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()
378  * For that case, it's meaningless to recover frozen PE. Intead,
381 static int eeh_phb_check_failure(struct eeh_pe *pe)  in eeh_phb_check_failure()  argument
390 	/* Find the PHB PE */  in eeh_phb_check_failure()
391 	phb_pe = eeh_phb_pe_get(pe->phb);  in eeh_phb_check_failure()
393 		pr_warn("%s Can't find PE for PHB#%x\n",  in eeh_phb_check_failure()
394 			__func__, pe->phb->global_number);  in eeh_phb_check_failure()
446 	struct eeh_pe *pe, *parent_pe;  in eeh_dev_check_failure()  local
460 	pe = eeh_dev_to_pe(edev);  in eeh_dev_check_failure()
463 	if (!pe) {  in eeh_dev_check_failure()
473 	ret = eeh_phb_check_failure(pe);  in eeh_dev_check_failure()
478 	 * If the PE isn't owned by us, we shouldn't check the  in eeh_dev_check_failure()
479 	 * state. Instead, let the owner handle it if the PE has  in eeh_dev_check_failure()
482 	if (eeh_pe_passed(pe))  in eeh_dev_check_failure()
493 	if (pe->state & EEH_PE_ISOLATED) {  in eeh_dev_check_failure()
494 		pe->check_count++;  in eeh_dev_check_failure()
495 		if (pe->check_count == EEH_MAX_FAILS) {  in eeh_dev_check_failure()
501 				pe->check_count,  in eeh_dev_check_failure()
518 	ret = eeh_ops->get_state(pe, NULL);  in eeh_dev_check_failure()
523 	 * PE's state, EEH not support and Permanently unavailable  in eeh_dev_check_failure()
524 	 * state, PE is in good state.  in eeh_dev_check_failure()
529 		pe->false_positives++;  in eeh_dev_check_failure()
535 	 * It should be corner case that the parent PE has been  in eeh_dev_check_failure()
539 	parent_pe = pe->parent;  in eeh_dev_check_failure()
545 		/* Frozen parent PE ? */  in eeh_dev_check_failure()
548 			pe = parent_pe;  in eeh_dev_check_failure()
549 			pr_err("EEH: Failure of PHB#%x-PE#%x will be handled at parent PHB#%x-PE#%x.\n",  in eeh_dev_check_failure()
550 			       pe->phb->global_number, pe->addr,  in eeh_dev_check_failure()
551 			       pe->phb->global_number, parent_pe->addr);  in eeh_dev_check_failure()
564 	eeh_pe_mark_isolated(pe);  in eeh_dev_check_failure()
571 	pr_debug("EEH: %s: Frozen PHB#%x-PE#%x detected\n",  in eeh_dev_check_failure()
572 		__func__, pe->phb->global_number, pe->addr);  in eeh_dev_check_failure()
573 	eeh_send_failure_event(pe);  in eeh_dev_check_failure()
615  * @pe: EEH PE
621 int eeh_pci_enable(struct eeh_pe *pe, int function)  in eeh_pci_enable()  argument
626 	 * pHyp doesn't allow to enable IO or DMA on unfrozen PE.  in eeh_pci_enable()
627 	 * Also, it's pointless to enable them on unfrozen PE. So  in eeh_pci_enable()
653 		rc = eeh_ops->get_state(pe, NULL);  in eeh_pci_enable()
668 	rc = eeh_ops->set_option(pe, function);  in eeh_pci_enable()
671 			"PHB#%x-PE#%x, err=%d\n",  in eeh_pci_enable()
672 			__func__, function, pe->phb->global_number,  in eeh_pci_enable()
673 			pe->addr, rc);  in eeh_pci_enable()
677 		rc = eeh_wait_state(pe, PCI_BUS_RESET_WAIT_MSEC);  in eeh_pci_enable()
744 	struct eeh_pe *pe = eeh_dev_to_pe(edev);  in pcibios_set_pcie_reset_state()  local
746 	if (!pe) {  in pcibios_set_pcie_reset_state()
747 		pr_err("%s: No PE found on PCI device %s\n",  in pcibios_set_pcie_reset_state()
754 		eeh_ops->reset(pe, EEH_RESET_DEACTIVATE);  in pcibios_set_pcie_reset_state()
755 		eeh_unfreeze_pe(pe);  in pcibios_set_pcie_reset_state()
756 		if (!(pe->type & EEH_PE_VF))  in pcibios_set_pcie_reset_state()
757 			eeh_pe_state_clear(pe, EEH_PE_CFG_BLOCKED, true);  in pcibios_set_pcie_reset_state()
758 		eeh_pe_dev_traverse(pe, eeh_restore_dev_state, dev);  in pcibios_set_pcie_reset_state()
759 		eeh_pe_state_clear(pe, EEH_PE_ISOLATED, true);  in pcibios_set_pcie_reset_state()
762 		eeh_pe_mark_isolated(pe);  in pcibios_set_pcie_reset_state()
763 		eeh_pe_state_clear(pe, EEH_PE_CFG_BLOCKED, true);  in pcibios_set_pcie_reset_state()
764 		eeh_ops->set_option(pe, EEH_OPT_FREEZE_PE);  in pcibios_set_pcie_reset_state()
765 		eeh_pe_dev_traverse(pe, eeh_disable_and_save_dev_state, dev);  in pcibios_set_pcie_reset_state()
766 		if (!(pe->type & EEH_PE_VF))  in pcibios_set_pcie_reset_state()
767 			eeh_pe_state_mark(pe, EEH_PE_CFG_BLOCKED);  in pcibios_set_pcie_reset_state()
768 		eeh_ops->reset(pe, EEH_RESET_HOT);  in pcibios_set_pcie_reset_state()
771 		eeh_pe_mark_isolated(pe);  in pcibios_set_pcie_reset_state()
772 		eeh_pe_state_clear(pe, EEH_PE_CFG_BLOCKED, true);  in pcibios_set_pcie_reset_state()
773 		eeh_ops->set_option(pe, EEH_OPT_FREEZE_PE);  in pcibios_set_pcie_reset_state()
774 		eeh_pe_dev_traverse(pe, eeh_disable_and_save_dev_state, dev);  in pcibios_set_pcie_reset_state()
775 		if (!(pe->type & EEH_PE_VF))  in pcibios_set_pcie_reset_state()
776 			eeh_pe_state_mark(pe, EEH_PE_CFG_BLOCKED);  in pcibios_set_pcie_reset_state()
777 		eeh_ops->reset(pe, EEH_RESET_FUNDAMENTAL);  in pcibios_set_pcie_reset_state()
780 		eeh_pe_state_clear(pe, EEH_PE_ISOLATED | EEH_PE_CFG_BLOCKED, true);  in pcibios_set_pcie_reset_state()
809 	struct eeh_pe *pe;  in eeh_pe_refreeze_passed()  local
812 	eeh_for_each_pe(root, pe) {  in eeh_pe_refreeze_passed()
813 		if (eeh_pe_passed(pe)) {  in eeh_pe_refreeze_passed()
814 			state = eeh_ops->get_state(pe, NULL);  in eeh_pe_refreeze_passed()
817 				pr_info("EEH: Passed-through PE PHB#%x-PE#%x was thawed by reset, re-freezing for safety.\n",  in eeh_pe_refreeze_passed()
818 					pe->phb->global_number, pe->addr);  in eeh_pe_refreeze_passed()
819 				eeh_pe_set_option(pe, EEH_OPT_FREEZE_PE);  in eeh_pe_refreeze_passed()
826  * eeh_pe_reset_full - Complete a full reset process on the indicated PE
827  * @pe: EEH PE
829  * This function executes a full reset procedure on a PE, including setting
835  * This function will attempt to reset a PE three times before failing.
837 int eeh_pe_reset_full(struct eeh_pe *pe, bool include_passed)  in eeh_pe_reset_full()  argument
847 	 * PE requires a fundamental reset.  in eeh_pe_reset_full()
849 	eeh_pe_dev_traverse(pe, eeh_set_dev_freset, &freset);  in eeh_pe_reset_full()
854 	/* Mark the PE as in reset state and block config space accesses */  in eeh_pe_reset_full()
855 	eeh_pe_state_mark(pe, reset_state);  in eeh_pe_reset_full()
859 		ret = eeh_pe_reset(pe, type, include_passed);  in eeh_pe_reset_full()
861 			ret = eeh_pe_reset(pe, EEH_RESET_DEACTIVATE,  in eeh_pe_reset_full()
865 			pr_warn("EEH: Failure %d resetting PHB#%x-PE#%x (attempt %d)\n\n",  in eeh_pe_reset_full()
866 				state, pe->phb->global_number, pe->addr, i + 1);  in eeh_pe_reset_full()
870 			pr_warn("EEH: PHB#%x-PE#%x: Successful reset (attempt %d)\n",  in eeh_pe_reset_full()
871 				pe->phb->global_number, pe->addr, i + 1);  in eeh_pe_reset_full()
873 		/* Wait until the PE is in a functioning state */  in eeh_pe_reset_full()
874 		state = eeh_wait_state(pe, PCI_BUS_RESET_WAIT_MSEC);  in eeh_pe_reset_full()
876 			pr_warn("EEH: Unrecoverable slot failure on PHB#%x-PE#%x",  in eeh_pe_reset_full()
877 				pe->phb->global_number, pe->addr);  in eeh_pe_reset_full()
884 			pr_warn("EEH: PHB#%x-PE#%x: Slot inactive after reset: 0x%x (attempt %d)\n",  in eeh_pe_reset_full()
885 				pe->phb->global_number, pe->addr, state, i + 1);  in eeh_pe_reset_full()
888 	/* Resetting the PE may have unfrozen child PEs. If those PEs have been  in eeh_pe_reset_full()
892 		eeh_pe_refreeze_passed(pe);  in eeh_pe_reset_full()
894 	eeh_pe_state_clear(pe, reset_state, true);  in eeh_pe_reset_full()
921 	 * we can restore it after reset, either PHB range or PE range.  in eeh_save_bars()
1082 	if (!edev || !edev->pdev || !edev->pe) {  in eeh_remove_device()
1089 	 * device attached to the parent PE in order for BAR restore  in eeh_remove_device()
1091 	 * from the parent PE during the BAR resotre.  in eeh_remove_device()
1120 	if (!(edev->pe->state & EEH_PE_KEEP))  in eeh_remove_device()
1126 int eeh_unfreeze_pe(struct eeh_pe *pe)  in eeh_unfreeze_pe()  argument
1130 	ret = eeh_pci_enable(pe, EEH_OPT_THAW_MMIO);  in eeh_unfreeze_pe()
1132 		pr_warn("%s: Failure %d enabling IO on PHB#%x-PE#%x\n",  in eeh_unfreeze_pe()
1133 			__func__, ret, pe->phb->global_number, pe->addr);  in eeh_unfreeze_pe()
1137 	ret = eeh_pci_enable(pe, EEH_OPT_THAW_DMA);  in eeh_unfreeze_pe()
1139 		pr_warn("%s: Failure %d enabling DMA on PHB#%x-PE#%x\n",  in eeh_unfreeze_pe()
1140 			__func__, ret, pe->phb->global_number, pe->addr);  in eeh_unfreeze_pe()
1155 static int eeh_pe_change_owner(struct eeh_pe *pe)  in eeh_pe_change_owner()  argument
1162 	/* Check PE state */  in eeh_pe_change_owner()
1163 	ret = eeh_ops->get_state(pe, NULL);  in eeh_pe_change_owner()
1167 	/* Unfrozen PE, nothing to do */  in eeh_pe_change_owner()
1171 	/* Frozen PE, check if it needs PE level reset */  in eeh_pe_change_owner()
1172 	eeh_pe_for_each_dev(pe, edev, tmp) {  in eeh_pe_change_owner()
1191 			return eeh_pe_reset_and_recover(pe);  in eeh_pe_change_owner()
1195 	ret = eeh_unfreeze_pe(pe);  in eeh_pe_change_owner()
1197 		eeh_pe_state_clear(pe, EEH_PE_ISOLATED, true);  in eeh_pe_change_owner()
1202  * eeh_dev_open - Increase count of pass through devices for PE
1206  * PE. In the result, the EEH errors detected on the PE won't be
1207  * reported. The PE owner will be responsible for detection
1221 	/* No EEH device or PE ? */  in eeh_dev_open()
1223 	if (!edev || !edev->pe)  in eeh_dev_open()
1227 	 * The PE might have been put into frozen state, but we  in eeh_dev_open()
1229 	 * in frozen PE won't work properly. Clear the frozen state  in eeh_dev_open()
1232 	ret = eeh_pe_change_owner(edev->pe);  in eeh_dev_open()
1236 	/* Increase PE's pass through count */  in eeh_dev_open()
1237 	atomic_inc(&edev->pe->pass_dev_cnt);  in eeh_dev_open()
1248  * eeh_dev_release - Decrease count of pass through devices for PE
1251  * Decrease count of pass through devices for the indicated PE. If
1252  * there is no passed through device in PE, the EEH errors detected
1253  * on the PE will be reported and handled as usual.
1267 	if (!edev || !edev->pe || !eeh_pe_passed(edev->pe))  in eeh_dev_release()
1270 	/* Decrease PE's pass through count */  in eeh_dev_release()
1271 	WARN_ON(atomic_dec_if_positive(&edev->pe->pass_dev_cnt) < 0);  in eeh_dev_release()
1272 	eeh_pe_change_owner(edev->pe);  in eeh_dev_release()
1297  * eeh_iommu_group_to_pe - Convert IOMMU group to EEH PE
1300  * The routine is called to convert IOMMU group to EEH PE.
1316 	/* No EEH device or PE ? */  in eeh_iommu_group_to_pe()
1318 	if (!edev || !edev->pe)  in eeh_iommu_group_to_pe()
1321 	return edev->pe;  in eeh_iommu_group_to_pe()
1328  * eeh_pe_set_option - Set options for the indicated PE
1329  * @pe: EEH PE
1333  * on the indicated PE, to enable IO or DMA for the frozen PE.
1335 int eeh_pe_set_option(struct eeh_pe *pe, int option)  in eeh_pe_set_option()  argument
1339 	/* Invalid PE ? */  in eeh_pe_set_option()
1340 	if (!pe)  in eeh_pe_set_option()
1346 	 * isn't expected to be disabled on one specific PE.  in eeh_pe_set_option()
1351 			ret = eeh_pe_change_owner(pe);  in eeh_pe_set_option()
1366 		ret = eeh_pci_enable(pe, option);  in eeh_pe_set_option()
1379  * eeh_pe_get_state - Retrieve PE's state
1380  * @pe: EEH PE
1382  * Retrieve the PE's state, which includes 3 aspects: enabled
1385 int eeh_pe_get_state(struct eeh_pe *pe)  in eeh_pe_get_state()  argument
1390 	/* Existing PE ? */  in eeh_pe_get_state()
1391 	if (!pe)  in eeh_pe_get_state()
1398 	 * If the parent PE is owned by the host kernel and is undergoing  in eeh_pe_get_state()
1399 	 * error recovery, we should return the PE state as temporarily  in eeh_pe_get_state()
1403 	if (pe->parent &&  in eeh_pe_get_state()
1404 	    !(pe->state & EEH_PE_REMOVED) &&  in eeh_pe_get_state()
1405 	    (pe->parent->state & (EEH_PE_ISOLATED | EEH_PE_RECOVERING)))  in eeh_pe_get_state()
1408 	result = eeh_ops->get_state(pe, NULL);  in eeh_pe_get_state()
1428 static int eeh_pe_reenable_devices(struct eeh_pe *pe, bool include_passed)  in eeh_pe_reenable_devices()  argument
1434 	eeh_pe_restore_bars(pe);  in eeh_pe_reenable_devices()
1440 	eeh_pe_for_each_dev(pe, edev, tmp) {  in eeh_pe_reenable_devices()
1453 	/* The PE is still in frozen state */  in eeh_pe_reenable_devices()
1454 	if (include_passed || !eeh_pe_passed(pe)) {  in eeh_pe_reenable_devices()
1455 		ret = eeh_unfreeze_pe(pe);  in eeh_pe_reenable_devices()
1457 		pr_info("EEH: Note: Leaving passthrough PHB#%x-PE#%x frozen.\n",  in eeh_pe_reenable_devices()
1458 			pe->phb->global_number, pe->addr);  in eeh_pe_reenable_devices()
1460 		eeh_pe_state_clear(pe, EEH_PE_ISOLATED, include_passed);  in eeh_pe_reenable_devices()
1466  * eeh_pe_reset - Issue PE reset according to specified type
1467  * @pe: EEH PE
1470  * The routine is called to reset the specified PE with the
1472  * PE reset is the most important part for error recovery.
1474 int eeh_pe_reset(struct eeh_pe *pe, int option, bool include_passed)  in eeh_pe_reset()  argument
1478 	/* Invalid PE ? */  in eeh_pe_reset()
1479 	if (!pe)  in eeh_pe_reset()
1487 		ret = eeh_ops->reset(pe, option);  in eeh_pe_reset()
1488 		eeh_pe_state_clear(pe, EEH_PE_CFG_BLOCKED, include_passed);  in eeh_pe_reset()
1492 		ret = eeh_pe_reenable_devices(pe, include_passed);  in eeh_pe_reset()
1497 		 * Proactively freeze the PE to drop all MMIO access  in eeh_pe_reset()
1501 		eeh_ops->set_option(pe, EEH_OPT_FREEZE_PE);  in eeh_pe_reset()
1503 		eeh_pe_state_mark(pe, EEH_PE_CFG_BLOCKED);  in eeh_pe_reset()
1504 		ret = eeh_ops->reset(pe, option);  in eeh_pe_reset()
1517  * eeh_pe_configure - Configure PCI bridges after PE reset
1518  * @pe: EEH PE
1521  * those PCI devices, especially PCI bridges affected by PE
1524 int eeh_pe_configure(struct eeh_pe *pe)  in eeh_pe_configure()  argument
1528 	/* Invalid PE ? */  in eeh_pe_configure()
1529 	if (!pe)  in eeh_pe_configure()
1537  * eeh_pe_inject_err - Injecting the specified PCI error to the indicated PE
1538  * @pe: the indicated PE
1545  * is determined by @type and @function, to the indicated PE for
1548 int eeh_pe_inject_err(struct eeh_pe *pe, int type, int func,  in eeh_pe_inject_err()  argument
1551 	/* Invalid PE ? */  in eeh_pe_inject_err()
1552 	if (!pe)  in eeh_pe_inject_err()
1567 	return eeh_ops->err_inject(pe, type, func, addr, mask);  in eeh_pe_inject_err()
1627 	struct eeh_pe *pe;  in eeh_force_recover_write()  local
1636 	 * When PE is NULL the event is a "special" event. Rather than  in eeh_force_recover_write()
1637 	 * recovering a specific PE it forces the EEH core to scan for failed  in eeh_force_recover_write()
1654 	/* Retrieve PE */  in eeh_force_recover_write()
1655 	pe = eeh_pe_get(hose, pe_no);  in eeh_force_recover_write()
1656 	if (!pe)  in eeh_force_recover_write()
1663 	 * from an odd state (e.g. PE removed, or recovery of a  in eeh_force_recover_write()
1664 	 * non-isolated PE)  in eeh_force_recover_write()
1666 	__eeh_send_failure_event(pe);  in eeh_force_recover_write()
1798 	 *    PE freeze. Using the in_8() accessor skips the eeh detection hook  in eeh_debugfs_break_device()
1801 	 *    where the HW will asyncronously freeze a PE and it's up to  in eeh_debugfs_break_device()