89  * PE would be created there.
95  * EEH allowed maximal frozen times. If one particular PE's
96  * frozen count in last hour exceeds this limit, the PE will
137 	u64 slot_resets;	/* PE reset			*/
176 			edev->pe->phb->global_number, edev->bdfn >> 8,  in eeh_dump_dev_log()
179 		edev->pe->phb->global_number, edev->bdfn >> 8,  in eeh_dump_dev_log()
269 static void *eeh_dump_pe_log(struct eeh_pe *pe, void *flag)  in eeh_dump_pe_log()  argument
274 	eeh_pe_for_each_dev(pe, edev, tmp)  in eeh_dump_pe_log()
283  * @pe: EEH PE
291 void eeh_slot_error_detail(struct eeh_pe *pe, int severity)  in eeh_slot_error_detail()  argument
304 	 * When the @severity is EEH_LOG_PERM, the PE is going to be  in eeh_slot_error_detail()
306 	 * the PE will be closed. The drivers rely on working IO path  in eeh_slot_error_detail()
311 	if (!(pe->type & EEH_PE_PHB)) {  in eeh_slot_error_detail()
314 			eeh_pci_enable(pe, EEH_OPT_THAW_MMIO);  in eeh_slot_error_detail()
321 		 * is set automatically when the PE is put to EEH_PE_ISOLATED.  in eeh_slot_error_detail()
328 		eeh_ops->configure_bridge(pe);  in eeh_slot_error_detail()
329 		if (!(pe->state & EEH_PE_CFG_BLOCKED)) {  in eeh_slot_error_detail()
330 			eeh_pe_restore_bars(pe);  in eeh_slot_error_detail()
333 			eeh_pe_traverse(pe, eeh_dump_pe_log, &loglen);  in eeh_slot_error_detail()
337 	eeh_ops->get_log(pe, severity, pci_regs_buf, loglen);  in eeh_slot_error_detail()
354  * For that case, it's meaningless to recover frozen PE. Intead,
357 static int eeh_phb_check_failure(struct eeh_pe *pe)  in eeh_phb_check_failure()  argument
366 	/* Find the PHB PE */  in eeh_phb_check_failure()
367 	phb_pe = eeh_phb_pe_get(pe->phb);  in eeh_phb_check_failure()
369 		pr_warn("%s Can't find PE for PHB#%x\n",  in eeh_phb_check_failure()
370 			__func__, pe->phb->global_number);  in eeh_phb_check_failure()
422 	struct eeh_pe *pe, *parent_pe;  in eeh_dev_check_failure()  local
436 	pe = eeh_dev_to_pe(edev);  in eeh_dev_check_failure()
439 	if (!pe) {  in eeh_dev_check_failure()
449 	ret = eeh_phb_check_failure(pe);  in eeh_dev_check_failure()
454 	 * If the PE isn't owned by us, we shouldn't check the  in eeh_dev_check_failure()
455 	 * state. Instead, let the owner handle it if the PE has  in eeh_dev_check_failure()
458 	if (eeh_pe_passed(pe))  in eeh_dev_check_failure()
469 	if (pe->state & EEH_PE_ISOLATED) {  in eeh_dev_check_failure()
470 		pe->check_count++;  in eeh_dev_check_failure()
471 		if (pe->check_count == EEH_MAX_FAILS) {  in eeh_dev_check_failure()
477 				pe->check_count,  in eeh_dev_check_failure()
494 	ret = eeh_ops->get_state(pe, NULL);  in eeh_dev_check_failure()
499 	 * PE's state, EEH not support and Permanently unavailable  in eeh_dev_check_failure()
500 	 * state, PE is in good state.  in eeh_dev_check_failure()
505 		pe->false_positives++;  in eeh_dev_check_failure()
511 	 * It should be corner case that the parent PE has been  in eeh_dev_check_failure()
515 	parent_pe = pe->parent;  in eeh_dev_check_failure()
521 		/* Frozen parent PE ? */  in eeh_dev_check_failure()
524 			pe = parent_pe;  in eeh_dev_check_failure()
525 			pr_err("EEH: Failure of PHB#%x-PE#%x will be handled at parent PHB#%x-PE#%x.\n",  in eeh_dev_check_failure()
526 			       pe->phb->global_number, pe->addr,  in eeh_dev_check_failure()
527 			       pe->phb->global_number, parent_pe->addr);  in eeh_dev_check_failure()
540 	eeh_pe_mark_isolated(pe);  in eeh_dev_check_failure()
547 	pr_debug("EEH: %s: Frozen PHB#%x-PE#%x detected\n",  in eeh_dev_check_failure()
548 		__func__, pe->phb->global_number, pe->addr);  in eeh_dev_check_failure()
549 	eeh_send_failure_event(pe);  in eeh_dev_check_failure()
591  * @pe: EEH PE
597 int eeh_pci_enable(struct eeh_pe *pe, int function)  in eeh_pci_enable()  argument
602 	 * pHyp doesn't allow to enable IO or DMA on unfrozen PE.  in eeh_pci_enable()
603 	 * Also, it's pointless to enable them on unfrozen PE. So  in eeh_pci_enable()
629 		rc = eeh_ops->get_state(pe, NULL);  in eeh_pci_enable()
644 	rc = eeh_ops->set_option(pe, function);  in eeh_pci_enable()
647 			"PHB#%x-PE#%x, err=%d\n",  in eeh_pci_enable()
648 			__func__, function, pe->phb->global_number,  in eeh_pci_enable()
649 			pe->addr, rc);  in eeh_pci_enable()
653 		rc = eeh_wait_state(pe, PCI_BUS_RESET_WAIT_MSEC);  in eeh_pci_enable()
720 	struct eeh_pe *pe = eeh_dev_to_pe(edev);  in pcibios_set_pcie_reset_state()  local
722 	if (!pe) {  in pcibios_set_pcie_reset_state()
723 		pr_err("%s: No PE found on PCI device %s\n",  in pcibios_set_pcie_reset_state()
730 		eeh_ops->reset(pe, EEH_RESET_DEACTIVATE);  in pcibios_set_pcie_reset_state()
731 		eeh_unfreeze_pe(pe);  in pcibios_set_pcie_reset_state()
732 		if (!(pe->type & EEH_PE_VF))  in pcibios_set_pcie_reset_state()
733 			eeh_pe_state_clear(pe, EEH_PE_CFG_BLOCKED, true);  in pcibios_set_pcie_reset_state()
734 		eeh_pe_dev_traverse(pe, eeh_restore_dev_state, dev);  in pcibios_set_pcie_reset_state()
735 		eeh_pe_state_clear(pe, EEH_PE_ISOLATED, true);  in pcibios_set_pcie_reset_state()
738 		eeh_pe_mark_isolated(pe);  in pcibios_set_pcie_reset_state()
739 		eeh_pe_state_clear(pe, EEH_PE_CFG_BLOCKED, true);  in pcibios_set_pcie_reset_state()
740 		eeh_ops->set_option(pe, EEH_OPT_FREEZE_PE);  in pcibios_set_pcie_reset_state()
741 		eeh_pe_dev_traverse(pe, eeh_disable_and_save_dev_state, dev);  in pcibios_set_pcie_reset_state()
742 		if (!(pe->type & EEH_PE_VF))  in pcibios_set_pcie_reset_state()
743 			eeh_pe_state_mark(pe, EEH_PE_CFG_BLOCKED);  in pcibios_set_pcie_reset_state()
744 		eeh_ops->reset(pe, EEH_RESET_HOT);  in pcibios_set_pcie_reset_state()
747 		eeh_pe_mark_isolated(pe);  in pcibios_set_pcie_reset_state()
748 		eeh_pe_state_clear(pe, EEH_PE_CFG_BLOCKED, true);  in pcibios_set_pcie_reset_state()
749 		eeh_ops->set_option(pe, EEH_OPT_FREEZE_PE);  in pcibios_set_pcie_reset_state()
750 		eeh_pe_dev_traverse(pe, eeh_disable_and_save_dev_state, dev);  in pcibios_set_pcie_reset_state()
751 		if (!(pe->type & EEH_PE_VF))  in pcibios_set_pcie_reset_state()
752 			eeh_pe_state_mark(pe, EEH_PE_CFG_BLOCKED);  in pcibios_set_pcie_reset_state()
753 		eeh_ops->reset(pe, EEH_RESET_FUNDAMENTAL);  in pcibios_set_pcie_reset_state()
756 		eeh_pe_state_clear(pe, EEH_PE_ISOLATED | EEH_PE_CFG_BLOCKED, true);  in pcibios_set_pcie_reset_state()
785 	struct eeh_pe *pe;  in eeh_pe_refreeze_passed()  local
788 	eeh_for_each_pe(root, pe) {  in eeh_pe_refreeze_passed()
789 		if (eeh_pe_passed(pe)) {  in eeh_pe_refreeze_passed()
790 			state = eeh_ops->get_state(pe, NULL);  in eeh_pe_refreeze_passed()
793 				pr_info("EEH: Passed-through PE PHB#%x-PE#%x was thawed by reset, re-freezing for safety.\n",  in eeh_pe_refreeze_passed()
794 					pe->phb->global_number, pe->addr);  in eeh_pe_refreeze_passed()
795 				eeh_pe_set_option(pe, EEH_OPT_FREEZE_PE);  in eeh_pe_refreeze_passed()
802  * eeh_pe_reset_full - Complete a full reset process on the indicated PE
803  * @pe: EEH PE
805  * This function executes a full reset procedure on a PE, including setting
811  * This function will attempt to reset a PE three times before failing.
813 int eeh_pe_reset_full(struct eeh_pe *pe, bool include_passed)  in eeh_pe_reset_full()  argument
823 	 * PE requires a fundamental reset.  in eeh_pe_reset_full()
825 	eeh_pe_dev_traverse(pe, eeh_set_dev_freset, &freset);  in eeh_pe_reset_full()
830 	/* Mark the PE as in reset state and block config space accesses */  in eeh_pe_reset_full()
831 	eeh_pe_state_mark(pe, reset_state);  in eeh_pe_reset_full()
835 		ret = eeh_pe_reset(pe, type, include_passed);  in eeh_pe_reset_full()
837 			ret = eeh_pe_reset(pe, EEH_RESET_DEACTIVATE,  in eeh_pe_reset_full()
841 			pr_warn("EEH: Failure %d resetting PHB#%x-PE#%x (attempt %d)\n\n",  in eeh_pe_reset_full()
842 				state, pe->phb->global_number, pe->addr, i + 1);  in eeh_pe_reset_full()
846 			pr_warn("EEH: PHB#%x-PE#%x: Successful reset (attempt %d)\n",  in eeh_pe_reset_full()
847 				pe->phb->global_number, pe->addr, i + 1);  in eeh_pe_reset_full()
849 		/* Wait until the PE is in a functioning state */  in eeh_pe_reset_full()
850 		state = eeh_wait_state(pe, PCI_BUS_RESET_WAIT_MSEC);  in eeh_pe_reset_full()
852 			pr_warn("EEH: Unrecoverable slot failure on PHB#%x-PE#%x",  in eeh_pe_reset_full()
853 				pe->phb->global_number, pe->addr);  in eeh_pe_reset_full()
860 			pr_warn("EEH: PHB#%x-PE#%x: Slot inactive after reset: 0x%x (attempt %d)\n",  in eeh_pe_reset_full()
861 				pe->phb->global_number, pe->addr, state, i + 1);  in eeh_pe_reset_full()
864 	/* Resetting the PE may have unfrozen child PEs. If those PEs have been  in eeh_pe_reset_full()
868 		eeh_pe_refreeze_passed(pe);  in eeh_pe_reset_full()
870 	eeh_pe_state_clear(pe, reset_state, true);  in eeh_pe_reset_full()
897 	 * we can restore it after reset, either PHB range or PE range.  in eeh_save_bars()
1058 	if (!edev || !edev->pdev || !edev->pe) {  in eeh_remove_device()
1065 	 * device attached to the parent PE in order for BAR restore  in eeh_remove_device()
1067 	 * from the parent PE during the BAR resotre.  in eeh_remove_device()
1096 	if (!(edev->pe->state & EEH_PE_KEEP))  in eeh_remove_device()
1102 int eeh_unfreeze_pe(struct eeh_pe *pe)  in eeh_unfreeze_pe()  argument
1106 	ret = eeh_pci_enable(pe, EEH_OPT_THAW_MMIO);  in eeh_unfreeze_pe()
1108 		pr_warn("%s: Failure %d enabling IO on PHB#%x-PE#%x\n",  in eeh_unfreeze_pe()
1109 			__func__, ret, pe->phb->global_number, pe->addr);  in eeh_unfreeze_pe()
1113 	ret = eeh_pci_enable(pe, EEH_OPT_THAW_DMA);  in eeh_unfreeze_pe()
1115 		pr_warn("%s: Failure %d enabling DMA on PHB#%x-PE#%x\n",  in eeh_unfreeze_pe()
1116 			__func__, ret, pe->phb->global_number, pe->addr);  in eeh_unfreeze_pe()
1131 static int eeh_pe_change_owner(struct eeh_pe *pe)  in eeh_pe_change_owner()  argument
1138 	/* Check PE state */  in eeh_pe_change_owner()
1139 	ret = eeh_ops->get_state(pe, NULL);  in eeh_pe_change_owner()
1143 	/* Unfrozen PE, nothing to do */  in eeh_pe_change_owner()
1147 	/* Frozen PE, check if it needs PE level reset */  in eeh_pe_change_owner()
1148 	eeh_pe_for_each_dev(pe, edev, tmp) {  in eeh_pe_change_owner()
1167 			return eeh_pe_reset_and_recover(pe);  in eeh_pe_change_owner()
1171 	ret = eeh_unfreeze_pe(pe);  in eeh_pe_change_owner()
1173 		eeh_pe_state_clear(pe, EEH_PE_ISOLATED, true);  in eeh_pe_change_owner()
1178  * eeh_dev_open - Increase count of pass through devices for PE
1182  * PE. In the result, the EEH errors detected on the PE won't be
1183  * reported. The PE owner will be responsible for detection
1197 	/* No EEH device or PE ? */  in eeh_dev_open()
1199 	if (!edev || !edev->pe)  in eeh_dev_open()
1203 	 * The PE might have been put into frozen state, but we  in eeh_dev_open()
1205 	 * in frozen PE won't work properly. Clear the frozen state  in eeh_dev_open()
1208 	ret = eeh_pe_change_owner(edev->pe);  in eeh_dev_open()
1212 	/* Increase PE's pass through count */  in eeh_dev_open()
1213 	atomic_inc(&edev->pe->pass_dev_cnt);  in eeh_dev_open()
1224  * eeh_dev_release - Decrease count of pass through devices for PE
1227  * Decrease count of pass through devices for the indicated PE. If
1228  * there is no passed through device in PE, the EEH errors detected
1229  * on the PE will be reported and handled as usual.
1243 	if (!edev || !edev->pe || !eeh_pe_passed(edev->pe))  in eeh_dev_release()
1246 	/* Decrease PE's pass through count */  in eeh_dev_release()
1247 	WARN_ON(atomic_dec_if_positive(&edev->pe->pass_dev_cnt) < 0);  in eeh_dev_release()
1248 	eeh_pe_change_owner(edev->pe);  in eeh_dev_release()
1273  * eeh_iommu_group_to_pe - Convert IOMMU group to EEH PE
1276  * The routine is called to convert IOMMU group to EEH PE.
1292 	/* No EEH device or PE ? */  in eeh_iommu_group_to_pe()
1294 	if (!edev || !edev->pe)  in eeh_iommu_group_to_pe()
1297 	return edev->pe;  in eeh_iommu_group_to_pe()
1304  * eeh_pe_set_option - Set options for the indicated PE
1305  * @pe: EEH PE
1309  * on the indicated PE, to enable IO or DMA for the frozen PE.
1311 int eeh_pe_set_option(struct eeh_pe *pe, int option)  in eeh_pe_set_option()  argument
1315 	/* Invalid PE ? */  in eeh_pe_set_option()
1316 	if (!pe)  in eeh_pe_set_option()
1322 	 * isn't expected to be disabled on one specific PE.  in eeh_pe_set_option()
1327 			ret = eeh_pe_change_owner(pe);  in eeh_pe_set_option()
1342 		ret = eeh_pci_enable(pe, option);  in eeh_pe_set_option()
1355  * eeh_pe_get_state - Retrieve PE's state
1356  * @pe: EEH PE
1358  * Retrieve the PE's state, which includes 3 aspects: enabled
1361 int eeh_pe_get_state(struct eeh_pe *pe)  in eeh_pe_get_state()  argument
1366 	/* Existing PE ? */  in eeh_pe_get_state()
1367 	if (!pe)  in eeh_pe_get_state()
1374 	 * If the parent PE is owned by the host kernel and is undergoing  in eeh_pe_get_state()
1375 	 * error recovery, we should return the PE state as temporarily  in eeh_pe_get_state()
1379 	if (pe->parent &&  in eeh_pe_get_state()
1380 	    !(pe->state & EEH_PE_REMOVED) &&  in eeh_pe_get_state()
1381 	    (pe->parent->state & (EEH_PE_ISOLATED | EEH_PE_RECOVERING)))  in eeh_pe_get_state()
1384 	result = eeh_ops->get_state(pe, NULL);  in eeh_pe_get_state()
1404 static int eeh_pe_reenable_devices(struct eeh_pe *pe, bool include_passed)  in eeh_pe_reenable_devices()  argument
1410 	eeh_pe_restore_bars(pe);  in eeh_pe_reenable_devices()
1416 	eeh_pe_for_each_dev(pe, edev, tmp) {  in eeh_pe_reenable_devices()
1429 	/* The PE is still in frozen state */  in eeh_pe_reenable_devices()
1430 	if (include_passed || !eeh_pe_passed(pe)) {  in eeh_pe_reenable_devices()
1431 		ret = eeh_unfreeze_pe(pe);  in eeh_pe_reenable_devices()
1433 		pr_info("EEH: Note: Leaving passthrough PHB#%x-PE#%x frozen.\n",  in eeh_pe_reenable_devices()
1434 			pe->phb->global_number, pe->addr);  in eeh_pe_reenable_devices()
1436 		eeh_pe_state_clear(pe, EEH_PE_ISOLATED, include_passed);  in eeh_pe_reenable_devices()
1442  * eeh_pe_reset - Issue PE reset according to specified type
1443  * @pe: EEH PE
1446  * The routine is called to reset the specified PE with the
1448  * PE reset is the most important part for error recovery.
1450 int eeh_pe_reset(struct eeh_pe *pe, int option, bool include_passed)  in eeh_pe_reset()  argument
1454 	/* Invalid PE ? */  in eeh_pe_reset()
1455 	if (!pe)  in eeh_pe_reset()
1463 		ret = eeh_ops->reset(pe, option);  in eeh_pe_reset()
1464 		eeh_pe_state_clear(pe, EEH_PE_CFG_BLOCKED, include_passed);  in eeh_pe_reset()
1468 		ret = eeh_pe_reenable_devices(pe, include_passed);  in eeh_pe_reset()
1473 		 * Proactively freeze the PE to drop all MMIO access  in eeh_pe_reset()
1477 		eeh_ops->set_option(pe, EEH_OPT_FREEZE_PE);  in eeh_pe_reset()
1479 		eeh_pe_state_mark(pe, EEH_PE_CFG_BLOCKED);  in eeh_pe_reset()
1480 		ret = eeh_ops->reset(pe, option);  in eeh_pe_reset()
1493  * eeh_pe_configure - Configure PCI bridges after PE reset
1494  * @pe: EEH PE
1497  * those PCI devices, especially PCI bridges affected by PE
1500 int eeh_pe_configure(struct eeh_pe *pe)  in eeh_pe_configure()  argument
1504 	/* Invalid PE ? */  in eeh_pe_configure()
1505 	if (!pe)  in eeh_pe_configure()
1513  * eeh_pe_inject_err - Injecting the specified PCI error to the indicated PE
1514  * @pe: the indicated PE
1521  * is determined by @type and @function, to the indicated PE for
1524 int eeh_pe_inject_err(struct eeh_pe *pe, int type, int func,  in eeh_pe_inject_err()  argument
1527 	/* Invalid PE ? */  in eeh_pe_inject_err()
1528 	if (!pe)  in eeh_pe_inject_err()
1543 	return eeh_ops->err_inject(pe, type, func, addr, mask);  in eeh_pe_inject_err()
1634 	struct eeh_pe *pe;  in eeh_force_recover_write()  local
1643 	 * When PE is NULL the event is a "special" event. Rather than  in eeh_force_recover_write()
1644 	 * recovering a specific PE it forces the EEH core to scan for failed  in eeh_force_recover_write()
1661 	/* Retrieve PE */  in eeh_force_recover_write()
1662 	pe = eeh_pe_get(hose, pe_no);  in eeh_force_recover_write()
1663 	if (!pe)  in eeh_force_recover_write()
1670 	 * from an odd state (e.g. PE removed, or recovery of a  in eeh_force_recover_write()
1671 	 * non-isolated PE)  in eeh_force_recover_write()
1673 	__eeh_send_failure_event(pe);  in eeh_force_recover_write()
1792 	 *    PE freeze. Using the in_8() accessor skips the eeh detection hook  in eeh_debugfs_break_device()
1795 	 *    where the HW will asyncronously freeze a PE and it's up to  in eeh_debugfs_break_device()