[SCSI] zfcp: fix race conditions when accessing erp_action lists
authorAndreas Herrmann <aherrman@de.ibm.com>
Tue, 13 Sep 2005 19:47:11 +0000 (21:47 +0200)
committerJames Bottomley <jejb@mulgrave.(none)>
Mon, 19 Sep 2005 18:00:50 +0000 (13:00 -0500)
o always use locking when changing erp_action lists,
o avoid escalation to ERP_ACTION_REOPEN_PORT_FORCED if erp_action is
  still in use for ERP_ACTION_REOPEN_PORT

Signed-off-by: Andreas Herrmann <aherrman@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
drivers/s390/scsi/zfcp_erp.c

index cb4f612550baf459ed752ba385dbeca823574958..376cb0f6cb744c354dbbbc4d17642f4d30f29d1a 100644 (file)
@@ -886,7 +886,7 @@ static int
 zfcp_erp_strategy_check_fsfreq(struct zfcp_erp_action *erp_action)
 {
        int retval = 0;
-       struct zfcp_fsf_req *fsf_req;
+       struct zfcp_fsf_req *fsf_req = NULL;
        struct zfcp_adapter *adapter = erp_action->adapter;
 
        if (erp_action->fsf_req) {
@@ -896,7 +896,7 @@ zfcp_erp_strategy_check_fsfreq(struct zfcp_erp_action *erp_action)
                list_for_each_entry(fsf_req, &adapter->fsf_req_list_head, list)
                    if (fsf_req == erp_action->fsf_req)
                        break;
-               if (fsf_req == erp_action->fsf_req) {
+               if (fsf_req && (fsf_req->erp_action == erp_action)) {
                        /* fsf_req still exists */
                        debug_text_event(adapter->erp_dbf, 3, "a_ca_req");
                        debug_event(adapter->erp_dbf, 3, &fsf_req,
@@ -2291,7 +2291,9 @@ zfcp_erp_adapter_strategy_open_fsf_xconfig(struct zfcp_erp_action *erp_action)
                atomic_clear_mask(ZFCP_STATUS_ADAPTER_HOST_CON_INIT,
                                  &adapter->status);
                ZFCP_LOG_DEBUG("Doing exchange config data\n");
+               write_lock(&adapter->erp_lock);
                zfcp_erp_action_to_running(erp_action);
+               write_unlock(&adapter->erp_lock);
                zfcp_erp_timeout_init(erp_action);
                if (zfcp_fsf_exchange_config_data(erp_action)) {
                        retval = ZFCP_ERP_FAILED;
@@ -3194,11 +3196,19 @@ zfcp_erp_action_enqueue(int action,
                /* fall through !!! */
 
        case ZFCP_ERP_ACTION_REOPEN_PORT_FORCED:
-               if (atomic_test_mask
-                   (ZFCP_STATUS_COMMON_ERP_INUSE, &port->status)
-                   && port->erp_action.action ==
-                   ZFCP_ERP_ACTION_REOPEN_PORT_FORCED) {
-                       debug_text_event(adapter->erp_dbf, 4, "pf_actenq_drp");
+               if (atomic_test_mask(ZFCP_STATUS_COMMON_ERP_INUSE,
+                                    &port->status)) {
+                       if (port->erp_action.action !=
+                           ZFCP_ERP_ACTION_REOPEN_PORT_FORCED) {
+                               ZFCP_LOG_INFO("dropped erp action %i (port "
+                                             "0x%016Lx, action in use: %i)\n",
+                                             action, port->wwpn,
+                                             port->erp_action.action);
+                               debug_text_event(adapter->erp_dbf, 4,
+                                                "pf_actenq_drp");
+                       } else 
+                               debug_text_event(adapter->erp_dbf, 4,
+                                                "pf_actenq_drpcp");
                        debug_event(adapter->erp_dbf, 4, &port->wwpn,
                                    sizeof (wwn_t));
                        goto out;