[people/pmueller/ipfire-2.x.git] / src / patches / suse-2.6.27.39 / patches.arch / s390-12-05-cio_introduce_ccw_device_boxed_notify.patch

From: Gerald Schaefer <geraldsc@de.ibm.com>
Subject: cio: notify ccw driver for boxed devices
References: bnc#487755,LTC#52560

Symptom:     Dasd driver keeps trying to start io on not accessible device.
Problem:     Missing notification for ccw device drivers.
Solution:    Add proper notification if a device goes into boxed state.

Acked-by: John Jolly <jjolly@suse.de>
---
 Documentation/kmsg/s390/zfcp  |   15 +++++++++++++++
 arch/s390/include/asm/cio.h   |    2 ++
 drivers/s390/block/dasd.c     |    1 +
 drivers/s390/cio/device.c     |   26 +++++++++++++++++---------
 drivers/s390/cio/device.h     |    1 +
 drivers/s390/cio/device_fsm.c |   29 ++++++++++++++++++-----------
 drivers/s390/scsi/zfcp_ccw.c  |    5 +++++
 7 files changed, 59 insertions(+), 20 deletions(-)

--- a/Documentation/kmsg/s390/zfcp
+++ b/Documentation/kmsg/s390/zfcp
@@ -100,6 +100,21 @@
  */
 
 /*?
+ * Text: "%s: The FCP device did not respond within the specified time\n"
+ * Severity: Warning
+ * Parameter:
+ *   @1: bus ID of the zfcp device
+ * Description:
+ * The common I/O layer waited for a response from the FCP adapter but
+ * no response was received within the specified time limit. This might
+ * indicate a hardware problem.
+ * User action:
+ * Consult your hardware administrator. If this problem persists,
+ * gather Linux debug data, collect the FCP adapter hardware logs, and
+ * report the problem to your support organization.
+ */
+
+/*?
  * Text: "%s: Registering the FCP device with the SCSI stack failed\n"
  * Severity: Error
  * Parameter:
--- a/arch/s390/include/asm/cio.h
+++ b/arch/s390/include/asm/cio.h
@@ -456,6 +456,8 @@ struct ciw {
 #define CIO_OPER       0x0004
 /* Sick revalidation of device. */
 #define CIO_REVALIDATE 0x0008
+/* Device did not respond in time. */
+#define CIO_BOXED      0x0010
 
 /**
  * struct ccw_dev_id - unique identifier for ccw devices
--- a/drivers/s390/block/dasd.c
+++ b/drivers/s390/block/dasd.c
@@ -2359,6 +2359,7 @@ int dasd_generic_notify(struct ccw_devic
 	ret = 0;
 	switch (event) {
 	case CIO_GONE:
+	case CIO_BOXED:
 	case CIO_NO_PATH:
 		/* First of all call extended error reporting. */
 		dasd_eer_write(device, NULL, DASD_EER_NOPATH);
--- a/drivers/s390/cio/device.c
+++ b/drivers/s390/cio/device.c
@@ -479,11 +479,15 @@ static int online_store_recog_and_online
 		}
 		wait_event(cdev->private->wait_q,
 			   cdev->private->flags.recog_done);
+		if (cdev->private->state != DEV_STATE_OFFLINE)
+			/* recognition failed */
+			return -EAGAIN;
 	}
 	if (cdev->drv && cdev->drv->set_online)
 		ccw_device_set_online(cdev);
 	return 0;
 }
+
 static int online_store_handle_online(struct ccw_device *cdev, int force)
 {
 	int ret;
@@ -497,7 +501,9 @@ static int online_store_handle_online(st
 			return ret;
 		if (cdev->id.cu_type == 0)
 			cdev->private->state = DEV_STATE_NOT_OPER;
-		online_store_recog_and_online(cdev);
+		ret = online_store_recog_and_online(cdev);
+		if (ret)
+			return ret;
 	}
 	return 0;
 }
@@ -1017,33 +1023,35 @@ static void ccw_device_call_sch_unregist
 	put_device(&sch->dev);
 }
 
+void ccw_device_schedule_sch_unregister(struct ccw_device *cdev)
+{
+	PREPARE_WORK(&cdev->private->kick_work,
+		     ccw_device_call_sch_unregister);
+	queue_work(slow_path_wq, &cdev->private->kick_work);
+}
+
 /*
  * subchannel recognition done. Called from the state machine.
  */
 void
 io_subchannel_recog_done(struct ccw_device *cdev)
 {
-	struct subchannel *sch;
-
 	if (css_init_done == 0) {
 		cdev->private->flags.recog_done = 1;
 		return;
 	}
 	switch (cdev->private->state) {
+	case DEV_STATE_BOXED:
+		/* Device did not respond in time. */
 	case DEV_STATE_NOT_OPER:
 		cdev->private->flags.recog_done = 1;
 		/* Remove device found not operational. */
 		if (!get_device(&cdev->dev))
 			break;
-		sch = to_subchannel(cdev->dev.parent);
-		PREPARE_WORK(&cdev->private->kick_work,
-			     ccw_device_call_sch_unregister);
-		queue_work(slow_path_wq, &cdev->private->kick_work);
+		ccw_device_schedule_sch_unregister(cdev);
 		if (atomic_dec_and_test(&ccw_device_init_count))
 			wake_up(&ccw_device_init_wq);
 		break;
-	case DEV_STATE_BOXED:
-		/* Device did not respond in time. */
 	case DEV_STATE_OFFLINE:
 		/* 
 		 * We can't register the device in interrupt context so
--- a/drivers/s390/cio/device.h
+++ b/drivers/s390/cio/device.h
@@ -86,6 +86,7 @@ int ccw_device_is_orphan(struct ccw_devi
 int ccw_device_recognition(struct ccw_device *);
 int ccw_device_online(struct ccw_device *);
 int ccw_device_offline(struct ccw_device *);
+void ccw_device_schedule_sch_unregister(struct ccw_device *);
 
 /* Function prototypes for device status and basic sense stuff. */
 void ccw_device_accumulate_irb(struct ccw_device *, struct irb *);
--- a/drivers/s390/cio/device_fsm.c
+++ b/drivers/s390/cio/device_fsm.c
@@ -253,13 +253,12 @@ ccw_device_recog_done(struct ccw_device
 		old_lpm = 0;
 	if (sch->lpm != old_lpm)
 		__recover_lost_chpids(sch, old_lpm);
-	if (cdev->private->state == DEV_STATE_DISCONNECTED_SENSE_ID) {
-		if (state == DEV_STATE_NOT_OPER) {
-			cdev->private->flags.recog_done = 1;
-			cdev->private->state = DEV_STATE_DISCONNECTED;
-			return;
-		}
-		/* Boxed devices don't need extra treatment. */
+	if (cdev->private->state == DEV_STATE_DISCONNECTED_SENSE_ID &&
+	    (state == DEV_STATE_NOT_OPER || state == DEV_STATE_BOXED)) {
+		cdev->private->flags.recog_done = 1;
+		cdev->private->state = DEV_STATE_DISCONNECTED;
+		wake_up(&cdev->private->wait_q);
+		return;
 	}
 	notify = 0;
 	same_dev = 0; /* Keep the compiler quiet... */
@@ -304,12 +303,17 @@ ccw_device_recog_done(struct ccw_device
 			      " subchannel 0.%x.%04x\n",
 			      cdev->private->dev_id.devno,
 			      sch->schid.ssid, sch->schid.sch_no);
+		if (cdev->id.cu_type != 0) { /* device was recognized before */
+			cdev->private->flags.recog_done = 1;
+			cdev->private->state = DEV_STATE_BOXED;
+			wake_up(&cdev->private->wait_q);
+			return;
+		}
 		break;
 	}
 	cdev->private->state = state;
 	io_subchannel_recog_done(cdev);
-	if (state != DEV_STATE_NOT_OPER)
-		wake_up(&cdev->private->wait_q);
+	wake_up(&cdev->private->wait_q);
 }
 
 /*
@@ -387,10 +391,13 @@ ccw_device_done(struct ccw_device *cdev,
 
 	cdev->private->state = state;
 
-
-	if (state == DEV_STATE_BOXED)
+	if (state == DEV_STATE_BOXED) {
 		CIO_MSG_EVENT(0, "Boxed device %04x on subchannel %04x\n",
 			      cdev->private->dev_id.devno, sch->schid.sch_no);
+		if (cdev->online && !ccw_device_notify(cdev, CIO_BOXED))
+			ccw_device_schedule_sch_unregister(cdev);
+		cdev->private->flags.donotify = 0;
+	}
 
 	if (cdev->private->flags.donotify) {
 		cdev->private->flags.donotify = 0;
--- a/drivers/s390/scsi/zfcp_ccw.c
+++ b/drivers/s390/scsi/zfcp_ccw.c
@@ -181,6 +181,11 @@ static int zfcp_ccw_notify(struct ccw_de
 		zfcp_erp_adapter_reopen(adapter, ZFCP_STATUS_COMMON_ERP_FAILED,
 					89, NULL);
 		break;
+	case CIO_BOXED:
+		dev_warn(&adapter->ccw_device->dev, "The FCP device "
+			 "did not respond within the specified time\n");
+		zfcp_erp_adapter_shutdown(adapter, 0, 91, NULL);
+		break;
 	}
 	return 1;
 }
Commit	Line	Data
2cb7cef9 BS	1	From: Gerald Schaefer <geraldsc@de.ibm.com>
	2	Subject: cio: notify ccw driver for boxed devices
	3	References: bnc#487755,LTC#52560
	4
	5	Symptom: Dasd driver keeps trying to start io on not accessible device.
	6	Problem: Missing notification for ccw device drivers.
	7	Solution: Add proper notification if a device goes into boxed state.
	8
	9	Acked-by: John Jolly <jjolly@suse.de>
	10	---
	11	Documentation/kmsg/s390/zfcp \| 15 +++++++++++++++
	12	arch/s390/include/asm/cio.h \| 2 ++
	13	drivers/s390/block/dasd.c \| 1 +
	14	drivers/s390/cio/device.c \| 26 +++++++++++++++++---------
	15	drivers/s390/cio/device.h \| 1 +
	16	drivers/s390/cio/device_fsm.c \| 29 ++++++++++++++++++-----------
	17	drivers/s390/scsi/zfcp_ccw.c \| 5 +++++
	18	7 files changed, 59 insertions(+), 20 deletions(-)
	19
	20	--- a/Documentation/kmsg/s390/zfcp
	21	+++ b/Documentation/kmsg/s390/zfcp
	22	@@ -100,6 +100,21 @@
	23	*/
	24
	25	/*?
	26	+ * Text: "%s: The FCP device did not respond within the specified time\n"
	27	+ * Severity: Warning
	28	+ * Parameter:
	29	+ * @1: bus ID of the zfcp device
	30	+ * Description:
	31	+ * The common I/O layer waited for a response from the FCP adapter but
	32	+ * no response was received within the specified time limit. This might
	33	+ * indicate a hardware problem.
	34	+ * User action:
	35	+ * Consult your hardware administrator. If this problem persists,
	36	+ * gather Linux debug data, collect the FCP adapter hardware logs, and
	37	+ * report the problem to your support organization.
	38	+ */
	39	+
	40	+/*?
	41	* Text: "%s: Registering the FCP device with the SCSI stack failed\n"
	42	* Severity: Error
	43	* Parameter:
	44	--- a/arch/s390/include/asm/cio.h
	45	+++ b/arch/s390/include/asm/cio.h
	46	@@ -456,6 +456,8 @@ struct ciw {
	47	#define CIO_OPER 0x0004
	48	/* Sick revalidation of device. */
	49	#define CIO_REVALIDATE 0x0008
	50	+/* Device did not respond in time. */
	51	+#define CIO_BOXED 0x0010
	52
	53	/**
	54	* struct ccw_dev_id - unique identifier for ccw devices
	55	--- a/drivers/s390/block/dasd.c
	56	+++ b/drivers/s390/block/dasd.c
	57	@@ -2359,6 +2359,7 @@ int dasd_generic_notify(struct ccw_devic
	58	ret = 0;
	59	switch (event) {
	60	case CIO_GONE:
	61	+ case CIO_BOXED:
	62	case CIO_NO_PATH:
	63	/* First of all call extended error reporting. */
	64	dasd_eer_write(device, NULL, DASD_EER_NOPATH);
65	--- a/drivers/s390/cio/device.c
66	+++ b/drivers/s390/cio/device.c
67	@@ -479,11 +479,15 @@ static int online_store_recog_and_online
68	}
69	wait_event(cdev->private->wait_q,
70	cdev->private->flags.recog_done);
71	+ if (cdev->private->state != DEV_STATE_OFFLINE)
72	+ /* recognition failed */
73	+ return -EAGAIN;
74	}
75	if (cdev->drv && cdev->drv->set_online)
76	ccw_device_set_online(cdev);
77	return 0;
78	}
79	+
80	static int online_store_handle_online(struct ccw_device *cdev, int force)
81	{
82	int ret;
83	@@ -497,7 +501,9 @@ static int online_store_handle_online(st
84	return ret;
85	if (cdev->id.cu_type == 0)
86	cdev->private->state = DEV_STATE_NOT_OPER;
87	- online_store_recog_and_online(cdev);
88	+ ret = online_store_recog_and_online(cdev);
89	+ if (ret)
90	+ return ret;
91	}
92	return 0;
93	}
94	@@ -1017,33 +1023,35 @@ static void ccw_device_call_sch_unregist
95	put_device(&sch->dev);
96	}
97
98	+void ccw_device_schedule_sch_unregister(struct ccw_device *cdev)
99	+{
100	+ PREPARE_WORK(&cdev->private->kick_work,
101	+ ccw_device_call_sch_unregister);
102	+ queue_work(slow_path_wq, &cdev->private->kick_work);
103	+}
104	+
105	/*
106	* subchannel recognition done. Called from the state machine.
107	*/
108	void
109	io_subchannel_recog_done(struct ccw_device *cdev)
110	{
111	- struct subchannel *sch;
112	-
113	if (css_init_done == 0) {
114	cdev->private->flags.recog_done = 1;
115	return;
116	}
117	switch (cdev->private->state) {
118	+ case DEV_STATE_BOXED:
119	+ /* Device did not respond in time. */
120	case DEV_STATE_NOT_OPER:
121	cdev->private->flags.recog_done = 1;
122	/* Remove device found not operational. */
123	if (!get_device(&cdev->dev))
124	break;
125	- sch = to_subchannel(cdev->dev.parent);
126	- PREPARE_WORK(&cdev->private->kick_work,
127	- ccw_device_call_sch_unregister);
128	- queue_work(slow_path_wq, &cdev->private->kick_work);
129	+ ccw_device_schedule_sch_unregister(cdev);
130	if (atomic_dec_and_test(&ccw_device_init_count))
131	wake_up(&ccw_device_init_wq);
132	break;
133	- case DEV_STATE_BOXED:
134	- /* Device did not respond in time. */
135	case DEV_STATE_OFFLINE:
136	/*
137	* We can't register the device in interrupt context so
138	--- a/drivers/s390/cio/device.h
139	+++ b/drivers/s390/cio/device.h
140	@@ -86,6 +86,7 @@ int ccw_device_is_orphan(struct ccw_devi
141	int ccw_device_recognition(struct ccw_device *);
142	int ccw_device_online(struct ccw_device *);
143	int ccw_device_offline(struct ccw_device *);
144	+void ccw_device_schedule_sch_unregister(struct ccw_device *);
145
146	/* Function prototypes for device status and basic sense stuff. */
147	void ccw_device_accumulate_irb(struct ccw_device , struct irb );
148	--- a/drivers/s390/cio/device_fsm.c
149	+++ b/drivers/s390/cio/device_fsm.c
150	@@ -253,13 +253,12 @@ ccw_device_recog_done(struct ccw_device
151	old_lpm = 0;
152	if (sch->lpm != old_lpm)
153	__recover_lost_chpids(sch, old_lpm);
154	- if (cdev->private->state == DEV_STATE_DISCONNECTED_SENSE_ID) {
155	- if (state == DEV_STATE_NOT_OPER) {
156	- cdev->private->flags.recog_done = 1;
157	- cdev->private->state = DEV_STATE_DISCONNECTED;
158	- return;
159	- }
160	- /* Boxed devices don't need extra treatment. */
161	+ if (cdev->private->state == DEV_STATE_DISCONNECTED_SENSE_ID &&
162	+ (state == DEV_STATE_NOT_OPER \|\| state == DEV_STATE_BOXED)) {
163	+ cdev->private->flags.recog_done = 1;
164	+ cdev->private->state = DEV_STATE_DISCONNECTED;
165	+ wake_up(&cdev->private->wait_q);
166	+ return;
167	}
168	notify = 0;
169	same_dev = 0; /* Keep the compiler quiet... */
170	@@ -304,12 +303,17 @@ ccw_device_recog_done(struct ccw_device
171	" subchannel 0.%x.%04x\n",
172	cdev->private->dev_id.devno,
173	sch->schid.ssid, sch->schid.sch_no);
174	+ if (cdev->id.cu_type != 0) { /* device was recognized before */
175	+ cdev->private->flags.recog_done = 1;
176	+ cdev->private->state = DEV_STATE_BOXED;
177	+ wake_up(&cdev->private->wait_q);
178	+ return;
179	+ }
180	break;
181	}
182	cdev->private->state = state;
183	io_subchannel_recog_done(cdev);
184	- if (state != DEV_STATE_NOT_OPER)
185	- wake_up(&cdev->private->wait_q);
186	+ wake_up(&cdev->private->wait_q);
187	}
188
189	/*
190	@@ -387,10 +391,13 @@ ccw_device_done(struct ccw_device *cdev,
191
192	cdev->private->state = state;
193
194	-
195	- if (state == DEV_STATE_BOXED)
196	+ if (state == DEV_STATE_BOXED) {
197	CIO_MSG_EVENT(0, "Boxed device %04x on subchannel %04x\n",
198	cdev->private->dev_id.devno, sch->schid.sch_no);
199	+ if (cdev->online && !ccw_device_notify(cdev, CIO_BOXED))
200	+ ccw_device_schedule_sch_unregister(cdev);
201	+ cdev->private->flags.donotify = 0;
202	+ }
203
204	if (cdev->private->flags.donotify) {
205	cdev->private->flags.donotify = 0;
206	--- a/drivers/s390/scsi/zfcp_ccw.c
207	+++ b/drivers/s390/scsi/zfcp_ccw.c
208	@@ -181,6 +181,11 @@ static int zfcp_ccw_notify(struct ccw_de
209	zfcp_erp_adapter_reopen(adapter, ZFCP_STATUS_COMMON_ERP_FAILED,
210	89, NULL);
211	break;
212	+ case CIO_BOXED:
213	+ dev_warn(&adapter->ccw_device->dev, "The FCP device "
214	+ "did not respond within the specified time\n");
215	+ zfcp_erp_adapter_shutdown(adapter, 0, 91, NULL);
216	+ break;
217	}
218	return 1;
219	}