[PATCH v2 13/15] isci: fix io failures while wide port links are coming up

[Dan Williams <dan.j.williams@intel.com>]

From: Marcin Tomczak <marcin.tomczak@intel.com>

When the first phy of a wide port comes up, don't report the port ready
yet, always wait for 250 miliseconds then config the port with all phys
added to the port. So that we can avoid reporting wide port device too
early to kernel, which caused the first IOs (report luns, inquirys)
failed due to not all the phys are configured into its port. Changes
also made that the phys in a wide port don't need to go through half
second wait time for consuming power.

Signed-off-by: Marcin Tomczak <marcin.tomczak@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/scsi/isci/host.c        |   59 ++++++++++++++++++++++++++++++++++++---
 drivers/scsi/isci/port_config.c |   33 ++++++++++++++--------
 2 files changed, 76 insertions(+), 16 deletions(-)

diff --git a/drivers/scsi/isci/host.c b/drivers/scsi/isci/host.c
index 7e4d709..1a65d65 100644
--- a/drivers/scsi/isci/host.c
+++ b/drivers/scsi/isci/host.c
@@ -899,7 +899,8 @@ static enum sci_status sci_controller_start_next_phy(struct isci_host *ihost)
 			 */
 			if ((iphy->is_in_link_training == false && state == SCI_PHY_INITIAL) ||
 			    (iphy->is_in_link_training == false && state == SCI_PHY_STOPPED) ||
-			    (iphy->is_in_link_training == true && is_phy_starting(iphy))) {
+			    (iphy->is_in_link_training == true && is_phy_starting(iphy)) ||
+			    (ihost->port_agent.phy_ready_mask != ihost->port_agent.phy_configured_mask)) {
 				is_controller_start_complete = false;
 				break;
 			}
@@ -1904,6 +1905,31 @@ static void power_control_timeout(unsigned long data)
 		ihost->power_control.phys_waiting--;
 		ihost->power_control.phys_granted_power++;
 		sci_phy_consume_power_handler(iphy);
+
+		if (iphy->protocol == SCIC_SDS_PHY_PROTOCOL_SAS) {
+			u8 j;
+
+			for (j = 0; j < SCI_MAX_PHYS; j++) {
+				struct isci_phy *requester = ihost->power_control.requesters[j];
+
+				/*
+				 * Search the power_control queue to see if there are other phys
+				 * attached to the same remote device. If found, take all of
+				 * them out of await_sas_power state.
+				 */
+				if (requester != NULL && requester != iphy) {
+					u8 other = memcmp(requester->frame_rcvd.iaf.sas_addr,
+							  iphy->frame_rcvd.iaf.sas_addr,
+							  sizeof(requester->frame_rcvd.iaf.sas_addr));
+
+					if (other == 0) {
+						ihost->power_control.requesters[j] = NULL;
+						ihost->power_control.phys_waiting--;
+						sci_phy_consume_power_handler(requester);
+					}
+				}
+			}
+		}
 	}
 
 	/*
@@ -1938,9 +1964,34 @@ void sci_controller_power_control_queue_insert(struct isci_host *ihost,
 		ihost->power_control.timer_started = true;
 
 	} else {
-		/* Add the phy in the waiting list */
-		ihost->power_control.requesters[iphy->phy_index] = iphy;
-		ihost->power_control.phys_waiting++;
+		/*
+		 * There are phys, attached to the same sas address as this phy, are
+		 * already in READY state, this phy don't need wait.
+		 */
+		u8 i;
+		struct isci_phy *current_phy;
+
+		for (i = 0; i < SCI_MAX_PHYS; i++) {
+			u8 other;
+			current_phy = &ihost->phys[i];
+
+			other = memcmp(current_phy->frame_rcvd.iaf.sas_addr,
+				       iphy->frame_rcvd.iaf.sas_addr,
+				       sizeof(current_phy->frame_rcvd.iaf.sas_addr));
+
+			if (current_phy->sm.current_state_id == SCI_PHY_READY &&
+			    current_phy->protocol == SCIC_SDS_PHY_PROTOCOL_SAS &&
+			    other == 0) {
+				sci_phy_consume_power_handler(iphy);
+				break;
+			}
+		}
+
+		if (i == SCI_MAX_PHYS) {
+			/* Add the phy in the waiting list */
+			ihost->power_control.requesters[iphy->phy_index] = iphy;
+			ihost->power_control.phys_waiting++;
+		}
 	}
 }
 
diff --git a/drivers/scsi/isci/port_config.c b/drivers/scsi/isci/port_config.c
index 37d15d3..6d1e954 100644
--- a/drivers/scsi/isci/port_config.c
+++ b/drivers/scsi/isci/port_config.c
@@ -466,6 +466,23 @@ sci_apc_agent_validate_phy_configuration(struct isci_host *ihost,
 	return sci_port_configuration_agent_validate_ports(ihost, port_agent);
 }
 
+/*
+ * This routine will restart the automatic port configuration timeout
+ * timer for the next time period. This could be caused by either a link
+ * down event or a link up event where we can not yet tell to which a phy
+ * belongs.
+ */
+static void sci_apc_agent_start_timer(
+	struct sci_port_configuration_agent *port_agent,
+	u32 timeout)
+{
+	if (port_agent->timer_pending)
+		sci_del_timer(&port_agent->timer);
+
+	port_agent->timer_pending = true;
+	sci_mod_timer(&port_agent->timer, timeout);
+}
+
 static void sci_apc_agent_configure_ports(struct isci_host *ihost,
 					       struct sci_port_configuration_agent *port_agent,
 					       struct isci_phy *iphy,
@@ -565,17 +582,8 @@ static void sci_apc_agent_configure_ports(struct isci_host *ihost,
 		break;
 
 	case SCIC_SDS_APC_START_TIMER:
-		/*
-		 * This can occur for either a link down event, or a link
-		 * up event where we cannot yet tell the port to which a
-		 * phy belongs.
-		 */
-		if (port_agent->timer_pending)
-			sci_del_timer(&port_agent->timer);
-
-		port_agent->timer_pending = true;
-		sci_mod_timer(&port_agent->timer,
-			      SCIC_SDS_APC_WAIT_LINK_UP_NOTIFICATION);
+		sci_apc_agent_start_timer(port_agent,
+					  SCIC_SDS_APC_WAIT_LINK_UP_NOTIFICATION);
 		break;
 
 	case SCIC_SDS_APC_SKIP_PHY:
@@ -607,7 +615,8 @@ static void sci_apc_agent_link_up(struct isci_host *ihost,
 	if (!iport) {
 		/* the phy is not the part of this port */
 		port_agent->phy_ready_mask |= 1 << phy_index;
-		sci_apc_agent_configure_ports(ihost, port_agent, iphy, true);
+		sci_apc_agent_start_timer(port_agent,
+					  SCIC_SDS_APC_WAIT_LINK_UP_NOTIFICATION);
 	} else {
 		/* the phy is already the part of the port */
 		u32 port_state = iport->sm.current_state_id;