eeh.h 15 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463
  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. #ifndef _POWERPC_EEH_H
  20. #define _POWERPC_EEH_H
  21. #ifdef __KERNEL__
  22. #include <linux/init.h>
  23. #include <linux/list.h>
  24. #include <linux/string.h>
  25. #include <linux/time.h>
  26. #include <linux/atomic.h>
  27. #include <uapi/asm/eeh.h>
  28. struct pci_dev;
  29. struct pci_bus;
  30. struct pci_dn;
  31. #ifdef CONFIG_EEH
  32. /* EEH subsystem flags */
  33. #define EEH_ENABLED 0x01 /* EEH enabled */
  34. #define EEH_FORCE_DISABLED 0x02 /* EEH disabled */
  35. #define EEH_PROBE_MODE_DEV 0x04 /* From PCI device */
  36. #define EEH_PROBE_MODE_DEVTREE 0x08 /* From device tree */
  37. #define EEH_VALID_PE_ZERO 0x10 /* PE#0 is valid */
  38. #define EEH_ENABLE_IO_FOR_LOG 0x20 /* Enable IO for log */
  39. #define EEH_EARLY_DUMP_LOG 0x40 /* Dump log immediately */
  40. /*
  41. * Delay for PE reset, all in ms
  42. *
  43. * PCI specification has reset hold time of 100 milliseconds.
  44. * We have 250 milliseconds here. The PCI bus settlement time
  45. * is specified as 1.5 seconds and we have 1.8 seconds.
  46. */
  47. #define EEH_PE_RST_HOLD_TIME 250
  48. #define EEH_PE_RST_SETTLE_TIME 1800
  49. /*
  50. * The struct is used to trace PE related EEH functionality.
  51. * In theory, there will have one instance of the struct to
  52. * be created against particular PE. In nature, PEs correlate
  53. * to each other. the struct has to reflect that hierarchy in
  54. * order to easily pick up those affected PEs when one particular
  55. * PE has EEH errors.
  56. *
  57. * Also, one particular PE might be composed of PCI device, PCI
  58. * bus and its subordinate components. The struct also need ship
  59. * the information. Further more, one particular PE is only meaingful
  60. * in the corresponding PHB. Therefore, the root PEs should be created
  61. * against existing PHBs in on-to-one fashion.
  62. */
  63. #define EEH_PE_INVALID (1 << 0) /* Invalid */
  64. #define EEH_PE_PHB (1 << 1) /* PHB PE */
  65. #define EEH_PE_DEVICE (1 << 2) /* Device PE */
  66. #define EEH_PE_BUS (1 << 3) /* Bus PE */
  67. #define EEH_PE_VF (1 << 4) /* VF PE */
  68. #define EEH_PE_ISOLATED (1 << 0) /* Isolated PE */
  69. #define EEH_PE_RECOVERING (1 << 1) /* Recovering PE */
  70. #define EEH_PE_CFG_BLOCKED (1 << 2) /* Block config access */
  71. #define EEH_PE_RESET (1 << 3) /* PE reset in progress */
  72. #define EEH_PE_KEEP (1 << 8) /* Keep PE on hotplug */
  73. #define EEH_PE_CFG_RESTRICTED (1 << 9) /* Block config on error */
  74. #define EEH_PE_REMOVED (1 << 10) /* Removed permanently */
  75. #define EEH_PE_PRI_BUS (1 << 11) /* Cached primary bus */
  76. struct eeh_pe {
  77. int type; /* PE type: PHB/Bus/Device */
  78. int state; /* PE EEH dependent mode */
  79. int config_addr; /* Traditional PCI address */
  80. int addr; /* PE configuration address */
  81. struct pci_controller *phb; /* Associated PHB */
  82. struct pci_bus *bus; /* Top PCI bus for bus PE */
  83. int check_count; /* Times of ignored error */
  84. int freeze_count; /* Times of froze up */
  85. struct timeval tstamp; /* Time on first-time freeze */
  86. int false_positives; /* Times of reported #ff's */
  87. atomic_t pass_dev_cnt; /* Count of passed through devs */
  88. struct eeh_pe *parent; /* Parent PE */
  89. void *data; /* PE auxillary data */
  90. struct list_head child_list; /* Link PE to the child list */
  91. struct list_head edevs; /* Link list of EEH devices */
  92. struct list_head child; /* Child PEs */
  93. };
  94. #define eeh_pe_for_each_dev(pe, edev, tmp) \
  95. list_for_each_entry_safe(edev, tmp, &pe->edevs, list)
  96. static inline bool eeh_pe_passed(struct eeh_pe *pe)
  97. {
  98. return pe ? !!atomic_read(&pe->pass_dev_cnt) : false;
  99. }
  100. /*
  101. * The struct is used to trace EEH state for the associated
  102. * PCI device node or PCI device. In future, it might
  103. * represent PE as well so that the EEH device to form
  104. * another tree except the currently existing tree of PCI
  105. * buses and PCI devices
  106. */
  107. #define EEH_DEV_BRIDGE (1 << 0) /* PCI bridge */
  108. #define EEH_DEV_ROOT_PORT (1 << 1) /* PCIe root port */
  109. #define EEH_DEV_DS_PORT (1 << 2) /* Downstream port */
  110. #define EEH_DEV_IRQ_DISABLED (1 << 3) /* Interrupt disabled */
  111. #define EEH_DEV_DISCONNECTED (1 << 4) /* Removing from PE */
  112. #define EEH_DEV_NO_HANDLER (1 << 8) /* No error handler */
  113. #define EEH_DEV_SYSFS (1 << 9) /* Sysfs created */
  114. #define EEH_DEV_REMOVED (1 << 10) /* Removed permanently */
  115. struct eeh_dev {
  116. int mode; /* EEH mode */
  117. int class_code; /* Class code of the device */
  118. int config_addr; /* Config address */
  119. int pe_config_addr; /* PE config address */
  120. u32 config_space[16]; /* Saved PCI config space */
  121. int pcix_cap; /* Saved PCIx capability */
  122. int pcie_cap; /* Saved PCIe capability */
  123. int aer_cap; /* Saved AER capability */
  124. int af_cap; /* Saved AF capability */
  125. struct eeh_pe *pe; /* Associated PE */
  126. struct list_head list; /* Form link list in the PE */
  127. struct list_head rmv_list; /* Record the removed edevs */
  128. struct pci_controller *phb; /* Associated PHB */
  129. struct pci_dn *pdn; /* Associated PCI device node */
  130. struct pci_dev *pdev; /* Associated PCI device */
  131. bool in_error; /* Error flag for edev */
  132. struct pci_dev *physfn; /* Associated SRIOV PF */
  133. struct pci_bus *bus; /* PCI bus for partial hotplug */
  134. };
  135. static inline struct pci_dn *eeh_dev_to_pdn(struct eeh_dev *edev)
  136. {
  137. return edev ? edev->pdn : NULL;
  138. }
  139. static inline struct pci_dev *eeh_dev_to_pci_dev(struct eeh_dev *edev)
  140. {
  141. return edev ? edev->pdev : NULL;
  142. }
  143. static inline struct eeh_pe *eeh_dev_to_pe(struct eeh_dev* edev)
  144. {
  145. return edev ? edev->pe : NULL;
  146. }
  147. /* Return values from eeh_ops::next_error */
  148. enum {
  149. EEH_NEXT_ERR_NONE = 0,
  150. EEH_NEXT_ERR_INF,
  151. EEH_NEXT_ERR_FROZEN_PE,
  152. EEH_NEXT_ERR_FENCED_PHB,
  153. EEH_NEXT_ERR_DEAD_PHB,
  154. EEH_NEXT_ERR_DEAD_IOC
  155. };
  156. /*
  157. * The struct is used to trace the registered EEH operation
  158. * callback functions. Actually, those operation callback
  159. * functions are heavily platform dependent. That means the
  160. * platform should register its own EEH operation callback
  161. * functions before any EEH further operations.
  162. */
  163. #define EEH_OPT_DISABLE 0 /* EEH disable */
  164. #define EEH_OPT_ENABLE 1 /* EEH enable */
  165. #define EEH_OPT_THAW_MMIO 2 /* MMIO enable */
  166. #define EEH_OPT_THAW_DMA 3 /* DMA enable */
  167. #define EEH_OPT_FREEZE_PE 4 /* Freeze PE */
  168. #define EEH_STATE_UNAVAILABLE (1 << 0) /* State unavailable */
  169. #define EEH_STATE_NOT_SUPPORT (1 << 1) /* EEH not supported */
  170. #define EEH_STATE_RESET_ACTIVE (1 << 2) /* Active reset */
  171. #define EEH_STATE_MMIO_ACTIVE (1 << 3) /* Active MMIO */
  172. #define EEH_STATE_DMA_ACTIVE (1 << 4) /* Active DMA */
  173. #define EEH_STATE_MMIO_ENABLED (1 << 5) /* MMIO enabled */
  174. #define EEH_STATE_DMA_ENABLED (1 << 6) /* DMA enabled */
  175. #define EEH_RESET_DEACTIVATE 0 /* Deactivate the PE reset */
  176. #define EEH_RESET_HOT 1 /* Hot reset */
  177. #define EEH_RESET_FUNDAMENTAL 3 /* Fundamental reset */
  178. #define EEH_LOG_TEMP 1 /* EEH temporary error log */
  179. #define EEH_LOG_PERM 2 /* EEH permanent error log */
  180. struct eeh_ops {
  181. char *name;
  182. int (*init)(void);
  183. int (*post_init)(void);
  184. void* (*probe)(struct pci_dn *pdn, void *data);
  185. int (*set_option)(struct eeh_pe *pe, int option);
  186. int (*get_pe_addr)(struct eeh_pe *pe);
  187. int (*get_state)(struct eeh_pe *pe, int *state);
  188. int (*reset)(struct eeh_pe *pe, int option);
  189. int (*wait_state)(struct eeh_pe *pe, int max_wait);
  190. int (*get_log)(struct eeh_pe *pe, int severity, char *drv_log, unsigned long len);
  191. int (*configure_bridge)(struct eeh_pe *pe);
  192. int (*err_inject)(struct eeh_pe *pe, int type, int func,
  193. unsigned long addr, unsigned long mask);
  194. int (*read_config)(struct pci_dn *pdn, int where, int size, u32 *val);
  195. int (*write_config)(struct pci_dn *pdn, int where, int size, u32 val);
  196. int (*next_error)(struct eeh_pe **pe);
  197. int (*restore_config)(struct pci_dn *pdn);
  198. };
  199. extern int eeh_subsystem_flags;
  200. extern int eeh_max_freezes;
  201. extern struct eeh_ops *eeh_ops;
  202. extern raw_spinlock_t confirm_error_lock;
  203. static inline void eeh_add_flag(int flag)
  204. {
  205. eeh_subsystem_flags |= flag;
  206. }
  207. static inline void eeh_clear_flag(int flag)
  208. {
  209. eeh_subsystem_flags &= ~flag;
  210. }
  211. static inline bool eeh_has_flag(int flag)
  212. {
  213. return !!(eeh_subsystem_flags & flag);
  214. }
  215. static inline bool eeh_enabled(void)
  216. {
  217. if (eeh_has_flag(EEH_FORCE_DISABLED) ||
  218. !eeh_has_flag(EEH_ENABLED))
  219. return false;
  220. return true;
  221. }
  222. static inline void eeh_serialize_lock(unsigned long *flags)
  223. {
  224. raw_spin_lock_irqsave(&confirm_error_lock, *flags);
  225. }
  226. static inline void eeh_serialize_unlock(unsigned long flags)
  227. {
  228. raw_spin_unlock_irqrestore(&confirm_error_lock, flags);
  229. }
  230. typedef void *(*eeh_traverse_func)(void *data, void *flag);
  231. void eeh_set_pe_aux_size(int size);
  232. int eeh_phb_pe_create(struct pci_controller *phb);
  233. struct eeh_pe *eeh_phb_pe_get(struct pci_controller *phb);
  234. struct eeh_pe *eeh_pe_get(struct eeh_dev *edev);
  235. int eeh_add_to_parent_pe(struct eeh_dev *edev);
  236. int eeh_rmv_from_parent_pe(struct eeh_dev *edev);
  237. void eeh_pe_update_time_stamp(struct eeh_pe *pe);
  238. void *eeh_pe_traverse(struct eeh_pe *root,
  239. eeh_traverse_func fn, void *flag);
  240. void *eeh_pe_dev_traverse(struct eeh_pe *root,
  241. eeh_traverse_func fn, void *flag);
  242. void eeh_pe_restore_bars(struct eeh_pe *pe);
  243. const char *eeh_pe_loc_get(struct eeh_pe *pe);
  244. struct pci_bus *eeh_pe_bus_get(struct eeh_pe *pe);
  245. struct eeh_dev *eeh_dev_init(struct pci_dn *pdn);
  246. void eeh_dev_phb_init_dynamic(struct pci_controller *phb);
  247. int eeh_init(void);
  248. int __init eeh_ops_register(struct eeh_ops *ops);
  249. int __exit eeh_ops_unregister(const char *name);
  250. int eeh_check_failure(const volatile void __iomem *token);
  251. int eeh_dev_check_failure(struct eeh_dev *edev);
  252. void eeh_addr_cache_build(void);
  253. void eeh_add_device_early(struct pci_dn *);
  254. void eeh_add_device_tree_early(struct pci_dn *);
  255. void eeh_add_device_late(struct pci_dev *);
  256. void eeh_add_device_tree_late(struct pci_bus *);
  257. void eeh_add_sysfs_files(struct pci_bus *);
  258. void eeh_remove_device(struct pci_dev *);
  259. int eeh_unfreeze_pe(struct eeh_pe *pe, bool sw_state);
  260. int eeh_pe_reset_and_recover(struct eeh_pe *pe);
  261. int eeh_dev_open(struct pci_dev *pdev);
  262. void eeh_dev_release(struct pci_dev *pdev);
  263. struct eeh_pe *eeh_iommu_group_to_pe(struct iommu_group *group);
  264. int eeh_pe_set_option(struct eeh_pe *pe, int option);
  265. int eeh_pe_get_state(struct eeh_pe *pe);
  266. int eeh_pe_reset(struct eeh_pe *pe, int option);
  267. int eeh_pe_configure(struct eeh_pe *pe);
  268. int eeh_pe_inject_err(struct eeh_pe *pe, int type, int func,
  269. unsigned long addr, unsigned long mask);
  270. /**
  271. * EEH_POSSIBLE_ERROR() -- test for possible MMIO failure.
  272. *
  273. * If this macro yields TRUE, the caller relays to eeh_check_failure()
  274. * which does further tests out of line.
  275. */
  276. #define EEH_POSSIBLE_ERROR(val, type) ((val) == (type)~0 && eeh_enabled())
  277. /*
  278. * Reads from a device which has been isolated by EEH will return
  279. * all 1s. This macro gives an all-1s value of the given size (in
  280. * bytes: 1, 2, or 4) for comparing with the result of a read.
  281. */
  282. #define EEH_IO_ERROR_VALUE(size) (~0U >> ((4 - (size)) * 8))
  283. #else /* !CONFIG_EEH */
  284. static inline bool eeh_enabled(void)
  285. {
  286. return false;
  287. }
  288. static inline int eeh_init(void)
  289. {
  290. return 0;
  291. }
  292. static inline void *eeh_dev_init(struct pci_dn *pdn, void *data)
  293. {
  294. return NULL;
  295. }
  296. static inline void eeh_dev_phb_init_dynamic(struct pci_controller *phb) { }
  297. static inline int eeh_check_failure(const volatile void __iomem *token)
  298. {
  299. return 0;
  300. }
  301. #define eeh_dev_check_failure(x) (0)
  302. static inline void eeh_addr_cache_build(void) { }
  303. static inline void eeh_add_device_early(struct pci_dn *pdn) { }
  304. static inline void eeh_add_device_tree_early(struct pci_dn *pdn) { }
  305. static inline void eeh_add_device_late(struct pci_dev *dev) { }
  306. static inline void eeh_add_device_tree_late(struct pci_bus *bus) { }
  307. static inline void eeh_add_sysfs_files(struct pci_bus *bus) { }
  308. static inline void eeh_remove_device(struct pci_dev *dev) { }
  309. #define EEH_POSSIBLE_ERROR(val, type) (0)
  310. #define EEH_IO_ERROR_VALUE(size) (-1UL)
  311. #endif /* CONFIG_EEH */
  312. #ifdef CONFIG_PPC64
  313. /*
  314. * MMIO read/write operations with EEH support.
  315. */
  316. static inline u8 eeh_readb(const volatile void __iomem *addr)
  317. {
  318. u8 val = in_8(addr);
  319. if (EEH_POSSIBLE_ERROR(val, u8))
  320. eeh_check_failure(addr);
  321. return val;
  322. }
  323. static inline u16 eeh_readw(const volatile void __iomem *addr)
  324. {
  325. u16 val = in_le16(addr);
  326. if (EEH_POSSIBLE_ERROR(val, u16))
  327. eeh_check_failure(addr);
  328. return val;
  329. }
  330. static inline u32 eeh_readl(const volatile void __iomem *addr)
  331. {
  332. u32 val = in_le32(addr);
  333. if (EEH_POSSIBLE_ERROR(val, u32))
  334. eeh_check_failure(addr);
  335. return val;
  336. }
  337. static inline u64 eeh_readq(const volatile void __iomem *addr)
  338. {
  339. u64 val = in_le64(addr);
  340. if (EEH_POSSIBLE_ERROR(val, u64))
  341. eeh_check_failure(addr);
  342. return val;
  343. }
  344. static inline u16 eeh_readw_be(const volatile void __iomem *addr)
  345. {
  346. u16 val = in_be16(addr);
  347. if (EEH_POSSIBLE_ERROR(val, u16))
  348. eeh_check_failure(addr);
  349. return val;
  350. }
  351. static inline u32 eeh_readl_be(const volatile void __iomem *addr)
  352. {
  353. u32 val = in_be32(addr);
  354. if (EEH_POSSIBLE_ERROR(val, u32))
  355. eeh_check_failure(addr);
  356. return val;
  357. }
  358. static inline u64 eeh_readq_be(const volatile void __iomem *addr)
  359. {
  360. u64 val = in_be64(addr);
  361. if (EEH_POSSIBLE_ERROR(val, u64))
  362. eeh_check_failure(addr);
  363. return val;
  364. }
  365. static inline void eeh_memcpy_fromio(void *dest, const
  366. volatile void __iomem *src,
  367. unsigned long n)
  368. {
  369. _memcpy_fromio(dest, src, n);
  370. /* Look for ffff's here at dest[n]. Assume that at least 4 bytes
  371. * were copied. Check all four bytes.
  372. */
  373. if (n >= 4 && EEH_POSSIBLE_ERROR(*((u32 *)(dest + n - 4)), u32))
  374. eeh_check_failure(src);
  375. }
  376. /* in-string eeh macros */
  377. static inline void eeh_readsb(const volatile void __iomem *addr, void * buf,
  378. int ns)
  379. {
  380. _insb(addr, buf, ns);
  381. if (EEH_POSSIBLE_ERROR((*(((u8*)buf)+ns-1)), u8))
  382. eeh_check_failure(addr);
  383. }
  384. static inline void eeh_readsw(const volatile void __iomem *addr, void * buf,
  385. int ns)
  386. {
  387. _insw(addr, buf, ns);
  388. if (EEH_POSSIBLE_ERROR((*(((u16*)buf)+ns-1)), u16))
  389. eeh_check_failure(addr);
  390. }
  391. static inline void eeh_readsl(const volatile void __iomem *addr, void * buf,
  392. int nl)
  393. {
  394. _insl(addr, buf, nl);
  395. if (EEH_POSSIBLE_ERROR((*(((u32*)buf)+nl-1)), u32))
  396. eeh_check_failure(addr);
  397. }
  398. #endif /* CONFIG_PPC64 */
  399. #endif /* __KERNEL__ */
  400. #endif /* _POWERPC_EEH_H */