == LDLM_TIMEOUT_DEFAULT)
+                ldlm_timeout = 6;
 
         RETURN(0);
 

diff --git a/lustre/mgs/mgs_handler.c b/lustre/mgs/mgs_handler.c
index e161ad3..e6ba9be 100644 (file)
--- a/lustre/mgs/mgs_handler.c
+++ b/lustre/mgs/mgs_handler.c
@@ -382,6 +382,8 @@ static int mgs_handle_target_reg(struct ptlrpc_request *req)
                                obd->obd_name, lockrc);
         }
 
+        OBD_FAIL_TIMEOUT(OBD_FAIL_MGS_SLOW_TARGET_REG, 10);
+
         /* Log writing contention is handled by the fsdb_sem */
 
         if (mti->mti_flags & LDD_F_WRITECONF) {

diff --git a/lustre/obdclass/class_obd.c b/lustre/obdclass/class_obd.c
index 21732e6..c2af448 100644 (file)
--- a/lustre/obdclass/class_obd.c
+++ b/lustre/obdclass/class_obd.c
@@ -64,7 +64,7 @@ unsigned int obd_debug_peer_on_timeout;
 unsigned int obd_dump_on_timeout;
 unsigned int obd_dump_on_eviction;
 unsigned int obd_timeout = 100; /* seconds */
-unsigned int ldlm_timeout = 20; /* seconds */
+unsigned int ldlm_timeout = LDLM_TIMEOUT_DEFAULT; /* seconds */
 unsigned int obd_health_check_timeout = 120; /* seconds */
 unsigned int obd_max_dirty_pages = 256;
 atomic_t obd_dirty_pages;