powerpc/powernv: Fix killed EEH event

On PowerNV platform, EEH errors are reported by IO accessors or poller
driven by interrupt. After the PE is isolated, we won't produce EEH
event for the PE. The current implementation has possibility of EEH
event lost in this way:

The interrupt handler queues one "special" event, which drives the poller.
EEH thread doesn't pick the special event yet. IO accessors kicks in, the
frozen PE is marked as "isolated" and EEH event is queued to the list.
EEH thread runs because of special event and purge all existing EEH events.
However, we never produce an other EEH event for the frozen PE. Eventually,
the PE is marked as "isolated" and we don't have EEH event to recover it.

The patch fixes the issue to keep EEH events for PEs that have been
marked as "isolated" with the help of additional "force" help to
eeh_remove_event().

Reported-by: Rolf Brudeseth <rolfb@us.ibm.com>
Signed-off-by: Gavin Shan <gwshan@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

2 files changed, 17 insertions, 8 deletions

diff --git a/arch/powerpc/kernel/eeh_driver.c b/arch/powerpc/kernel/eeh_driver.c
index 8bb40e7cdeb6..420da61d4ce0 100644
--- a/arch/powerpc/kernel/eeh_driver.c
+++ b/arch/powerpc/kernel/eeh_driver.c
@@ -770,7 +770,7 @@ static void eeh_handle_special_event(void)
 			eeh_serialize_lock(&flags);
 
 			/* Purge all events */
-			eeh_remove_event(NULL);
+			eeh_remove_event(NULL, true);
 
 			list_for_each_entry(hose, &hose_list, list_node) {
 				phb_pe = eeh_phb_pe_get(hose);
@@ -789,7 +789,7 @@ static void eeh_handle_special_event(void)
 			eeh_serialize_lock(&flags);
 
 			/* Purge all events of the PHB */
-			eeh_remove_event(pe);
+			eeh_remove_event(pe, true);
 
 			if (rc == EEH_NEXT_ERR_DEAD_PHB)
 				eeh_pe_state_mark(pe, EEH_PE_ISOLATED);
diff --git a/arch/powerpc/kernel/eeh_event.c b/arch/powerpc/kernel/eeh_event.c
index 72d748b56c86..4eefb6e34dbb 100644
--- a/arch/powerpc/kernel/eeh_event.c
+++ b/arch/powerpc/kernel/eeh_event.c
@@ -152,24 +152,33 @@ int eeh_send_failure_event(struct eeh_pe *pe)
 /**
  * eeh_remove_event - Remove EEH event from the queue
  * @pe: Event binding to the PE
+ * @force: Event will be removed unconditionally
  *
  * On PowerNV platform, we might have subsequent coming events
  * is part of the former one. For that case, those subsequent
  * coming events are totally duplicated and unnecessary, thus
  * they should be removed.
  */
-void eeh_remove_event(struct eeh_pe *pe)
+void eeh_remove_event(struct eeh_pe *pe, bool force)
 {
 	unsigned long flags;
 	struct eeh_event *event, *tmp;
 
+	/*
+	 * If we have NULL PE passed in, we have dead IOC
+	 * or we're sure we can report all existing errors
+	 * by the caller.
+	 *
+	 * With "force", the event with associated PE that
+	 * have been isolated, the event won't be removed
+	 * to avoid event lost.
+	 */
 	spin_lock_irqsave(&eeh_eventlist_lock, flags);
 	list_for_each_entry_safe(event, tmp, &eeh_eventlist, list) {
-		/*
-		 * If we don't have valid PE passed in, that means
-		 * we already have event corresponding to dead IOC
-		 * and all events should be purged.
-		 */
+		if (!force && event->pe &&
+		    (event->pe->state & EEH_PE_ISOLATED))
+			continue;
+
 		if (!pe) {
 			list_del(&event->list);
 			kfree(event);