powerpc/eeh: Trace eeh device from I/O cache

The idea comes from Benjamin Herrenschmidt. The eeh cache helps fetching the pci device according to the given I/O address. Since the eeh cache is serving for eeh, it's reasonable for eeh cache to trace eeh device except pci device. The patch make eeh cache to trace eeh device. Also, the major eeh entry function eeh_dn_check_failure has been renamed to eeh_dev_check_failure since it will take eeh device as input parameter. Signed-off-by: Gavin Shan <shangw@linux.vnet.ibm.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

powerpc/eeh: Trace eeh device from I/O cache
The idea comes from Benjamin Herrenschmidt. The eeh cache helps fetching the pci device according to the given I/O address. Since the eeh cache is serving for eeh, it's reasonable for eeh cache to trace eeh device except pci device. The patch make eeh cache to trace eeh device. Also, the major eeh entry function eeh_dn_check_failure has been renamed to eeh_dev_check_failure since it will take eeh device as input parameter. Signed-off-by: Gavin Shan <shangw@linux.vnet.ibm.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Gavin Shan · Benjamin Herrenschmidt
1 parent d7bb88629d
Showing 6 changed files with 28 additions and 32 deletions Side-by-side Diff
arch/powerpc/include/asm/eeh.h
arch/powerpc/include/asm/pci-bridge.h
arch/powerpc/include/asm/ppc-pci.h
arch/powerpc/kernel/rtas_pci.c
arch/powerpc/platforms/pseries/eeh.c
arch/powerpc/platforms/pseries/eeh_cache.c
@@ -196,7 +196,7 @@
 int __exit eeh_ops_unregister(const char *name);
 unsigned long eeh_check_failure(const volatile void __iomem *token,
 				unsigned long val);
-int eeh_dn_check_failure(struct device_node *dn, struct pci_dev *dev);
+int eeh_dev_check_failure(struct eeh_dev *edev);
 void __init pci_addr_cache_build(void);
 void eeh_add_device_tree_early(struct device_node *);
 void eeh_add_device_tree_late(struct pci_bus *);
@@ -231,10 +231,7 @@
 	return val;
 }
  
-static inline int eeh_dn_check_failure(struct device_node *dn, struct pci_dev *dev)
-{
-	return 0;
-}
+#define eeh_dev_check_failure(x) (0)
  
 static inline void pci_addr_cache_build(void) { }
  
@@ -184,6 +184,8 @@
 {
 	return PCI_DN(dn)->edev;
 }
+#else
+#define of_node_to_eeh_dev(x) (NULL)
 #endif
  
 /** Find the bus corresponding to the indicated device node */
@@ -50,7 +50,7 @@
 void pci_addr_cache_build(void);
 void pci_addr_cache_insert_device(struct pci_dev *dev);
 void pci_addr_cache_remove_device(struct pci_dev *dev);
-struct pci_dev *pci_addr_cache_get_device(unsigned long addr);
+struct eeh_dev *pci_addr_cache_get_device(unsigned long addr);
 void eeh_slot_error_detail(struct eeh_pe *pe, int severity);
 int eeh_pci_enable(struct eeh_pe *pe, int function);
 int eeh_reset_pe(struct eeh_pe *);
@@ -81,7 +81,7 @@
 		return PCIBIOS_DEVICE_NOT_FOUND;
  
 	if (returnval == EEH_IO_ERROR_VALUE(size) &&
-	    eeh_dn_check_failure (pdn->node, NULL))
+	    eeh_dev_check_failure(of_node_to_eeh_dev(pdn->node)))
 		return PCIBIOS_DEVICE_NOT_FOUND;
  
 	return PCIBIOS_SUCCESSFUL;
@@ -270,9 +270,8 @@
 }
  
 /**
- * eeh_dn_check_failure - Check if all 1's data is due to EEH slot freeze
- * @dn: device node
- * @dev: pci device, if known
+ * eeh_dev_check_failure - Check if all 1's data is due to EEH slot freeze
+ * @edev: eeh device
  *
  * Check for an EEH failure for the given device node.  Call this
  * routine if the result of a read was all 0xff's and you want to
  
  
@@ -284,12 +283,13 @@
  *
  * It is safe to call this routine in an interrupt context.
  */
-int eeh_dn_check_failure(struct device_node *dn, struct pci_dev *dev)
+int eeh_dev_check_failure(struct eeh_dev *edev)
 {
 	int ret;
 	unsigned long flags;
+	struct device_node *dn;
+	struct pci_dev *dev;
 	struct eeh_pe *pe;
-	struct eeh_dev *edev;
 	int rc = 0;
 	const char *location;
  
  
@@ -298,15 +298,12 @@
 	if (!eeh_subsystem_enabled)
 		return 0;
  
-	if (dn) {
-		edev = of_node_to_eeh_dev(dn);
-	} else if (dev) {
-		edev = pci_dev_to_eeh_dev(dev);
-		dn = pci_device_to_OF_node(dev);
-	} else {
+	if (!edev) {
 		eeh_stats.no_dn++;
 		return 0;
 	}
+	dn = eeh_dev_to_of_node(edev);
+	dev = eeh_dev_to_pci_dev(edev);
 	pe = edev->pe;
  
 	/* Access to IO BARs might get this far and still not want checking. */
@@ -393,7 +390,7 @@
 	return rc;
 }
  
-EXPORT_SYMBOL_GPL(eeh_dn_check_failure);
+EXPORT_SYMBOL_GPL(eeh_dev_check_failure);
  
 /**
  * eeh_check_failure - Check if all 1's data is due to EEH slot freeze
  
  
  
@@ -410,21 +407,19 @@
 unsigned long eeh_check_failure(const volatile void __iomem *token, unsigned long val)
 {
 	unsigned long addr;
-	struct pci_dev *dev;
-	struct device_node *dn;
+	struct eeh_dev *edev;
  
 	/* Finding the phys addr + pci device; this is pretty quick. */
 	addr = eeh_token_to_phys((unsigned long __force) token);
-	dev = pci_addr_cache_get_device(addr);
-	if (!dev) {
+	edev = pci_addr_cache_get_device(addr);
+	if (!edev) {
 		eeh_stats.no_device++;
 		return val;
 	}
  
-	dn = pci_device_to_OF_node(dev);
-	eeh_dn_check_failure(dn, dev);
+	eeh_dev_check_failure(edev);
  
-	pci_dev_put(dev);
+	pci_dev_put(eeh_dev_to_pci_dev(edev));
 	return val;
 }
  
@@ -50,6 +50,7 @@
 	struct rb_node rb_node;
 	unsigned long addr_lo;
 	unsigned long addr_hi;
+	struct eeh_dev *edev;
 	struct pci_dev *pcidev;
 	unsigned int flags;
 };
@@ -59,7 +60,7 @@
 	spinlock_t piar_lock;
 } pci_io_addr_cache_root;
  
-static inline struct pci_dev *__pci_addr_cache_get_device(unsigned long addr)
+static inline struct eeh_dev *__pci_addr_cache_get_device(unsigned long addr)
 {
 	struct rb_node *n = pci_io_addr_cache_root.rb_root.rb_node;
  
@@ -74,7 +75,7 @@
 				n = n->rb_right;
 			} else {
 				pci_dev_get(piar->pcidev);
-				return piar->pcidev;
+				return piar->edev;
 			}
 		}
 	}
  
  
  
@@ -92,15 +93,15 @@
  * from zero (that is, they do *not* have pci_io_addr added in).
  * It is safe to call this function within an interrupt.
  */
-struct pci_dev *pci_addr_cache_get_device(unsigned long addr)
+struct eeh_dev *pci_addr_cache_get_device(unsigned long addr)
 {
-	struct pci_dev *dev;
+	struct eeh_dev *edev;
 	unsigned long flags;
  
 	spin_lock_irqsave(&pci_io_addr_cache_root.piar_lock, flags);
-	dev = __pci_addr_cache_get_device(addr);
+	edev = __pci_addr_cache_get_device(addr);
 	spin_unlock_irqrestore(&pci_io_addr_cache_root.piar_lock, flags);
-	return dev;
+	return edev;
 }
  
 #ifdef DEBUG
@@ -158,6 +159,7 @@
 	pci_dev_get(dev);
 	piar->addr_lo = alo;
 	piar->addr_hi = ahi;
+	piar->edev = pci_dev_to_eeh_dev(dev);
 	piar->pcidev = dev;
 	piar->flags = flags;
...	...	@@ -196,7 +196,7 @@
196	196	int __exit eeh_ops_unregister(const char *name);
197	197	unsigned long eeh_check_failure(const volatile void __iomem *token,
198	198	unsigned long val);
199		-int eeh_dn_check_failure(struct device_node dn, struct pci_dev dev);
	199	+int eeh_dev_check_failure(struct eeh_dev *edev);
200	200	void __init pci_addr_cache_build(void);
201	201	void eeh_add_device_tree_early(struct device_node *);
202	202	void eeh_add_device_tree_late(struct pci_bus *);
...	...	@@ -231,10 +231,7 @@
231	231	return val;
232	232	}
233	233
234		-static inline int eeh_dn_check_failure(struct device_node dn, struct pci_dev dev)
235		-{
236		- return 0;
237		-}
	234	+#define eeh_dev_check_failure(x) (0)
238	235
239	236	static inline void pci_addr_cache_build(void) { }
240	237
...	...	@@ -184,6 +184,8 @@
184	184	{
185	185	return PCI_DN(dn)->edev;
186	186	}
	187	+#else
	188	+#define of_node_to_eeh_dev(x) (NULL)
187	189	#endif
188	190
189	191	/** Find the bus corresponding to the indicated device node */
...	...	@@ -50,7 +50,7 @@
50	50	void pci_addr_cache_build(void);
51	51	void pci_addr_cache_insert_device(struct pci_dev *dev);
52	52	void pci_addr_cache_remove_device(struct pci_dev *dev);
53		-struct pci_dev *pci_addr_cache_get_device(unsigned long addr);
	53	+struct eeh_dev *pci_addr_cache_get_device(unsigned long addr);
54	54	void eeh_slot_error_detail(struct eeh_pe *pe, int severity);
55	55	int eeh_pci_enable(struct eeh_pe *pe, int function);
56	56	int eeh_reset_pe(struct eeh_pe *);
...	...	@@ -81,7 +81,7 @@
81	81	return PCIBIOS_DEVICE_NOT_FOUND;
82	82
83	83	if (returnval == EEH_IO_ERROR_VALUE(size) &&
84		- eeh_dn_check_failure (pdn->node, NULL))
	84	+ eeh_dev_check_failure(of_node_to_eeh_dev(pdn->node)))
85	85	return PCIBIOS_DEVICE_NOT_FOUND;
86	86
87	87	return PCIBIOS_SUCCESSFUL;
...	...	@@ -270,9 +270,8 @@
270	270	}
271	271
272	272	/**
273		- * eeh_dn_check_failure - Check if all 1's data is due to EEH slot freeze
274		- * @dn: device node
275		- * @dev: pci device, if known
	273	+ * eeh_dev_check_failure - Check if all 1's data is due to EEH slot freeze
	274	+ * @edev: eeh device
276	275	*
277	276	* Check for an EEH failure for the given device node. Call this
278	277	* routine if the result of a read was all 0xff's and you want to
279	278
280	279
...	...	@@ -284,12 +283,13 @@
284	283	*
285	284	* It is safe to call this routine in an interrupt context.
286	285	*/
287		-int eeh_dn_check_failure(struct device_node dn, struct pci_dev dev)
	286	+int eeh_dev_check_failure(struct eeh_dev *edev)
288	287	{
289	288	int ret;
290	289	unsigned long flags;
	290	+ struct device_node *dn;
	291	+ struct pci_dev *dev;
291	292	struct eeh_pe *pe;
292		- struct eeh_dev *edev;
293	293	int rc = 0;
294	294	const char *location;
295	295
296	296
...	...	@@ -298,15 +298,12 @@
298	298	if (!eeh_subsystem_enabled)
299	299	return 0;
300	300
301		- if (dn) {
302		- edev = of_node_to_eeh_dev(dn);
303		- } else if (dev) {
304		- edev = pci_dev_to_eeh_dev(dev);
305		- dn = pci_device_to_OF_node(dev);
306		- } else {
	301	+ if (!edev) {
307	302	eeh_stats.no_dn++;
308	303	return 0;
309	304	}
	305	+ dn = eeh_dev_to_of_node(edev);
	306	+ dev = eeh_dev_to_pci_dev(edev);
310	307	pe = edev->pe;
311	308
312	309	/* Access to IO BARs might get this far and still not want checking. */
...	...	@@ -393,7 +390,7 @@
393	390	return rc;
394	391	}
395	392
396		-EXPORT_SYMBOL_GPL(eeh_dn_check_failure);
	393	+EXPORT_SYMBOL_GPL(eeh_dev_check_failure);
397	394
398	395	/**
399	396	* eeh_check_failure - Check if all 1's data is due to EEH slot freeze
400	397
401	398
402	399
...	...	@@ -410,21 +407,19 @@
410	407	unsigned long eeh_check_failure(const volatile void __iomem *token, unsigned long val)
411	408	{
412	409	unsigned long addr;
413		- struct pci_dev *dev;
414		- struct device_node *dn;
	410	+ struct eeh_dev *edev;
415	411
416	412	/* Finding the phys addr + pci device; this is pretty quick. */
417	413	addr = eeh_token_to_phys((unsigned long __force) token);
418		- dev = pci_addr_cache_get_device(addr);
419		- if (!dev) {
	414	+ edev = pci_addr_cache_get_device(addr);
	415	+ if (!edev) {
420	416	eeh_stats.no_device++;
421	417	return val;
422	418	}
423	419
424		- dn = pci_device_to_OF_node(dev);
425		- eeh_dn_check_failure(dn, dev);
	420	+ eeh_dev_check_failure(edev);
426	421
427		- pci_dev_put(dev);
	422	+ pci_dev_put(eeh_dev_to_pci_dev(edev));
428	423	return val;
429	424	}
430	425
...	...	@@ -50,6 +50,7 @@
50	50	struct rb_node rb_node;
51	51	unsigned long addr_lo;
52	52	unsigned long addr_hi;
	53	+ struct eeh_dev *edev;
53	54	struct pci_dev *pcidev;
54	55	unsigned int flags;
55	56	};
...	...	@@ -59,7 +60,7 @@
59	60	spinlock_t piar_lock;
60	61	} pci_io_addr_cache_root;
61	62
62		-static inline struct pci_dev *__pci_addr_cache_get_device(unsigned long addr)
	63	+static inline struct eeh_dev *__pci_addr_cache_get_device(unsigned long addr)
63	64	{
64	65	struct rb_node *n = pci_io_addr_cache_root.rb_root.rb_node;
65	66
...	...	@@ -74,7 +75,7 @@
74	75	n = n->rb_right;
75	76	} else {
76	77	pci_dev_get(piar->pcidev);
77		- return piar->pcidev;
	78	+ return piar->edev;
78	79	}
79	80	}
80	81	}
81	82
82	83
83	84
...	...	@@ -92,15 +93,15 @@
92	93	* from zero (that is, they do not have pci_io_addr added in).
93	94	* It is safe to call this function within an interrupt.
94	95	*/
95		-struct pci_dev *pci_addr_cache_get_device(unsigned long addr)
	96	+struct eeh_dev *pci_addr_cache_get_device(unsigned long addr)
96	97	{
97		- struct pci_dev *dev;
	98	+ struct eeh_dev *edev;
98	99	unsigned long flags;
99	100
100	101	spin_lock_irqsave(&pci_io_addr_cache_root.piar_lock, flags);
101		- dev = __pci_addr_cache_get_device(addr);
	102	+ edev = __pci_addr_cache_get_device(addr);
102	103	spin_unlock_irqrestore(&pci_io_addr_cache_root.piar_lock, flags);
103		- return dev;
	104	+ return edev;
104	105	}
105	106
106	107	#ifdef DEBUG
...	...	@@ -158,6 +159,7 @@
158	159	pci_dev_get(dev);
159	160	piar->addr_lo = alo;
160	161	piar->addr_hi = ahi;
	162	+ piar->edev = pci_dev_to_eeh_dev(dev);
161	163	piar->pcidev = dev;
162	164	piar->flags = flags;
163	165