Branch HEAD

1. delete the lc_watchdog_touch_ms()
2. changing to seconds instead of ms
3. delete lcw_time in lc_watchdog
b=18582
i=nathan
i=johann

diff --git a/libcfs/include/libcfs/libcfs.h b/libcfs/include/libcfs/libcfs.h
index 745f3a0..598276f 100644 (file)
--- a/libcfs/include/libcfs/libcfs.h
+++ b/libcfs/include/libcfs/libcfs.h
@@ -140,8 +140,10 @@ struct lc_watchdog *lc_watchdog_add(int time,
                                     void *data);
 
 /* Enables a watchdog and resets its timer. */
-void lc_watchdog_touch_ms(struct lc_watchdog *lcw, int timeout_ms);
-void lc_watchdog_touch(struct lc_watchdog *lcw);
+void lc_watchdog_touch(struct lc_watchdog *lcw, int timeout);
+#define GET_TIMEOUT(svc) (max_t(int, obd_timeout,                       \
+                          AT_OFF ? 0 : at_get(&svc->srv_at_estimate)) * \
+                          svc->srv_watchdog_factor)
 
 /* Disable a watchdog; touch it to restart it. */
 void lc_watchdog_disable(struct lc_watchdog *lcw);

diff --git a/libcfs/libcfs/watchdog.c b/libcfs/libcfs/watchdog.c
index 6c1801a..24fb8ee 100644 (file)
--- a/libcfs/libcfs/watchdog.c
+++ b/libcfs/libcfs/watchdog.c
@@ -53,7 +53,6 @@ struct lc_watchdog {
         void             *lcw_data;
 
         pid_t             lcw_pid;
-        cfs_duration_t    lcw_time; /* time until watchdog fires, jiffies */
 
         enum {
                 LC_WATCHDOG_DISABLED,
@@ -255,8 +254,7 @@ 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 " LPPID ": inactive for "
-                               "%lds\n", lcw->lcw_pid, cfs_duration_sec(lcw->lcw_time));
+                        CDEBUG(D_INFO, "found lcw for pid " LPPID "\n", lcw->lcw_pid);
 
                         if (lcw->lcw_state != LC_WATCHDOG_DISABLED)
                                 lcw->lcw_callback(lcw->lcw_pid, lcw->lcw_data);
@@ -312,7 +310,7 @@ static void lcw_dispatch_stop(void)
         EXIT;
 }
 
-struct lc_watchdog *lc_watchdog_add(int timeout_ms,
+struct lc_watchdog *lc_watchdog_add(int timeout,
                                     void (*callback)(pid_t, void *),
                                     void *data)
 {
@@ -327,7 +325,6 @@ struct lc_watchdog *lc_watchdog_add(int timeout_ms,
 
         lcw->lcw_task     = cfs_current();
         lcw->lcw_pid      = cfs_curproc_pid();
-        lcw->lcw_time     = cfs_time_seconds(timeout_ms) / 1000;
         lcw->lcw_callback = (callback != NULL) ? callback : lc_watchdog_dumplog;
         lcw->lcw_data     = data;
         lcw->lcw_state    = LC_WATCHDOG_DISABLED;
@@ -343,7 +340,7 @@ struct lc_watchdog *lc_watchdog_add(int timeout_ms,
         /* Keep this working in case we enable them by default */
         if (lcw->lcw_state == LC_WATCHDOG_ENABLED) {
                 lcw->lcw_last_touched = cfs_time_current();
-                cfs_timer_arm(&lcw->lcw_timer, lcw->lcw_time + 
+                cfs_timer_arm(&lcw->lcw_timer, cfs_time_seconds(timeout) +
                               cfs_time_current());
         }
 
@@ -368,7 +365,7 @@ static void lcw_update_time(struct lc_watchdog *lcw, const char *message)
         lcw->lcw_last_touched = newtime;
 }
 
-void lc_watchdog_touch_ms(struct lc_watchdog *lcw, int timeout_ms)
+void lc_watchdog_touch(struct lc_watchdog *lcw, int timeout)
 {
         ENTRY;
         LASSERT(lcw != NULL);
@@ -381,17 +378,10 @@ void lc_watchdog_touch_ms(struct lc_watchdog *lcw, int timeout_ms)
         lcw->lcw_state = LC_WATCHDOG_ENABLED;
 
         cfs_timer_arm(&lcw->lcw_timer, cfs_time_current() +
-                      cfs_time_seconds(timeout_ms) / 1000);
+                      cfs_time_seconds(timeout));
 
         EXIT;
 }
-EXPORT_SYMBOL(lc_watchdog_touch_ms);
-
-/* deprecated - use above instead */
-void lc_watchdog_touch(struct lc_watchdog *lcw)
-{
-        lc_watchdog_touch_ms(lcw, (int)cfs_duration_sec(lcw->lcw_time) * 1000);
-}
 EXPORT_SYMBOL(lc_watchdog_touch);
 
 void lc_watchdog_disable(struct lc_watchdog *lcw)
@@ -448,7 +438,7 @@ EXPORT_SYMBOL(lc_watchdog_dumplog);
 
 #else   /* !defined(WITH_WATCHDOG) */
 
-struct lc_watchdog *lc_watchdog_add(int timeout_ms,
+struct lc_watchdog *lc_watchdog_add(int timeout,
                                     void (*callback)(pid_t pid, void *),
                                     void *data)
 {
@@ -457,12 +447,7 @@ struct lc_watchdog *lc_watchdog_add(int timeout_ms,
 }
 EXPORT_SYMBOL(lc_watchdog_add);
 
-void lc_watchdog_touch_ms(struct lc_watchdog *lcw, int timeout_ms)
-{
-}
-EXPORT_SYMBOL(lc_watchdog_touch_ms);
-
-void lc_watchdog_touch(struct lc_watchdog *lcw)
+void lc_watchdog_touch(struct lc_watchdog *lcw, int timeout)
 {
 }
 EXPORT_SYMBOL(lc_watchdog_touch);

diff --git a/lustre/include/lustre_net.h b/lustre/include/lustre_net.h
index b1dcac3..f94116b 100644 (file)
--- a/lustre/include/lustre_net.h
+++ b/lustre/include/lustre_net.h
@@ -658,6 +658,10 @@ struct ptlrpc_thread {
          * put watchdog in the structure per thread b=14840
          */
         struct lc_watchdog *t_watchdog;
+        /**
+         * the svc this thread belonged to b=18582
+         */
+        struct ptlrpc_service *t_svc;
         cfs_waitq_t t_ctl_waitq;
         struct lu_env *t_env;
 };

diff --git a/lustre/ldlm/ldlm_lockd.c b/lustre/ldlm/ldlm_lockd.c
index f6fe4a7..6237590 100644 (file)
--- a/lustre/ldlm/ldlm_lockd.c
+++ b/lustre/ldlm/ldlm_lockd.c
@@ -2260,7 +2260,7 @@ static int ldlm_setup(void)
         ldlm_state->ldlm_cb_service =
                 ptlrpc_init_svc(LDLM_NBUFS, LDLM_BUFSIZE, LDLM_MAXREQSIZE,
                                 LDLM_MAXREPSIZE, LDLM_CB_REQUEST_PORTAL,
-                                LDLM_CB_REPLY_PORTAL, 1800,
+                                LDLM_CB_REPLY_PORTAL, 2,
                                 ldlm_callback_handler, "ldlm_cbd",
                                 ldlm_svc_proc_dir, NULL,
                                 ldlm_min_threads, ldlm_max_threads,
@@ -2275,7 +2275,7 @@ static int ldlm_setup(void)
         ldlm_state->ldlm_cancel_service =
                 ptlrpc_init_svc(LDLM_NBUFS, LDLM_BUFSIZE, LDLM_MAXREQSIZE,
                                 LDLM_MAXREPSIZE, LDLM_CANCEL_REQUEST_PORTAL,
-                                LDLM_CANCEL_REPLY_PORTAL, 6000,
+                                LDLM_CANCEL_REPLY_PORTAL, 6,
                                 ldlm_cancel_handler, "ldlm_canceld",
                                 ldlm_svc_proc_dir, NULL,
                                 ldlm_min_threads, ldlm_max_threads,

diff --git a/lustre/mdt/mdt_handler.c b/lustre/mdt/mdt_handler.c
index 661ad8e..3212962 100644 (file)
--- a/lustre/mdt/mdt_handler.c
+++ b/lustre/mdt/mdt_handler.c
@@ -349,7 +349,7 @@ static int mdt_statfs(struct mdt_thread_info *info)
         /* This will trigger a watchdog timeout */
         OBD_FAIL_TIMEOUT(OBD_FAIL_MDS_STATFS_LCW_SLEEP,
                          (MDT_SERVICE_WATCHDOG_FACTOR *
-                          at_get(&svc->srv_at_estimate) / 1000) + 1);
+                          at_get(&svc->srv_at_estimate)) + 1);
 
         rc = mdt_check_ucred(info);
         if (rc)

diff --git a/lustre/mdt/mdt_internal.h b/lustre/mdt/mdt_internal.h
index 953ede9..aa6c6cc 100644 (file)
--- a/lustre/mdt/mdt_internal.h
+++ b/lustre/mdt/mdt_internal.h
@@ -175,7 +175,7 @@ struct mdt_device {
 #define mdt_client_bitmap_lock  mdt_lut.lut_client_bitmap_lock
 #define mdt_client_bitmap       mdt_lut.lut_client_bitmap
 
-#define MDT_SERVICE_WATCHDOG_FACTOR     (2000)
+#define MDT_SERVICE_WATCHDOG_FACTOR     (2)
 #define MDT_ROCOMPAT_SUPP       (OBD_ROCOMPAT_LOVOBJID)
 #define MDT_INCOMPAT_SUPP       (OBD_INCOMPAT_MDT | OBD_INCOMPAT_COMMON_LR)
 #define MDT_COS_DEFAULT         (1)

diff --git a/lustre/mgs/mgs_handler.c b/lustre/mgs/mgs_handler.c
index 97878c8..0df7efc 100644 (file)
--- a/lustre/mgs/mgs_handler.c
+++ b/lustre/mgs/mgs_handler.c
@@ -250,7 +250,7 @@ static int mgs_setup(struct obd_device *obd, struct lustre_cfg *lcfg)
         mgs->mgs_service =
                 ptlrpc_init_svc(MGS_NBUFS, MGS_BUFSIZE, MGS_MAXREQSIZE,
                                 MGS_MAXREPSIZE, MGS_REQUEST_PORTAL,
-                                MGC_REPLY_PORTAL, 2000,
+                                MGC_REPLY_PORTAL, 2,
                                 mgs_handle, LUSTRE_MGS_NAME,
                                 obd->obd_proc_entry, target_print_req,
                                 MGS_THREADS_AUTO_MIN, MGS_THREADS_AUTO_MAX,

diff --git a/lustre/ost/ost_internal.h b/lustre/ost/ost_internal.h
index 519bbb3..6775362 100644 (file)
--- a/lustre/ost/ost_internal.h
+++ b/lustre/ost/ost_internal.h
@@ -37,7 +37,7 @@
 #ifndef OST_INTERNAL_H
 #define OST_INTERNAL_H
 
-#define OSS_SERVICE_WATCHDOG_FACTOR 2000
+#define OSS_SERVICE_WATCHDOG_FACTOR 2
 
 /*
  * tunables for per-thread page pool (bug 5137)

diff --git a/lustre/ptlrpc/service.c b/lustre/ptlrpc/service.c
index 35a2eb4..d28db7b 100644 (file)
--- a/lustre/ptlrpc/service.c
+++ b/lustre/ptlrpc/service.c
@@ -1917,10 +1917,7 @@ static int ptlrpc_main(void *arg)
          */
         cfs_waitq_signal(&thread->t_ctl_waitq);
 
-        thread->t_watchdog = lc_watchdog_add(max_t(int, obd_timeout, AT_OFF ? 0 :
-                                                   at_get(&svc->srv_at_estimate))
-                                             *  svc->srv_watchdog_factor,
-                                             NULL, NULL);
+        thread->t_watchdog = lc_watchdog_add(GET_TIMEOUT(svc), NULL, NULL);
 
         spin_lock(&svc->srv_lock);
         svc->srv_threads_running++;
@@ -1953,10 +1950,7 @@ static int ptlrpc_main(void *arg)
                               svc->srv_at_check,
                               &lwi);
 
-                lc_watchdog_touch_ms(thread->t_watchdog, max_t(int, obd_timeout,
-                                     AT_OFF ? 0 :
-                                     at_get(&svc->srv_at_estimate)) *
-                                     svc->srv_watchdog_factor);
+                lc_watchdog_touch(thread->t_watchdog, GET_TIMEOUT(svc));
 
                 ptlrpc_check_rqbd_pool(svc);
 
@@ -2256,6 +2250,7 @@ int ptlrpc_start_thread(struct obd_device *dev, struct ptlrpc_service *svc)
         id = svc->srv_threads_started++;
         spin_unlock(&svc->srv_lock);
 
+        thread->t_svc = svc;
         thread->t_id = id;
         sprintf(name, "%s_%02d", svc->srv_thread_name, id);
         d.dev = dev;

diff --git a/lustre/quota/quota_context.c b/lustre/quota/quota_context.c
index cccc3fd..537f101 100644 (file)
--- a/lustre/quota/quota_context.c
+++ b/lustre/quota/quota_context.c
@@ -940,7 +940,8 @@ schedule_dqacq(struct obd_device *obd, struct lustre_quota_ctxt *qctxt,
                         l_wait_event(qctxt->lqc_wait_for_qmaster,
                                      check_qm(qctxt), &lwi);
                         CDEBUG(D_QUOTA, "wake up when quota master is back\n");
-                        lc_watchdog_touch(oti->oti_thread->t_watchdog);
+                        lc_watchdog_touch(oti->oti_thread->t_watchdog,
+                                 GET_TIMEOUT(oti->oti_thread->t_svc));
                 } else {
                         spin_unlock(&qctxt->lqc_lock);
                 }

diff --git a/lustre/quota/quota_interface.c b/lustre/quota/quota_interface.c
index ef40fd8..16ed64e 100644 (file)
--- a/lustre/quota/quota_interface.c
+++ b/lustre/quota/quota_interface.c
@@ -393,7 +393,8 @@ static int quota_chk_acq_common(struct obd_device *obd, unsigned int uid,
                         l_wait_event(qctxt->lqc_wait_for_qmaster, check_qm(qctxt),
                                      &lwi);
                         CDEBUG(D_QUOTA, "wake up when quota master is back\n");
-                        lc_watchdog_touch(oti->oti_thread->t_watchdog);
+                        lc_watchdog_touch(oti->oti_thread->t_watchdog,
+                                 GET_TIMEOUT(oti->oti_thread->t_svc));
                 } else {
                         spin_unlock(&qctxt->lqc_lock);
                 }
@@ -424,7 +425,8 @@ static int quota_chk_acq_common(struct obd_device *obd, unsigned int uid,
                         struct l_wait_info lwi;
 
                         if (oti && oti->oti_thread && oti->oti_thread->t_watchdog)
-                                lc_watchdog_touch(oti->oti_thread->t_watchdog);
+                                lc_watchdog_touch(oti->oti_thread->t_watchdog,
+                                         GET_TIMEOUT(oti->oti_thread->t_svc));
                         CDEBUG(D_QUOTA, "rc: %d, count_err: %d\n", rc,
                                count_err++);