From: Hans-Joachim Picht <hpicht@redhat.com>
Date: Wed, 22 Oct 2008 14:17:35 +0200
Subject: [s390] qdio: repair timeout handling for qdio_shutdown
Message-id: 20081022121735.GC26181@redhat.com
O-Subject: [RHEL5 U4 PATCH 3/5] s390 - qdio: Repair timeout handling for qdio_shutdown
Bugzilla: 463164
RH-Acked-by: Pete Zaitcev <zaitcev@redhat.com>

Description
============

If qdio shutdown runs in parallel with a channel error,
the qdio_timeout_handler might not be triggered.
In this case neither state INACTIVE nor state ERR
is reached and the following wait_event hangs forever.
Solution: do not make use of ccw_device_set_timeout(),
but add a timeout to the following wait_event.
And make sure, wake_up is called in case of an
i/o error on the qdio-device.

Bugzilla
=========

BZ 463164
https://bugzilla.redhat.com/show_bug.cgi?id=463164

Upstream status of the patch:
=============================

The problem has been fixed upstream as part of the qdio-driver rewrite.
See: http://lkml.org/lkml/2008/7/1/240

Test status:
============

The patch has been tested and fixes the problem.
The fix has been verified by the IBM test department.

Please ACK.

With best regards,

	--Hans

diff --git a/drivers/s390/cio/qdio.c b/drivers/s390/cio/qdio.c
index df55883..c10667e 100644
--- a/drivers/s390/cio/qdio.c
+++ b/drivers/s390/cio/qdio.c
@@ -2089,7 +2089,6 @@ qdio_timeout_handler(struct ccw_device *cdev)
 	default:
 		BUG();
 	}
-	ccw_device_set_timeout(cdev, 0);
 	wake_up(&cdev->private->wait_q);
 }
 
@@ -2128,6 +2127,8 @@ qdio_handler(struct ccw_device *cdev, unsigned long intparm, struct irb *irb)
 		case -EIO:
 			QDIO_PRINT_ERR("i/o error on device %s\n",
 				       cdev->dev.bus_id);
+			qdio_set_state(irq_ptr, QDIO_IRQ_STATE_ERR);
+			wake_up(&cdev->private->wait_q);
 			return;
 		case -ETIMEDOUT:
 			qdio_timeout_handler(cdev);
@@ -2674,12 +2675,12 @@ qdio_shutdown(struct ccw_device *cdev, int how)
 		spin_unlock_irqrestore(get_ccwdev_lock(cdev), flags);
 	} else if (rc == 0) {
 		qdio_set_state(irq_ptr, QDIO_IRQ_STATE_CLEANUP);
-		ccw_device_set_timeout(cdev, timeout);
 		spin_unlock_irqrestore(get_ccwdev_lock(cdev),flags);
 
-		wait_event(cdev->private->wait_q,
-			   irq_ptr->state == QDIO_IRQ_STATE_INACTIVE ||
-			   irq_ptr->state == QDIO_IRQ_STATE_ERR);
+		wait_event_interruptible_timeout(cdev->private->wait_q,
+			irq_ptr->state == QDIO_IRQ_STATE_INACTIVE ||
+			irq_ptr->state == QDIO_IRQ_STATE_ERR,
+			timeout);
 	} else {
 		QDIO_PRINT_INFO("ccw_device_{halt,clear} returned %d for "
 				"device %s\n", result, cdev->dev.bus_id);
@@ -2699,7 +2700,6 @@ qdio_shutdown(struct ccw_device *cdev, int how)
 
 	/* Ignore errors. */
 	qdio_set_state(irq_ptr, QDIO_IRQ_STATE_INACTIVE);
-	ccw_device_set_timeout(cdev, 0);
 out:
 	up(&irq_ptr->setting_up_sema);
 	return result;
@@ -2913,13 +2913,10 @@ qdio_establish_handle_irq(struct ccw_device *cdev, int cstat, int dstat)
 	QDIO_DBF_TEXT0(0,setup,dbf_text);
 	QDIO_DBF_TEXT0(0,trace,dbf_text);
 
-	if (qdio_establish_irq_check_for_errors(cdev, cstat, dstat)) {
-		ccw_device_set_timeout(cdev, 0);
+	if (qdio_establish_irq_check_for_errors(cdev, cstat, dstat))
 		return;
-	}
 
 	qdio_set_state(irq_ptr,QDIO_IRQ_STATE_ESTABLISHED);
-	ccw_device_set_timeout(cdev, 0);
 }
 
 int
@@ -3195,8 +3192,6 @@ qdio_establish(struct qdio_initialize *init_data)
 				irq_ptr->schid.ssid, irq_ptr->schid.sch_no,
 				result, result2);
 		result=result2;
-		if (result)
-			ccw_device_set_timeout(cdev, 0);
 	}
 
 	spin_unlock_irqrestore(get_ccwdev_lock(cdev),saveflags);
@@ -3279,7 +3274,6 @@ qdio_activate(struct ccw_device *cdev, int flags)
 
 	spin_lock_irqsave(get_ccwdev_lock(cdev),saveflags);
 
-	ccw_device_set_timeout(cdev, 0);
 	ccw_device_set_options(cdev, CCWDEV_REPORT_ALL);
 	result=ccw_device_start(cdev,&irq_ptr->ccw,QDIO_DOING_ACTIVATE,
 				0, DOIO_DENY_PREFETCH);