xref: /openbmc/linux/arch/powerpc/include/asm/eeh.h (revision 483eb062)
1 /*
2  * Copyright (C) 2001  Dave Engebretsen & Todd Inglett IBM Corporation.
3  * Copyright 2001-2012 IBM Corporation.
4  *
5  * This program is free software; you can redistribute it and/or modify
6  * it under the terms of the GNU General Public License as published by
7  * the Free Software Foundation; either version 2 of the License, or
8  * (at your option) any later version.
9  *
10  * This program is distributed in the hope that it will be useful,
11  * but WITHOUT ANY WARRANTY; without even the implied warranty of
12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
13  * GNU General Public License for more details.
14  *
15  * You should have received a copy of the GNU General Public License
16  * along with this program; if not, write to the Free Software
17  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
18  */
19 
20 #ifndef _POWERPC_EEH_H
21 #define _POWERPC_EEH_H
22 #ifdef __KERNEL__
23 
24 #include <linux/init.h>
25 #include <linux/list.h>
26 #include <linux/string.h>
27 #include <linux/time.h>
28 
29 struct pci_dev;
30 struct pci_bus;
31 struct device_node;
32 
33 #ifdef CONFIG_EEH
34 
35 /*
36  * The struct is used to trace PE related EEH functionality.
37  * In theory, there will have one instance of the struct to
38  * be created against particular PE. In nature, PEs corelate
39  * to each other. the struct has to reflect that hierarchy in
40  * order to easily pick up those affected PEs when one particular
41  * PE has EEH errors.
42  *
43  * Also, one particular PE might be composed of PCI device, PCI
44  * bus and its subordinate components. The struct also need ship
45  * the information. Further more, one particular PE is only meaingful
46  * in the corresponding PHB. Therefore, the root PEs should be created
47  * against existing PHBs in on-to-one fashion.
48  */
49 #define EEH_PE_INVALID	(1 << 0)	/* Invalid   */
50 #define EEH_PE_PHB	(1 << 1)	/* PHB PE    */
51 #define EEH_PE_DEVICE 	(1 << 2)	/* Device PE */
52 #define EEH_PE_BUS	(1 << 3)	/* Bus PE    */
53 
54 #define EEH_PE_ISOLATED		(1 << 0)	/* Isolated PE		*/
55 #define EEH_PE_RECOVERING	(1 << 1)	/* Recovering PE	*/
56 #define EEH_PE_PHB_DEAD		(1 << 2)	/* Dead PHB		*/
57 
58 #define EEH_PE_KEEP		(1 << 8)	/* Keep PE on hotplug	*/
59 
60 struct eeh_pe {
61 	int type;			/* PE type: PHB/Bus/Device	*/
62 	int state;			/* PE EEH dependent mode	*/
63 	int config_addr;		/* Traditional PCI address	*/
64 	int addr;			/* PE configuration address	*/
65 	struct pci_controller *phb;	/* Associated PHB		*/
66 	struct pci_bus *bus;		/* Top PCI bus for bus PE	*/
67 	int check_count;		/* Times of ignored error	*/
68 	int freeze_count;		/* Times of froze up		*/
69 	struct timeval tstamp;		/* Time on first-time freeze	*/
70 	int false_positives;		/* Times of reported #ff's	*/
71 	struct eeh_pe *parent;		/* Parent PE			*/
72 	struct list_head child_list;	/* Link PE to the child list	*/
73 	struct list_head edevs;		/* Link list of EEH devices	*/
74 	struct list_head child;		/* Child PEs			*/
75 };
76 
77 #define eeh_pe_for_each_dev(pe, edev, tmp) \
78 		list_for_each_entry_safe(edev, tmp, &pe->edevs, list)
79 
80 /*
81  * The struct is used to trace EEH state for the associated
82  * PCI device node or PCI device. In future, it might
83  * represent PE as well so that the EEH device to form
84  * another tree except the currently existing tree of PCI
85  * buses and PCI devices
86  */
87 #define EEH_DEV_BRIDGE		(1 << 0)	/* PCI bridge		*/
88 #define EEH_DEV_ROOT_PORT	(1 << 1)	/* PCIe root port	*/
89 #define EEH_DEV_DS_PORT		(1 << 2)	/* Downstream port	*/
90 #define EEH_DEV_IRQ_DISABLED	(1 << 3)	/* Interrupt disabled	*/
91 #define EEH_DEV_DISCONNECTED	(1 << 4)	/* Removing from PE	*/
92 
93 #define EEH_DEV_NO_HANDLER	(1 << 8)	/* No error handler	*/
94 #define EEH_DEV_SYSFS		(1 << 9)	/* Sysfs created	*/
95 
96 struct eeh_dev {
97 	int mode;			/* EEH mode			*/
98 	int class_code;			/* Class code of the device	*/
99 	int config_addr;		/* Config address		*/
100 	int pe_config_addr;		/* PE config address		*/
101 	u32 config_space[16];		/* Saved PCI config space	*/
102 	u8 pcie_cap;			/* Saved PCIe capability	*/
103 	struct eeh_pe *pe;		/* Associated PE		*/
104 	struct list_head list;		/* Form link list in the PE	*/
105 	struct pci_controller *phb;	/* Associated PHB		*/
106 	struct device_node *dn;		/* Associated device node	*/
107 	struct pci_dev *pdev;		/* Associated PCI device	*/
108 	struct pci_bus *bus;		/* PCI bus for partial hotplug	*/
109 };
110 
111 static inline struct device_node *eeh_dev_to_of_node(struct eeh_dev *edev)
112 {
113 	return edev ? edev->dn : NULL;
114 }
115 
116 static inline struct pci_dev *eeh_dev_to_pci_dev(struct eeh_dev *edev)
117 {
118 	return edev ? edev->pdev : NULL;
119 }
120 
121 /* Return values from eeh_ops::next_error */
122 enum {
123 	EEH_NEXT_ERR_NONE = 0,
124 	EEH_NEXT_ERR_INF,
125 	EEH_NEXT_ERR_FROZEN_PE,
126 	EEH_NEXT_ERR_FENCED_PHB,
127 	EEH_NEXT_ERR_DEAD_PHB,
128 	EEH_NEXT_ERR_DEAD_IOC
129 };
130 
131 /*
132  * The struct is used to trace the registered EEH operation
133  * callback functions. Actually, those operation callback
134  * functions are heavily platform dependent. That means the
135  * platform should register its own EEH operation callback
136  * functions before any EEH further operations.
137  */
138 #define EEH_OPT_DISABLE		0	/* EEH disable	*/
139 #define EEH_OPT_ENABLE		1	/* EEH enable	*/
140 #define EEH_OPT_THAW_MMIO	2	/* MMIO enable	*/
141 #define EEH_OPT_THAW_DMA	3	/* DMA enable	*/
142 #define EEH_STATE_UNAVAILABLE	(1 << 0)	/* State unavailable	*/
143 #define EEH_STATE_NOT_SUPPORT	(1 << 1)	/* EEH not supported	*/
144 #define EEH_STATE_RESET_ACTIVE	(1 << 2)	/* Active reset		*/
145 #define EEH_STATE_MMIO_ACTIVE	(1 << 3)	/* Active MMIO		*/
146 #define EEH_STATE_DMA_ACTIVE	(1 << 4)	/* Active DMA		*/
147 #define EEH_STATE_MMIO_ENABLED	(1 << 5)	/* MMIO enabled		*/
148 #define EEH_STATE_DMA_ENABLED	(1 << 6)	/* DMA enabled		*/
149 #define EEH_RESET_DEACTIVATE	0	/* Deactivate the PE reset	*/
150 #define EEH_RESET_HOT		1	/* Hot reset			*/
151 #define EEH_RESET_FUNDAMENTAL	3	/* Fundamental reset		*/
152 #define EEH_LOG_TEMP		1	/* EEH temporary error log	*/
153 #define EEH_LOG_PERM		2	/* EEH permanent error log	*/
154 
155 struct eeh_ops {
156 	char *name;
157 	int (*init)(void);
158 	int (*post_init)(void);
159 	void* (*of_probe)(struct device_node *dn, void *flag);
160 	int (*dev_probe)(struct pci_dev *dev, void *flag);
161 	int (*set_option)(struct eeh_pe *pe, int option);
162 	int (*get_pe_addr)(struct eeh_pe *pe);
163 	int (*get_state)(struct eeh_pe *pe, int *state);
164 	int (*reset)(struct eeh_pe *pe, int option);
165 	int (*wait_state)(struct eeh_pe *pe, int max_wait);
166 	int (*get_log)(struct eeh_pe *pe, int severity, char *drv_log, unsigned long len);
167 	int (*configure_bridge)(struct eeh_pe *pe);
168 	int (*read_config)(struct device_node *dn, int where, int size, u32 *val);
169 	int (*write_config)(struct device_node *dn, int where, int size, u32 val);
170 	int (*next_error)(struct eeh_pe **pe);
171 	int (*restore_config)(struct device_node *dn);
172 };
173 
174 extern struct eeh_ops *eeh_ops;
175 extern int eeh_subsystem_enabled;
176 extern raw_spinlock_t confirm_error_lock;
177 extern int eeh_probe_mode;
178 
179 #define EEH_PROBE_MODE_DEV	(1<<0)	/* From PCI device	*/
180 #define EEH_PROBE_MODE_DEVTREE	(1<<1)	/* From device tree	*/
181 
182 static inline void eeh_probe_mode_set(int flag)
183 {
184 	eeh_probe_mode = flag;
185 }
186 
187 static inline int eeh_probe_mode_devtree(void)
188 {
189 	return (eeh_probe_mode == EEH_PROBE_MODE_DEVTREE);
190 }
191 
192 static inline int eeh_probe_mode_dev(void)
193 {
194 	return (eeh_probe_mode == EEH_PROBE_MODE_DEV);
195 }
196 
197 static inline void eeh_serialize_lock(unsigned long *flags)
198 {
199 	raw_spin_lock_irqsave(&confirm_error_lock, *flags);
200 }
201 
202 static inline void eeh_serialize_unlock(unsigned long flags)
203 {
204 	raw_spin_unlock_irqrestore(&confirm_error_lock, flags);
205 }
206 
207 /*
208  * Max number of EEH freezes allowed before we consider the device
209  * to be permanently disabled.
210  */
211 #define EEH_MAX_ALLOWED_FREEZES 5
212 
213 typedef void *(*eeh_traverse_func)(void *data, void *flag);
214 int eeh_phb_pe_create(struct pci_controller *phb);
215 struct eeh_pe *eeh_phb_pe_get(struct pci_controller *phb);
216 struct eeh_pe *eeh_pe_get(struct eeh_dev *edev);
217 int eeh_add_to_parent_pe(struct eeh_dev *edev);
218 int eeh_rmv_from_parent_pe(struct eeh_dev *edev);
219 void eeh_pe_update_time_stamp(struct eeh_pe *pe);
220 void *eeh_pe_traverse(struct eeh_pe *root,
221 		eeh_traverse_func fn, void *flag);
222 void *eeh_pe_dev_traverse(struct eeh_pe *root,
223 		eeh_traverse_func fn, void *flag);
224 void eeh_pe_restore_bars(struct eeh_pe *pe);
225 struct pci_bus *eeh_pe_bus_get(struct eeh_pe *pe);
226 
227 void *eeh_dev_init(struct device_node *dn, void *data);
228 void eeh_dev_phb_init_dynamic(struct pci_controller *phb);
229 int eeh_init(void);
230 int __init eeh_ops_register(struct eeh_ops *ops);
231 int __exit eeh_ops_unregister(const char *name);
232 unsigned long eeh_check_failure(const volatile void __iomem *token,
233 				unsigned long val);
234 int eeh_dev_check_failure(struct eeh_dev *edev);
235 void eeh_addr_cache_build(void);
236 void eeh_add_device_early(struct device_node *);
237 void eeh_add_device_tree_early(struct device_node *);
238 void eeh_add_device_late(struct pci_dev *);
239 void eeh_add_device_tree_late(struct pci_bus *);
240 void eeh_add_sysfs_files(struct pci_bus *);
241 void eeh_remove_device(struct pci_dev *);
242 
243 /**
244  * EEH_POSSIBLE_ERROR() -- test for possible MMIO failure.
245  *
246  * If this macro yields TRUE, the caller relays to eeh_check_failure()
247  * which does further tests out of line.
248  */
249 #define EEH_POSSIBLE_ERROR(val, type)	((val) == (type)~0 && eeh_subsystem_enabled)
250 
251 /*
252  * Reads from a device which has been isolated by EEH will return
253  * all 1s.  This macro gives an all-1s value of the given size (in
254  * bytes: 1, 2, or 4) for comparing with the result of a read.
255  */
256 #define EEH_IO_ERROR_VALUE(size)	(~0U >> ((4 - (size)) * 8))
257 
258 #else /* !CONFIG_EEH */
259 
260 static inline int eeh_init(void)
261 {
262 	return 0;
263 }
264 
265 static inline void *eeh_dev_init(struct device_node *dn, void *data)
266 {
267 	return NULL;
268 }
269 
270 static inline void eeh_dev_phb_init_dynamic(struct pci_controller *phb) { }
271 
272 static inline unsigned long eeh_check_failure(const volatile void __iomem *token, unsigned long val)
273 {
274 	return val;
275 }
276 
277 #define eeh_dev_check_failure(x) (0)
278 
279 static inline void eeh_addr_cache_build(void) { }
280 
281 static inline void eeh_add_device_early(struct device_node *dn) { }
282 
283 static inline void eeh_add_device_tree_early(struct device_node *dn) { }
284 
285 static inline void eeh_add_device_late(struct pci_dev *dev) { }
286 
287 static inline void eeh_add_device_tree_late(struct pci_bus *bus) { }
288 
289 static inline void eeh_add_sysfs_files(struct pci_bus *bus) { }
290 
291 static inline void eeh_remove_device(struct pci_dev *dev) { }
292 
293 #define EEH_POSSIBLE_ERROR(val, type) (0)
294 #define EEH_IO_ERROR_VALUE(size) (-1UL)
295 #endif /* CONFIG_EEH */
296 
297 #ifdef CONFIG_PPC64
298 /*
299  * MMIO read/write operations with EEH support.
300  */
301 static inline u8 eeh_readb(const volatile void __iomem *addr)
302 {
303 	u8 val = in_8(addr);
304 	if (EEH_POSSIBLE_ERROR(val, u8))
305 		return eeh_check_failure(addr, val);
306 	return val;
307 }
308 
309 static inline u16 eeh_readw(const volatile void __iomem *addr)
310 {
311 	u16 val = in_le16(addr);
312 	if (EEH_POSSIBLE_ERROR(val, u16))
313 		return eeh_check_failure(addr, val);
314 	return val;
315 }
316 
317 static inline u32 eeh_readl(const volatile void __iomem *addr)
318 {
319 	u32 val = in_le32(addr);
320 	if (EEH_POSSIBLE_ERROR(val, u32))
321 		return eeh_check_failure(addr, val);
322 	return val;
323 }
324 
325 static inline u64 eeh_readq(const volatile void __iomem *addr)
326 {
327 	u64 val = in_le64(addr);
328 	if (EEH_POSSIBLE_ERROR(val, u64))
329 		return eeh_check_failure(addr, val);
330 	return val;
331 }
332 
333 static inline u16 eeh_readw_be(const volatile void __iomem *addr)
334 {
335 	u16 val = in_be16(addr);
336 	if (EEH_POSSIBLE_ERROR(val, u16))
337 		return eeh_check_failure(addr, val);
338 	return val;
339 }
340 
341 static inline u32 eeh_readl_be(const volatile void __iomem *addr)
342 {
343 	u32 val = in_be32(addr);
344 	if (EEH_POSSIBLE_ERROR(val, u32))
345 		return eeh_check_failure(addr, val);
346 	return val;
347 }
348 
349 static inline u64 eeh_readq_be(const volatile void __iomem *addr)
350 {
351 	u64 val = in_be64(addr);
352 	if (EEH_POSSIBLE_ERROR(val, u64))
353 		return eeh_check_failure(addr, val);
354 	return val;
355 }
356 
357 static inline void eeh_memcpy_fromio(void *dest, const
358 				     volatile void __iomem *src,
359 				     unsigned long n)
360 {
361 	_memcpy_fromio(dest, src, n);
362 
363 	/* Look for ffff's here at dest[n].  Assume that at least 4 bytes
364 	 * were copied. Check all four bytes.
365 	 */
366 	if (n >= 4 && EEH_POSSIBLE_ERROR(*((u32 *)(dest + n - 4)), u32))
367 		eeh_check_failure(src, *((u32 *)(dest + n - 4)));
368 }
369 
370 /* in-string eeh macros */
371 static inline void eeh_readsb(const volatile void __iomem *addr, void * buf,
372 			      int ns)
373 {
374 	_insb(addr, buf, ns);
375 	if (EEH_POSSIBLE_ERROR((*(((u8*)buf)+ns-1)), u8))
376 		eeh_check_failure(addr, *(u8*)buf);
377 }
378 
379 static inline void eeh_readsw(const volatile void __iomem *addr, void * buf,
380 			      int ns)
381 {
382 	_insw(addr, buf, ns);
383 	if (EEH_POSSIBLE_ERROR((*(((u16*)buf)+ns-1)), u16))
384 		eeh_check_failure(addr, *(u16*)buf);
385 }
386 
387 static inline void eeh_readsl(const volatile void __iomem *addr, void * buf,
388 			      int nl)
389 {
390 	_insl(addr, buf, nl);
391 	if (EEH_POSSIBLE_ERROR((*(((u32*)buf)+nl-1)), u32))
392 		eeh_check_failure(addr, *(u32*)buf);
393 }
394 
395 #endif /* CONFIG_PPC64 */
396 #endif /* __KERNEL__ */
397 #endif /* _POWERPC_EEH_H */
398