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 #include <linux/atomic.h>
29
30 struct pci_dev;
31 struct pci_bus;
32 struct device_node;
33
34 #ifdef CONFIG_EEH
35
36 /* EEH subsystem flags */
37 #define EEH_ENABLED 0x01 /* EEH enabled */
38 #define EEH_FORCE_DISABLED 0x02 /* EEH disabled */
39 #define EEH_PROBE_MODE_DEV 0x04 /* From PCI device */
40 #define EEH_PROBE_MODE_DEVTREE 0x08 /* From device tree */
41 #define EEH_ENABLE_IO_FOR_LOG 0x10 /* Enable IO for log */
42
43 /*
44 * Delay for PE reset, all in ms
45 *
46 * PCI specification has reset hold time of 100 milliseconds.
47 * We have 250 milliseconds here. The PCI bus settlement time
48 * is specified as 1.5 seconds and we have 1.8 seconds.
49 */
50 #define EEH_PE_RST_HOLD_TIME 250
51 #define EEH_PE_RST_SETTLE_TIME 1800
52
53 /*
54 * The struct is used to trace PE related EEH functionality.
55 * In theory, there will have one instance of the struct to
56 * be created against particular PE. In nature, PEs corelate
57 * to each other. the struct has to reflect that hierarchy in
58 * order to easily pick up those affected PEs when one particular
59 * PE has EEH errors.
60 *
61 * Also, one particular PE might be composed of PCI device, PCI
62 * bus and its subordinate components. The struct also need ship
63 * the information. Further more, one particular PE is only meaingful
64 * in the corresponding PHB. Therefore, the root PEs should be created
65 * against existing PHBs in on-to-one fashion.
66 */
67 #define EEH_PE_INVALID (1 << 0) /* Invalid */
68 #define EEH_PE_PHB (1 << 1) /* PHB PE */
69 #define EEH_PE_DEVICE (1 << 2) /* Device PE */
70 #define EEH_PE_BUS (1 << 3) /* Bus PE */
71
72 #define EEH_PE_ISOLATED (1 << 0) /* Isolated PE */
73 #define EEH_PE_RECOVERING (1 << 1) /* Recovering PE */
74 #define EEH_PE_CFG_BLOCKED (1 << 2) /* Block config access */
75
76 #define EEH_PE_KEEP (1 << 8) /* Keep PE on hotplug */
77 #define EEH_PE_CFG_RESTRICTED (1 << 9) /* Block config on error */
78
79 struct eeh_pe {
80 int type; /* PE type: PHB/Bus/Device */
81 int state; /* PE EEH dependent mode */
82 int config_addr; /* Traditional PCI address */
83 int addr; /* PE configuration address */
84 struct pci_controller *phb; /* Associated PHB */
85 struct pci_bus *bus; /* Top PCI bus for bus PE */
86 int check_count; /* Times of ignored error */
87 int freeze_count; /* Times of froze up */
88 struct timeval tstamp; /* Time on first-time freeze */
89 int false_positives; /* Times of reported #ff's */
90 atomic_t pass_dev_cnt; /* Count of passed through devs */
91 struct eeh_pe *parent; /* Parent PE */
92 void *data; /* PE auxillary data */
93 struct list_head child_list; /* Link PE to the child list */
94 struct list_head edevs; /* Link list of EEH devices */
95 struct list_head child; /* Child PEs */
96 };
97
98 #define eeh_pe_for_each_dev(pe, edev, tmp) \
99 list_for_each_entry_safe(edev, tmp, &pe->edevs, list)
100
eeh_pe_passed(struct eeh_pe * pe)101 static inline bool eeh_pe_passed(struct eeh_pe *pe)
102 {
103 return pe ? !!atomic_read(&pe->pass_dev_cnt) : false;
104 }
105
106 /*
107 * The struct is used to trace EEH state for the associated
108 * PCI device node or PCI device. In future, it might
109 * represent PE as well so that the EEH device to form
110 * another tree except the currently existing tree of PCI
111 * buses and PCI devices
112 */
113 #define EEH_DEV_BRIDGE (1 << 0) /* PCI bridge */
114 #define EEH_DEV_ROOT_PORT (1 << 1) /* PCIe root port */
115 #define EEH_DEV_DS_PORT (1 << 2) /* Downstream port */
116 #define EEH_DEV_IRQ_DISABLED (1 << 3) /* Interrupt disabled */
117 #define EEH_DEV_DISCONNECTED (1 << 4) /* Removing from PE */
118
119 #define EEH_DEV_NO_HANDLER (1 << 8) /* No error handler */
120 #define EEH_DEV_SYSFS (1 << 9) /* Sysfs created */
121 #define EEH_DEV_REMOVED (1 << 10) /* Removed permanently */
122
123 struct eeh_dev {
124 int mode; /* EEH mode */
125 int class_code; /* Class code of the device */
126 int config_addr; /* Config address */
127 int pe_config_addr; /* PE config address */
128 u32 config_space[16]; /* Saved PCI config space */
129 int pcix_cap; /* Saved PCIx capability */
130 int pcie_cap; /* Saved PCIe capability */
131 int aer_cap; /* Saved AER capability */
132 struct eeh_pe *pe; /* Associated PE */
133 struct list_head list; /* Form link list in the PE */
134 struct pci_controller *phb; /* Associated PHB */
135 struct device_node *dn; /* Associated device node */
136 struct pci_dev *pdev; /* Associated PCI device */
137 struct pci_bus *bus; /* PCI bus for partial hotplug */
138 };
139
eeh_dev_to_of_node(struct eeh_dev * edev)140 static inline struct device_node *eeh_dev_to_of_node(struct eeh_dev *edev)
141 {
142 return edev ? edev->dn : NULL;
143 }
144
eeh_dev_to_pci_dev(struct eeh_dev * edev)145 static inline struct pci_dev *eeh_dev_to_pci_dev(struct eeh_dev *edev)
146 {
147 return edev ? edev->pdev : NULL;
148 }
149
eeh_dev_to_pe(struct eeh_dev * edev)150 static inline struct eeh_pe *eeh_dev_to_pe(struct eeh_dev* edev)
151 {
152 return edev ? edev->pe : NULL;
153 }
154
155 /* Return values from eeh_ops::next_error */
156 enum {
157 EEH_NEXT_ERR_NONE = 0,
158 EEH_NEXT_ERR_INF,
159 EEH_NEXT_ERR_FROZEN_PE,
160 EEH_NEXT_ERR_FENCED_PHB,
161 EEH_NEXT_ERR_DEAD_PHB,
162 EEH_NEXT_ERR_DEAD_IOC
163 };
164
165 /*
166 * The struct is used to trace the registered EEH operation
167 * callback functions. Actually, those operation callback
168 * functions are heavily platform dependent. That means the
169 * platform should register its own EEH operation callback
170 * functions before any EEH further operations.
171 */
172 #define EEH_OPT_DISABLE 0 /* EEH disable */
173 #define EEH_OPT_ENABLE 1 /* EEH enable */
174 #define EEH_OPT_THAW_MMIO 2 /* MMIO enable */
175 #define EEH_OPT_THAW_DMA 3 /* DMA enable */
176 #define EEH_OPT_FREEZE_PE 4 /* Freeze PE */
177 #define EEH_STATE_UNAVAILABLE (1 << 0) /* State unavailable */
178 #define EEH_STATE_NOT_SUPPORT (1 << 1) /* EEH not supported */
179 #define EEH_STATE_RESET_ACTIVE (1 << 2) /* Active reset */
180 #define EEH_STATE_MMIO_ACTIVE (1 << 3) /* Active MMIO */
181 #define EEH_STATE_DMA_ACTIVE (1 << 4) /* Active DMA */
182 #define EEH_STATE_MMIO_ENABLED (1 << 5) /* MMIO enabled */
183 #define EEH_STATE_DMA_ENABLED (1 << 6) /* DMA enabled */
184 #define EEH_PE_STATE_NORMAL 0 /* Normal state */
185 #define EEH_PE_STATE_RESET 1 /* PE reset asserted */
186 #define EEH_PE_STATE_STOPPED_IO_DMA 2 /* Frozen PE */
187 #define EEH_PE_STATE_STOPPED_DMA 4 /* Stopped DMA, Enabled IO */
188 #define EEH_PE_STATE_UNAVAIL 5 /* Unavailable */
189 #define EEH_RESET_DEACTIVATE 0 /* Deactivate the PE reset */
190 #define EEH_RESET_HOT 1 /* Hot reset */
191 #define EEH_RESET_FUNDAMENTAL 3 /* Fundamental reset */
192 #define EEH_LOG_TEMP 1 /* EEH temporary error log */
193 #define EEH_LOG_PERM 2 /* EEH permanent error log */
194
195 struct eeh_ops {
196 char *name;
197 int (*init)(void);
198 int (*post_init)(void);
199 void* (*of_probe)(struct device_node *dn, void *flag);
200 int (*dev_probe)(struct pci_dev *dev, void *flag);
201 int (*set_option)(struct eeh_pe *pe, int option);
202 int (*get_pe_addr)(struct eeh_pe *pe);
203 int (*get_state)(struct eeh_pe *pe, int *state);
204 int (*reset)(struct eeh_pe *pe, int option);
205 int (*wait_state)(struct eeh_pe *pe, int max_wait);
206 int (*get_log)(struct eeh_pe *pe, int severity, char *drv_log, unsigned long len);
207 int (*configure_bridge)(struct eeh_pe *pe);
208 int (*err_inject)(struct eeh_pe *pe, int type, int func,
209 unsigned long addr, unsigned long mask);
210 int (*read_config)(struct device_node *dn, int where, int size, u32 *val);
211 int (*write_config)(struct device_node *dn, int where, int size, u32 val);
212 int (*next_error)(struct eeh_pe **pe);
213 int (*restore_config)(struct device_node *dn);
214 };
215
216 extern int eeh_subsystem_flags;
217 extern struct eeh_ops *eeh_ops;
218 extern raw_spinlock_t confirm_error_lock;
219
eeh_add_flag(int flag)220 static inline void eeh_add_flag(int flag)
221 {
222 eeh_subsystem_flags |= flag;
223 }
224
eeh_clear_flag(int flag)225 static inline void eeh_clear_flag(int flag)
226 {
227 eeh_subsystem_flags &= ~flag;
228 }
229
eeh_has_flag(int flag)230 static inline bool eeh_has_flag(int flag)
231 {
232 return !!(eeh_subsystem_flags & flag);
233 }
234
eeh_enabled(void)235 static inline bool eeh_enabled(void)
236 {
237 if (eeh_has_flag(EEH_FORCE_DISABLED) ||
238 !eeh_has_flag(EEH_ENABLED))
239 return false;
240
241 return true;
242 }
243
eeh_serialize_lock(unsigned long * flags)244 static inline void eeh_serialize_lock(unsigned long *flags)
245 {
246 raw_spin_lock_irqsave(&confirm_error_lock, *flags);
247 }
248
eeh_serialize_unlock(unsigned long flags)249 static inline void eeh_serialize_unlock(unsigned long flags)
250 {
251 raw_spin_unlock_irqrestore(&confirm_error_lock, flags);
252 }
253
254 /*
255 * Max number of EEH freezes allowed before we consider the device
256 * to be permanently disabled.
257 */
258 #define EEH_MAX_ALLOWED_FREEZES 5
259
260 typedef void *(*eeh_traverse_func)(void *data, void *flag);
261 void eeh_set_pe_aux_size(int size);
262 int eeh_phb_pe_create(struct pci_controller *phb);
263 struct eeh_pe *eeh_phb_pe_get(struct pci_controller *phb);
264 struct eeh_pe *eeh_pe_get(struct eeh_dev *edev);
265 int eeh_add_to_parent_pe(struct eeh_dev *edev);
266 int eeh_rmv_from_parent_pe(struct eeh_dev *edev);
267 void eeh_pe_update_time_stamp(struct eeh_pe *pe);
268 void *eeh_pe_traverse(struct eeh_pe *root,
269 eeh_traverse_func fn, void *flag);
270 void *eeh_pe_dev_traverse(struct eeh_pe *root,
271 eeh_traverse_func fn, void *flag);
272 void eeh_pe_restore_bars(struct eeh_pe *pe);
273 const char *eeh_pe_loc_get(struct eeh_pe *pe);
274 struct pci_bus *eeh_pe_bus_get(struct eeh_pe *pe);
275
276 void *eeh_dev_init(struct device_node *dn, void *data);
277 void eeh_dev_phb_init_dynamic(struct pci_controller *phb);
278 int eeh_init(void);
279 int __init eeh_ops_register(struct eeh_ops *ops);
280 int __exit eeh_ops_unregister(const char *name);
281 int eeh_check_failure(const volatile void __iomem *token);
282 int eeh_dev_check_failure(struct eeh_dev *edev);
283 void eeh_addr_cache_build(void);
284 void eeh_add_device_early(struct device_node *);
285 void eeh_add_device_tree_early(struct device_node *);
286 void eeh_add_device_late(struct pci_dev *);
287 void eeh_add_device_tree_late(struct pci_bus *);
288 void eeh_add_sysfs_files(struct pci_bus *);
289 void eeh_remove_device(struct pci_dev *);
290 int eeh_unfreeze_pe(struct eeh_pe *pe, bool sw_state);
291 int eeh_pe_reset_and_recover(struct eeh_pe *pe);
292 int eeh_dev_open(struct pci_dev *pdev);
293 void eeh_dev_release(struct pci_dev *pdev);
294 struct eeh_pe *eeh_iommu_group_to_pe(struct iommu_group *group);
295 int eeh_pe_set_option(struct eeh_pe *pe, int option);
296 int eeh_pe_get_state(struct eeh_pe *pe);
297 int eeh_pe_reset(struct eeh_pe *pe, int option);
298 int eeh_pe_configure(struct eeh_pe *pe);
299
300 /**
301 * EEH_POSSIBLE_ERROR() -- test for possible MMIO failure.
302 *
303 * If this macro yields TRUE, the caller relays to eeh_check_failure()
304 * which does further tests out of line.
305 */
306 #define EEH_POSSIBLE_ERROR(val, type) ((val) == (type)~0 && eeh_enabled())
307
308 /*
309 * Reads from a device which has been isolated by EEH will return
310 * all 1s. This macro gives an all-1s value of the given size (in
311 * bytes: 1, 2, or 4) for comparing with the result of a read.
312 */
313 #define EEH_IO_ERROR_VALUE(size) (~0U >> ((4 - (size)) * 8))
314
315 #else /* !CONFIG_EEH */
316
eeh_enabled(void)317 static inline bool eeh_enabled(void)
318 {
319 return false;
320 }
321
eeh_init(void)322 static inline int eeh_init(void)
323 {
324 return 0;
325 }
326
eeh_dev_init(struct device_node * dn,void * data)327 static inline void *eeh_dev_init(struct device_node *dn, void *data)
328 {
329 return NULL;
330 }
331
eeh_dev_phb_init_dynamic(struct pci_controller * phb)332 static inline void eeh_dev_phb_init_dynamic(struct pci_controller *phb) { }
333
eeh_check_failure(const volatile void __iomem * token)334 static inline int eeh_check_failure(const volatile void __iomem *token)
335 {
336 return 0;
337 }
338
339 #define eeh_dev_check_failure(x) (0)
340
eeh_addr_cache_build(void)341 static inline void eeh_addr_cache_build(void) { }
342
eeh_add_device_early(struct device_node * dn)343 static inline void eeh_add_device_early(struct device_node *dn) { }
344
eeh_add_device_tree_early(struct device_node * dn)345 static inline void eeh_add_device_tree_early(struct device_node *dn) { }
346
eeh_add_device_late(struct pci_dev * dev)347 static inline void eeh_add_device_late(struct pci_dev *dev) { }
348
eeh_add_device_tree_late(struct pci_bus * bus)349 static inline void eeh_add_device_tree_late(struct pci_bus *bus) { }
350
eeh_add_sysfs_files(struct pci_bus * bus)351 static inline void eeh_add_sysfs_files(struct pci_bus *bus) { }
352
eeh_remove_device(struct pci_dev * dev)353 static inline void eeh_remove_device(struct pci_dev *dev) { }
354
355 #define EEH_POSSIBLE_ERROR(val, type) (0)
356 #define EEH_IO_ERROR_VALUE(size) (-1UL)
357 #endif /* CONFIG_EEH */
358
359 #ifdef CONFIG_PPC64
360 /*
361 * MMIO read/write operations with EEH support.
362 */
eeh_readb(const volatile void __iomem * addr)363 static inline u8 eeh_readb(const volatile void __iomem *addr)
364 {
365 u8 val = in_8(addr);
366 if (EEH_POSSIBLE_ERROR(val, u8))
367 eeh_check_failure(addr);
368 return val;
369 }
370
eeh_readw(const volatile void __iomem * addr)371 static inline u16 eeh_readw(const volatile void __iomem *addr)
372 {
373 u16 val = in_le16(addr);
374 if (EEH_POSSIBLE_ERROR(val, u16))
375 eeh_check_failure(addr);
376 return val;
377 }
378
eeh_readl(const volatile void __iomem * addr)379 static inline u32 eeh_readl(const volatile void __iomem *addr)
380 {
381 u32 val = in_le32(addr);
382 if (EEH_POSSIBLE_ERROR(val, u32))
383 eeh_check_failure(addr);
384 return val;
385 }
386
eeh_readq(const volatile void __iomem * addr)387 static inline u64 eeh_readq(const volatile void __iomem *addr)
388 {
389 u64 val = in_le64(addr);
390 if (EEH_POSSIBLE_ERROR(val, u64))
391 eeh_check_failure(addr);
392 return val;
393 }
394
eeh_readw_be(const volatile void __iomem * addr)395 static inline u16 eeh_readw_be(const volatile void __iomem *addr)
396 {
397 u16 val = in_be16(addr);
398 if (EEH_POSSIBLE_ERROR(val, u16))
399 eeh_check_failure(addr);
400 return val;
401 }
402
eeh_readl_be(const volatile void __iomem * addr)403 static inline u32 eeh_readl_be(const volatile void __iomem *addr)
404 {
405 u32 val = in_be32(addr);
406 if (EEH_POSSIBLE_ERROR(val, u32))
407 eeh_check_failure(addr);
408 return val;
409 }
410
eeh_readq_be(const volatile void __iomem * addr)411 static inline u64 eeh_readq_be(const volatile void __iomem *addr)
412 {
413 u64 val = in_be64(addr);
414 if (EEH_POSSIBLE_ERROR(val, u64))
415 eeh_check_failure(addr);
416 return val;
417 }
418
eeh_memcpy_fromio(void * dest,const volatile void __iomem * src,unsigned long n)419 static inline void eeh_memcpy_fromio(void *dest, const
420 volatile void __iomem *src,
421 unsigned long n)
422 {
423 _memcpy_fromio(dest, src, n);
424
425 /* Look for ffff's here at dest[n]. Assume that at least 4 bytes
426 * were copied. Check all four bytes.
427 */
428 if (n >= 4 && EEH_POSSIBLE_ERROR(*((u32 *)(dest + n - 4)), u32))
429 eeh_check_failure(src);
430 }
431
432 /* in-string eeh macros */
eeh_readsb(const volatile void __iomem * addr,void * buf,int ns)433 static inline void eeh_readsb(const volatile void __iomem *addr, void * buf,
434 int ns)
435 {
436 _insb(addr, buf, ns);
437 if (EEH_POSSIBLE_ERROR((*(((u8*)buf)+ns-1)), u8))
438 eeh_check_failure(addr);
439 }
440
eeh_readsw(const volatile void __iomem * addr,void * buf,int ns)441 static inline void eeh_readsw(const volatile void __iomem *addr, void * buf,
442 int ns)
443 {
444 _insw(addr, buf, ns);
445 if (EEH_POSSIBLE_ERROR((*(((u16*)buf)+ns-1)), u16))
446 eeh_check_failure(addr);
447 }
448
eeh_readsl(const volatile void __iomem * addr,void * buf,int nl)449 static inline void eeh_readsl(const volatile void __iomem *addr, void * buf,
450 int nl)
451 {
452 _insl(addr, buf, nl);
453 if (EEH_POSSIBLE_ERROR((*(((u32*)buf)+nl-1)), u32))
454 eeh_check_failure(addr);
455 }
456
457 #endif /* CONFIG_PPC64 */
458 #endif /* __KERNEL__ */
459 #endif /* _POWERPC_EEH_H */
460