[S390] cio: call ccw driver notify function with lock held

Calling a ccw driver's notify function without the ccw device lock
held opens up a race window between discovery and handling of a change
in the device operational state. As a result, the device driver may
encounter unexpected device malfunction, leading to out-of-retry
situations or similar.

Remove race by extending the ccw device lock from state change
discovery to the calling of the notify function.

Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>

5 files changed, 41 insertions, 38 deletions

diff --git a/drivers/s390/block/dasd.c b/drivers/s390/block/dasd.c
index 1b6c52ef7339..acb78017e7d0 100644
--- a/drivers/s390/block/dasd.c
+++ b/drivers/s390/block/dasd.c
@@ -2333,13 +2333,11 @@ int dasd_generic_notify(struct ccw_device *cdev, int event)
 {
         struct dasd_device *device;
         struct dasd_ccw_req *cqr;
-        unsigned long flags;
         int ret;
 
-        device = dasd_device_from_cdev(cdev);
+        device = dasd_device_from_cdev_locked(cdev);
         if (IS_ERR(device))
                 return 0;
-        spin_lock_irqsave(get_ccwdev_lock(cdev), flags);
         ret = 0;
         switch (event) {
         case CIO_GONE:
@@ -2369,7 +2367,6 @@ int dasd_generic_notify(struct ccw_device *cdev, int event)
                 ret = 1;
                 break;
         }
-        spin_unlock_irqrestore(get_ccwdev_lock(cdev), flags);
         dasd_put_device(device);
         return ret;
 }
diff --git a/drivers/s390/cio/css.c b/drivers/s390/cio/css.c
index 46c021d880dc..51489eff6b0b 100644
--- a/drivers/s390/cio/css.c
+++ b/drivers/s390/cio/css.c
@@ -477,7 +477,6 @@ void css_schedule_eval_all(void)
 
 void css_wait_for_slow_path(void)
 {
-        flush_workqueue(ccw_device_notify_work);
         flush_workqueue(slow_path_wq);
 }
 
diff --git a/drivers/s390/cio/device.c b/drivers/s390/cio/device.c
index e818d0c54c09..28221030b886 100644
--- a/drivers/s390/cio/device.c
+++ b/drivers/s390/cio/device.c
@@ -150,7 +150,6 @@ static struct css_driver io_subchannel_driver = {
 };
 
 struct workqueue_struct *ccw_device_work;
-struct workqueue_struct *ccw_device_notify_work;
 wait_queue_head_t ccw_device_init_wq;
 atomic_t ccw_device_init_count;
 
@@ -168,11 +167,6 @@ init_ccw_bus_type (void)
         ccw_device_work = create_singlethread_workqueue("cio");
         if (!ccw_device_work)
                 return -ENOMEM; /* FIXME: better errno ? */
-        ccw_device_notify_work = create_singlethread_workqueue("cio_notify");
-        if (!ccw_device_notify_work) {
-                ret = -ENOMEM; /* FIXME: better errno ? */
-                goto out_err;
-        }
         slow_path_wq = create_singlethread_workqueue("kslowcrw");
         if (!slow_path_wq) {
                 ret = -ENOMEM; /* FIXME: better errno ? */
@@ -192,8 +186,6 @@ init_ccw_bus_type (void)
 out_err:
         if (ccw_device_work)
                 destroy_workqueue(ccw_device_work);
-        if (ccw_device_notify_work)
-                destroy_workqueue(ccw_device_notify_work);
         if (slow_path_wq)
                 destroy_workqueue(slow_path_wq);
         return ret;
@@ -204,7 +196,6 @@ cleanup_ccw_bus_type (void)
 {
         css_driver_unregister(&io_subchannel_driver);
         bus_unregister(&ccw_bus_type);
-        destroy_workqueue(ccw_device_notify_work);
         destroy_workqueue(ccw_device_work);
 }
 
@@ -1496,11 +1487,22 @@ static void device_set_disconnected(struct ccw_device *cdev)
                 ccw_device_schedule_recovery();
 }
 
+void ccw_device_set_notoper(struct ccw_device *cdev)
+{
+        struct subchannel *sch = to_subchannel(cdev->dev.parent);
+
+        CIO_TRACE_EVENT(2, "notoper");
+        CIO_TRACE_EVENT(2, sch->dev.bus_id);
+        ccw_device_set_timeout(cdev, 0);
+        cio_disable_subchannel(sch);
+        cdev->private->state = DEV_STATE_NOT_OPER;
+}
+
 static int io_subchannel_sch_event(struct subchannel *sch, int slow)
 {
         int event, ret, disc;
         unsigned long flags;
-        enum { NONE, UNREGISTER, UNREGISTER_PROBE, REPROBE } action;
+        enum { NONE, UNREGISTER, UNREGISTER_PROBE, REPROBE, DISC } action;
         struct ccw_device *cdev;
 
         spin_lock_irqsave(sch->lock, flags);
@@ -1535,16 +1537,11 @@ static int io_subchannel_sch_event(struct subchannel *sch, int slow)
                 }
                 /* fall through */
         case CIO_GONE:
-                /* Prevent unwanted effects when opening lock. */
-                cio_disable_subchannel(sch);
-                device_set_disconnected(cdev);
                 /* Ask driver what to do with device. */
-                action = UNREGISTER;
-                spin_unlock_irqrestore(sch->lock, flags);
-                ret = io_subchannel_notify(sch, event);
-                spin_lock_irqsave(sch->lock, flags);
-                if (ret)
-                        action = NONE;
+                if (io_subchannel_notify(sch, event))
+                        action = DISC;
+                else
+                        action = UNREGISTER;
                 break;
         case CIO_REVALIDATE:
                 /* Device will be removed, so no notify necessary. */
@@ -1565,6 +1562,7 @@ static int io_subchannel_sch_event(struct subchannel *sch, int slow)
         switch (action) {
         case UNREGISTER:
         case UNREGISTER_PROBE:
+                ccw_device_set_notoper(cdev);
                 /* Unregister device (will use subchannel lock). */
                 spin_unlock_irqrestore(sch->lock, flags);
                 css_sch_device_unregister(sch);
@@ -1577,6 +1575,9 @@ static int io_subchannel_sch_event(struct subchannel *sch, int slow)
         case REPROBE:
                 ccw_device_trigger_reprobe(cdev);
                 break;
+        case DISC:
+                device_set_disconnected(cdev);
+                break;
         default:
                 break;
         }
@@ -1828,5 +1829,4 @@ EXPORT_SYMBOL(ccw_driver_unregister);
 EXPORT_SYMBOL(get_ccwdev_by_busid);
 EXPORT_SYMBOL(ccw_bus_type);
 EXPORT_SYMBOL(ccw_device_work);
-EXPORT_SYMBOL(ccw_device_notify_work);
 EXPORT_SYMBOL_GPL(ccw_device_get_subchannel_id);
diff --git a/drivers/s390/cio/device.h b/drivers/s390/cio/device.h
index 9800a8335a3f..6f5c3f2b3587 100644
--- a/drivers/s390/cio/device.h
+++ b/drivers/s390/cio/device.h
@@ -72,7 +72,6 @@ dev_fsm_final_state(struct ccw_device *cdev)
 }
 
 extern struct workqueue_struct *ccw_device_work;
-extern struct workqueue_struct *ccw_device_notify_work;
 extern wait_queue_head_t ccw_device_init_wq;
 extern atomic_t ccw_device_init_count;
 
@@ -120,6 +119,7 @@ int ccw_device_stlck(struct ccw_device *);
 void ccw_device_trigger_reprobe(struct ccw_device *);
 void ccw_device_kill_io(struct ccw_device *);
 int ccw_device_notify(struct ccw_device *, int);
+void ccw_device_set_notoper(struct ccw_device *cdev);
 
 /* qdio needs this. */
 void ccw_device_set_timeout(struct ccw_device *, int);
diff --git a/drivers/s390/cio/device_fsm.c b/drivers/s390/cio/device_fsm.c
index 8b5fe57fb2f3..550508df952b 100644
--- a/drivers/s390/cio/device_fsm.c
+++ b/drivers/s390/cio/device_fsm.c
@@ -337,26 +337,34 @@ int ccw_device_notify(struct ccw_device *cdev, int event)
                 return 0;
         if (!cdev->online)
                 return 0;
+        CIO_MSG_EVENT(2, "notify called for 0.%x.%04x, event=%d\n",
+                      cdev->private->dev_id.ssid, cdev->private->dev_id.devno,
+                      event);
         return cdev->drv->notify ? cdev->drv->notify(cdev, event) : 0;
 }
 
-static void
-ccw_device_oper_notify(struct work_struct *work)
+static void cmf_reenable_delayed(struct work_struct *work)
 {
         struct ccw_device_private *priv;
         struct ccw_device *cdev;
-        int ret;
 
         priv = container_of(work, struct ccw_device_private, kick_work);
         cdev = priv->cdev;
-        ret = ccw_device_notify(cdev, CIO_OPER);
-        if (ret) {
+        cmf_reenable(cdev);
+}
+
+static void ccw_device_oper_notify(struct ccw_device *cdev)
+{
+        if (ccw_device_notify(cdev, CIO_OPER)) {
                 /* Reenable channel measurements, if needed. */
-                cmf_reenable(cdev);
-                wake_up(&cdev->private->wait_q);
-        } else
-                /* Driver doesn't want device back. */
-                ccw_device_do_unreg_rereg(work);
+                PREPARE_WORK(&cdev->private->kick_work, cmf_reenable_delayed);
+                queue_work(ccw_device_work, &cdev->private->kick_work);
+                return;
+        }
+        /* Driver doesn't want device back. */
+        ccw_device_set_notoper(cdev);
+        PREPARE_WORK(&cdev->private->kick_work, ccw_device_do_unreg_rereg);
+        queue_work(ccw_device_work, &cdev->private->kick_work);
 }
 
 /*
@@ -386,8 +394,7 @@ ccw_device_done(struct ccw_device *cdev, int state)
 
         if (cdev->private->flags.donotify) {
                 cdev->private->flags.donotify = 0;
-                PREPARE_WORK(&cdev->private->kick_work, ccw_device_oper_notify);
-                queue_work(ccw_device_notify_work, &cdev->private->kick_work);
+                ccw_device_oper_notify(cdev);
         }
         wake_up(&cdev->private->wait_q);