2 * The file intends to implement the functions needed by EEH, which is
3 * built on IODA compliant chip. Actually, lots of functions related
4 * to EEH would be built based on the OPAL APIs.
6 * Copyright Benjamin Herrenschmidt & Gavin Shan, IBM Corporation 2013.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License as published by
10 * the Free Software Foundation; either version 2 of the License, or
11 * (at your option) any later version.
14 #include <linux/bootmem.h>
15 #include <linux/debugfs.h>
16 #include <linux/delay.h>
17 #include <linux/init.h>
19 #include <linux/irq.h>
20 #include <linux/kernel.h>
21 #include <linux/msi.h>
22 #include <linux/notifier.h>
23 #include <linux/pci.h>
24 #include <linux/string.h>
27 #include <asm/eeh_event.h>
29 #include <asm/iommu.h>
30 #include <asm/msi_bitmap.h>
32 #include <asm/pci-bridge.h>
33 #include <asm/ppc-pci.h>
39 static int ioda_eeh_nb_init = 0;
41 static int ioda_eeh_event(struct notifier_block *nb,
42 unsigned long events, void *change)
44 uint64_t changed_evts = (uint64_t)change;
46 /* We simply send special EEH event */
47 if ((changed_evts & OPAL_EVENT_PCI_ERROR) &&
48 (events & OPAL_EVENT_PCI_ERROR))
49 eeh_send_failure_event(NULL);
54 static struct notifier_block ioda_eeh_nb = {
55 .notifier_call = ioda_eeh_event,
60 #ifdef CONFIG_DEBUG_FS
61 static int ioda_eeh_dbgfs_set(void *data, int offset, u64 val)
63 struct pci_controller *hose = data;
64 struct pnv_phb *phb = hose->private_data;
66 out_be64(phb->regs + offset, val);
70 static int ioda_eeh_dbgfs_get(void *data, int offset, u64 *val)
72 struct pci_controller *hose = data;
73 struct pnv_phb *phb = hose->private_data;
75 *val = in_be64(phb->regs + offset);
79 static int ioda_eeh_outb_dbgfs_set(void *data, u64 val)
81 return ioda_eeh_dbgfs_set(data, 0xD10, val);
84 static int ioda_eeh_outb_dbgfs_get(void *data, u64 *val)
86 return ioda_eeh_dbgfs_get(data, 0xD10, val);
89 static int ioda_eeh_inbA_dbgfs_set(void *data, u64 val)
91 return ioda_eeh_dbgfs_set(data, 0xD90, val);
94 static int ioda_eeh_inbA_dbgfs_get(void *data, u64 *val)
96 return ioda_eeh_dbgfs_get(data, 0xD90, val);
99 static int ioda_eeh_inbB_dbgfs_set(void *data, u64 val)
101 return ioda_eeh_dbgfs_set(data, 0xE10, val);
104 static int ioda_eeh_inbB_dbgfs_get(void *data, u64 *val)
106 return ioda_eeh_dbgfs_get(data, 0xE10, val);
109 DEFINE_SIMPLE_ATTRIBUTE(ioda_eeh_outb_dbgfs_ops, ioda_eeh_outb_dbgfs_get,
110 ioda_eeh_outb_dbgfs_set, "0x%llx\n");
111 DEFINE_SIMPLE_ATTRIBUTE(ioda_eeh_inbA_dbgfs_ops, ioda_eeh_inbA_dbgfs_get,
112 ioda_eeh_inbA_dbgfs_set, "0x%llx\n");
113 DEFINE_SIMPLE_ATTRIBUTE(ioda_eeh_inbB_dbgfs_ops, ioda_eeh_inbB_dbgfs_get,
114 ioda_eeh_inbB_dbgfs_set, "0x%llx\n");
115 #endif /* CONFIG_DEBUG_FS */
118 * ioda_eeh_post_init - Chip dependent post initialization
119 * @hose: PCI controller
121 * The function will be called after eeh PEs and devices
122 * have been built. That means the EEH is ready to supply
123 * service with I/O cache.
125 static int ioda_eeh_post_init(struct pci_controller *hose)
127 struct pnv_phb *phb = hose->private_data;
130 /* Register OPAL event notifier */
131 if (!ioda_eeh_nb_init) {
132 ret = opal_notifier_register(&ioda_eeh_nb);
134 pr_err("%s: Can't register OPAL event notifier (%d)\n",
139 ioda_eeh_nb_init = 1;
142 #ifdef CONFIG_DEBUG_FS
144 debugfs_create_file("err_injct_outbound", 0600,
146 &ioda_eeh_outb_dbgfs_ops);
147 debugfs_create_file("err_injct_inboundA", 0600,
149 &ioda_eeh_inbA_dbgfs_ops);
150 debugfs_create_file("err_injct_inboundB", 0600,
152 &ioda_eeh_inbB_dbgfs_ops);
156 phb->eeh_state |= PNV_EEH_STATE_ENABLED;
162 * ioda_eeh_set_option - Set EEH operation or I/O setting
166 * Enable or disable EEH option for the indicated PE. The
167 * function also can be used to enable I/O or DMA for the
170 static int ioda_eeh_set_option(struct eeh_pe *pe, int option)
174 struct pci_controller *hose = pe->phb;
175 struct pnv_phb *phb = hose->private_data;
177 /* Check on PE number */
178 if (pe->addr < 0 || pe->addr >= phb->ioda.total_pe) {
179 pr_err("%s: PE address %x out of range [0, %x] "
181 __func__, pe->addr, phb->ioda.total_pe,
182 hose->global_number);
188 case EEH_OPT_DISABLE:
194 case EEH_OPT_THAW_MMIO:
195 ret = opal_pci_eeh_freeze_clear(phb->opal_id, pe_no,
196 OPAL_EEH_ACTION_CLEAR_FREEZE_MMIO);
198 pr_warning("%s: Failed to enable MMIO for "
199 "PHB#%x-PE#%x, err=%lld\n",
200 __func__, hose->global_number, pe_no, ret);
205 case EEH_OPT_THAW_DMA:
206 ret = opal_pci_eeh_freeze_clear(phb->opal_id, pe_no,
207 OPAL_EEH_ACTION_CLEAR_FREEZE_DMA);
209 pr_warning("%s: Failed to enable DMA for "
210 "PHB#%x-PE#%x, err=%lld\n",
211 __func__, hose->global_number, pe_no, ret);
217 pr_warning("%s: Invalid option %d\n", __func__, option);
225 * ioda_eeh_get_state - Retrieve the state of PE
228 * The PE's state should be retrieved from the PEEV, PEST
229 * IODA tables. Since the OPAL has exported the function
230 * to do it, it'd better to use that.
232 static int ioda_eeh_get_state(struct eeh_pe *pe)
239 struct pci_controller *hose = pe->phb;
240 struct pnv_phb *phb = hose->private_data;
243 * Sanity check on PE address. The PHB PE address should
246 if (pe->addr < 0 || pe->addr >= phb->ioda.total_pe) {
247 pr_err("%s: PE address %x out of range [0, %x] "
249 __func__, pe->addr, phb->ioda.total_pe,
250 hose->global_number);
251 return EEH_STATE_NOT_SUPPORT;
254 /* Retrieve PE status through OPAL */
256 ret = opal_pci_eeh_freeze_status(phb->opal_id, pe_no,
257 &fstate, &pcierr, NULL);
259 pr_err("%s: Failed to get EEH status on "
260 "PHB#%x-PE#%x\n, err=%lld\n",
261 __func__, hose->global_number, pe_no, ret);
262 return EEH_STATE_NOT_SUPPORT;
265 /* Check PHB status */
266 if (pe->type & EEH_PE_PHB) {
268 result &= ~EEH_STATE_RESET_ACTIVE;
270 if (pcierr != OPAL_EEH_PHB_ERROR) {
271 result |= EEH_STATE_MMIO_ACTIVE;
272 result |= EEH_STATE_DMA_ACTIVE;
273 result |= EEH_STATE_MMIO_ENABLED;
274 result |= EEH_STATE_DMA_ENABLED;
280 /* Parse result out */
283 case OPAL_EEH_STOPPED_NOT_FROZEN:
284 result &= ~EEH_STATE_RESET_ACTIVE;
285 result |= EEH_STATE_MMIO_ACTIVE;
286 result |= EEH_STATE_DMA_ACTIVE;
287 result |= EEH_STATE_MMIO_ENABLED;
288 result |= EEH_STATE_DMA_ENABLED;
290 case OPAL_EEH_STOPPED_MMIO_FREEZE:
291 result &= ~EEH_STATE_RESET_ACTIVE;
292 result |= EEH_STATE_DMA_ACTIVE;
293 result |= EEH_STATE_DMA_ENABLED;
295 case OPAL_EEH_STOPPED_DMA_FREEZE:
296 result &= ~EEH_STATE_RESET_ACTIVE;
297 result |= EEH_STATE_MMIO_ACTIVE;
298 result |= EEH_STATE_MMIO_ENABLED;
300 case OPAL_EEH_STOPPED_MMIO_DMA_FREEZE:
301 result &= ~EEH_STATE_RESET_ACTIVE;
303 case OPAL_EEH_STOPPED_RESET:
304 result |= EEH_STATE_RESET_ACTIVE;
306 case OPAL_EEH_STOPPED_TEMP_UNAVAIL:
307 result |= EEH_STATE_UNAVAILABLE;
309 case OPAL_EEH_STOPPED_PERM_UNAVAIL:
310 result |= EEH_STATE_NOT_SUPPORT;
313 pr_warning("%s: Unexpected EEH status 0x%x "
315 __func__, fstate, hose->global_number, pe_no);
321 static int ioda_eeh_pe_clear(struct eeh_pe *pe)
323 struct pci_controller *hose;
332 phb = pe->phb->private_data;
334 /* Clear the EEH error on the PE */
335 ret = opal_pci_eeh_freeze_clear(phb->opal_id,
336 pe_no, OPAL_EEH_ACTION_CLEAR_FREEZE_ALL);
338 pr_err("%s: Failed to clear EEH error for "
339 "PHB#%x-PE#%x, err=%lld\n",
340 __func__, hose->global_number, pe_no, ret);
345 * Read the PE state back and verify that the frozen
346 * state has been removed.
348 ret = opal_pci_eeh_freeze_status(phb->opal_id, pe_no,
349 &fstate, &pcierr, NULL);
351 pr_err("%s: Failed to get EEH status on "
352 "PHB#%x-PE#%x\n, err=%lld\n",
353 __func__, hose->global_number, pe_no, ret);
357 if (fstate != OPAL_EEH_STOPPED_NOT_FROZEN) {
358 pr_err("%s: Frozen state not cleared on "
359 "PHB#%x-PE#%x, sts=%x\n",
360 __func__, hose->global_number, pe_no, fstate);
367 static s64 ioda_eeh_phb_poll(struct pnv_phb *phb)
369 s64 rc = OPAL_HARDWARE;
372 rc = opal_pci_poll(phb->opal_id);
382 static int ioda_eeh_phb_reset(struct pci_controller *hose, int option)
384 struct pnv_phb *phb = hose->private_data;
385 s64 rc = OPAL_HARDWARE;
387 pr_debug("%s: Reset PHB#%x, option=%d\n",
388 __func__, hose->global_number, option);
390 /* Issue PHB complete reset request */
391 if (option == EEH_RESET_FUNDAMENTAL ||
392 option == EEH_RESET_HOT)
393 rc = opal_pci_reset(phb->opal_id,
396 else if (option == EEH_RESET_DEACTIVATE)
397 rc = opal_pci_reset(phb->opal_id,
399 OPAL_DEASSERT_RESET);
404 * Poll state of the PHB until the request is done
407 rc = ioda_eeh_phb_poll(phb);
409 if (rc != OPAL_SUCCESS)
415 static int ioda_eeh_root_reset(struct pci_controller *hose, int option)
417 struct pnv_phb *phb = hose->private_data;
418 s64 rc = OPAL_SUCCESS;
420 pr_debug("%s: Reset PHB#%x, option=%d\n",
421 __func__, hose->global_number, option);
424 * During the reset deassert time, we needn't care
425 * the reset scope because the firmware does nothing
426 * for fundamental or hot reset during deassert phase.
428 if (option == EEH_RESET_FUNDAMENTAL)
429 rc = opal_pci_reset(phb->opal_id,
430 OPAL_PCI_FUNDAMENTAL_RESET,
432 else if (option == EEH_RESET_HOT)
433 rc = opal_pci_reset(phb->opal_id,
436 else if (option == EEH_RESET_DEACTIVATE)
437 rc = opal_pci_reset(phb->opal_id,
439 OPAL_DEASSERT_RESET);
443 /* Poll state of the PHB until the request is done */
444 rc = ioda_eeh_phb_poll(phb);
446 if (rc != OPAL_SUCCESS)
452 static int ioda_eeh_bridge_reset(struct pci_controller *hose,
453 struct pci_dev *dev, int option)
457 pr_debug("%s: Reset device %04x:%02x:%02x.%01x with option %d\n",
458 __func__, hose->global_number, dev->bus->number,
459 PCI_SLOT(dev->devfn), PCI_FUNC(dev->devfn), option);
462 case EEH_RESET_FUNDAMENTAL:
464 pci_read_config_word(dev, PCI_BRIDGE_CONTROL, &ctrl);
465 ctrl |= PCI_BRIDGE_CTL_BUS_RESET;
466 pci_write_config_word(dev, PCI_BRIDGE_CONTROL, ctrl);
468 case EEH_RESET_DEACTIVATE:
469 pci_read_config_word(dev, PCI_BRIDGE_CONTROL, &ctrl);
470 ctrl &= ~PCI_BRIDGE_CTL_BUS_RESET;
471 pci_write_config_word(dev, PCI_BRIDGE_CONTROL, ctrl);
479 * ioda_eeh_reset - Reset the indicated PE
481 * @option: reset option
483 * Do reset on the indicated PE. For PCI bus sensitive PE,
484 * we need to reset the parent p2p bridge. The PHB has to
485 * be reinitialized if the p2p bridge is root bridge. For
486 * PCI device sensitive PE, we will try to reset the device
487 * through FLR. For now, we don't have OPAL APIs to do HARD
488 * reset yet, so all reset would be SOFT (HOT) reset.
490 static int ioda_eeh_reset(struct eeh_pe *pe, int option)
492 struct pci_controller *hose = pe->phb;
493 struct eeh_dev *edev;
498 * Anyway, we have to clear the problematic state for the
499 * corresponding PE. However, we needn't do it if the PE
500 * is PHB associated. That means the PHB is having fatal
501 * errors and it needs reset. Further more, the AIB interface
502 * isn't reliable any more.
504 if (!(pe->type & EEH_PE_PHB) &&
505 (option == EEH_RESET_HOT ||
506 option == EEH_RESET_FUNDAMENTAL)) {
507 ret = ioda_eeh_pe_clear(pe);
513 * The rules applied to reset, either fundamental or hot reset:
515 * We always reset the direct upstream bridge of the PE. If the
516 * direct upstream bridge isn't root bridge, we always take hot
517 * reset no matter what option (fundamental or hot) is. Otherwise,
518 * we should do the reset according to the required option.
520 if (pe->type & EEH_PE_PHB) {
521 ret = ioda_eeh_phb_reset(hose, option);
523 if (pe->type & EEH_PE_DEVICE) {
525 * If it's device PE, we didn't refer to the parent
526 * PCI bus yet. So we have to figure it out indirectly.
528 edev = list_first_entry(&pe->edevs,
529 struct eeh_dev, list);
530 dev = eeh_dev_to_pci_dev(edev);
531 dev = dev->bus->self;
534 * If it's bus PE, the parent PCI bus is already there
535 * and just pick it up.
541 * Do reset based on the fact that the direct upstream bridge
542 * is root bridge (port) or not.
544 if (dev->bus->number == 0)
545 ret = ioda_eeh_root_reset(hose, option);
547 ret = ioda_eeh_bridge_reset(hose, dev, option);
554 * ioda_eeh_get_log - Retrieve error log
556 * @severity: Severity level of the log
557 * @drv_log: buffer to store the log
558 * @len: space of the log buffer
560 * The function is used to retrieve error log from P7IOC.
562 static int ioda_eeh_get_log(struct eeh_pe *pe, int severity,
563 char *drv_log, unsigned long len)
567 struct pci_controller *hose = pe->phb;
568 struct pnv_phb *phb = hose->private_data;
570 spin_lock_irqsave(&phb->lock, flags);
572 ret = opal_pci_get_phb_diag_data2(phb->opal_id,
573 phb->diag.blob, PNV_PCI_DIAG_BUF_SIZE);
575 spin_unlock_irqrestore(&phb->lock, flags);
576 pr_warning("%s: Can't get log for PHB#%x-PE#%x (%lld)\n",
577 __func__, hose->global_number, pe->addr, ret);
582 * FIXME: We probably need log the error in somewhere.
583 * Lets make it up in future.
585 /* pr_info("%s", phb->diag.blob); */
587 spin_unlock_irqrestore(&phb->lock, flags);
593 * ioda_eeh_configure_bridge - Configure the PCI bridges for the indicated PE
596 * For particular PE, it might have included PCI bridges. In order
597 * to make the PE work properly, those PCI bridges should be configured
598 * correctly. However, we need do nothing on P7IOC since the reset
599 * function will do everything that should be covered by the function.
601 static int ioda_eeh_configure_bridge(struct eeh_pe *pe)
606 static void ioda_eeh_hub_diag_common(struct OpalIoP7IOCErrorData *data)
609 pr_info(" GEM XFIR: %016llx\n", data->gemXfir);
610 pr_info(" GEM RFIR: %016llx\n", data->gemRfir);
611 pr_info(" GEM RIRQFIR: %016llx\n", data->gemRirqfir);
612 pr_info(" GEM Mask: %016llx\n", data->gemMask);
613 pr_info(" GEM RWOF: %016llx\n", data->gemRwof);
616 pr_info(" LEM FIR: %016llx\n", data->lemFir);
617 pr_info(" LEM Error Mask: %016llx\n", data->lemErrMask);
618 pr_info(" LEM Action 0: %016llx\n", data->lemAction0);
619 pr_info(" LEM Action 1: %016llx\n", data->lemAction1);
620 pr_info(" LEM WOF: %016llx\n", data->lemWof);
623 static void ioda_eeh_hub_diag(struct pci_controller *hose)
625 struct pnv_phb *phb = hose->private_data;
626 struct OpalIoP7IOCErrorData *data = &phb->diag.hub_diag;
629 rc = opal_pci_get_hub_diag_data(phb->hub_id, data, sizeof(*data));
630 if (rc != OPAL_SUCCESS) {
631 pr_warning("%s: Failed to get HUB#%llx diag-data (%ld)\n",
632 __func__, phb->hub_id, rc);
636 switch (data->type) {
637 case OPAL_P7IOC_DIAG_TYPE_RGC:
638 pr_info("P7IOC diag-data for RGC\n\n");
639 ioda_eeh_hub_diag_common(data);
640 pr_info(" RGC Status: %016llx\n", data->rgc.rgcStatus);
641 pr_info(" RGC LDCP: %016llx\n", data->rgc.rgcLdcp);
643 case OPAL_P7IOC_DIAG_TYPE_BI:
644 pr_info("P7IOC diag-data for BI %s\n\n",
645 data->bi.biDownbound ? "Downbound" : "Upbound");
646 ioda_eeh_hub_diag_common(data);
647 pr_info(" BI LDCP 0: %016llx\n", data->bi.biLdcp0);
648 pr_info(" BI LDCP 1: %016llx\n", data->bi.biLdcp1);
649 pr_info(" BI LDCP 2: %016llx\n", data->bi.biLdcp2);
650 pr_info(" BI Fence Status: %016llx\n", data->bi.biFenceStatus);
652 case OPAL_P7IOC_DIAG_TYPE_CI:
653 pr_info("P7IOC diag-data for CI Port %d\\nn",
655 ioda_eeh_hub_diag_common(data);
656 pr_info(" CI Port Status: %016llx\n", data->ci.ciPortStatus);
657 pr_info(" CI Port LDCP: %016llx\n", data->ci.ciPortLdcp);
659 case OPAL_P7IOC_DIAG_TYPE_MISC:
660 pr_info("P7IOC diag-data for MISC\n\n");
661 ioda_eeh_hub_diag_common(data);
663 case OPAL_P7IOC_DIAG_TYPE_I2C:
664 pr_info("P7IOC diag-data for I2C\n\n");
665 ioda_eeh_hub_diag_common(data);
668 pr_warning("%s: Invalid type of HUB#%llx diag-data (%d)\n",
669 __func__, phb->hub_id, data->type);
673 static void ioda_eeh_p7ioc_phb_diag(struct pci_controller *hose,
674 struct OpalIoPhbErrorCommon *common)
676 struct OpalIoP7IOCPhbErrorData *data;
679 data = (struct OpalIoP7IOCPhbErrorData *)common;
681 pr_info("P7IOC PHB#%x Diag-data (Version: %d)\n\n",
682 hose->global_number, common->version);
684 pr_info(" brdgCtl: %08x\n", data->brdgCtl);
686 pr_info(" portStatusReg: %08x\n", data->portStatusReg);
687 pr_info(" rootCmplxStatus: %08x\n", data->rootCmplxStatus);
688 pr_info(" busAgentStatus: %08x\n", data->busAgentStatus);
690 pr_info(" deviceStatus: %08x\n", data->deviceStatus);
691 pr_info(" slotStatus: %08x\n", data->slotStatus);
692 pr_info(" linkStatus: %08x\n", data->linkStatus);
693 pr_info(" devCmdStatus: %08x\n", data->devCmdStatus);
694 pr_info(" devSecStatus: %08x\n", data->devSecStatus);
696 pr_info(" rootErrorStatus: %08x\n", data->rootErrorStatus);
697 pr_info(" uncorrErrorStatus: %08x\n", data->uncorrErrorStatus);
698 pr_info(" corrErrorStatus: %08x\n", data->corrErrorStatus);
699 pr_info(" tlpHdr1: %08x\n", data->tlpHdr1);
700 pr_info(" tlpHdr2: %08x\n", data->tlpHdr2);
701 pr_info(" tlpHdr3: %08x\n", data->tlpHdr3);
702 pr_info(" tlpHdr4: %08x\n", data->tlpHdr4);
703 pr_info(" sourceId: %08x\n", data->sourceId);
705 pr_info(" errorClass: %016llx\n", data->errorClass);
706 pr_info(" correlator: %016llx\n", data->correlator);
707 pr_info(" p7iocPlssr: %016llx\n", data->p7iocPlssr);
708 pr_info(" p7iocCsr: %016llx\n", data->p7iocCsr);
709 pr_info(" lemFir: %016llx\n", data->lemFir);
710 pr_info(" lemErrorMask: %016llx\n", data->lemErrorMask);
711 pr_info(" lemWOF: %016llx\n", data->lemWOF);
712 pr_info(" phbErrorStatus: %016llx\n", data->phbErrorStatus);
713 pr_info(" phbFirstErrorStatus: %016llx\n", data->phbFirstErrorStatus);
714 pr_info(" phbErrorLog0: %016llx\n", data->phbErrorLog0);
715 pr_info(" phbErrorLog1: %016llx\n", data->phbErrorLog1);
716 pr_info(" mmioErrorStatus: %016llx\n", data->mmioErrorStatus);
717 pr_info(" mmioFirstErrorStatus: %016llx\n", data->mmioFirstErrorStatus);
718 pr_info(" mmioErrorLog0: %016llx\n", data->mmioErrorLog0);
719 pr_info(" mmioErrorLog1: %016llx\n", data->mmioErrorLog1);
720 pr_info(" dma0ErrorStatus: %016llx\n", data->dma0ErrorStatus);
721 pr_info(" dma0FirstErrorStatus: %016llx\n", data->dma0FirstErrorStatus);
722 pr_info(" dma0ErrorLog0: %016llx\n", data->dma0ErrorLog0);
723 pr_info(" dma0ErrorLog1: %016llx\n", data->dma0ErrorLog1);
724 pr_info(" dma1ErrorStatus: %016llx\n", data->dma1ErrorStatus);
725 pr_info(" dma1FirstErrorStatus: %016llx\n", data->dma1FirstErrorStatus);
726 pr_info(" dma1ErrorLog0: %016llx\n", data->dma1ErrorLog0);
727 pr_info(" dma1ErrorLog1: %016llx\n", data->dma1ErrorLog1);
729 for (i = 0; i < OPAL_P7IOC_NUM_PEST_REGS; i++) {
730 if ((data->pestA[i] >> 63) == 0 &&
731 (data->pestB[i] >> 63) == 0)
734 pr_info(" PE[%3d] PESTA: %016llx\n", i, data->pestA[i]);
735 pr_info(" PESTB: %016llx\n", data->pestB[i]);
739 static void ioda_eeh_phb3_phb_diag(struct pci_controller *hose,
740 struct OpalIoPhbErrorCommon *common)
742 struct OpalIoPhb3ErrorData *data;
745 data = (struct OpalIoPhb3ErrorData*)common;
746 pr_info("PHB3 PHB#%x Diag-data (Version: %d)\n\n",
747 hose->global_number, common->version);
749 pr_info(" brdgCtl: %08x\n", data->brdgCtl);
751 pr_info(" portStatusReg: %08x\n", data->portStatusReg);
752 pr_info(" rootCmplxStatus: %08x\n", data->rootCmplxStatus);
753 pr_info(" busAgentStatus: %08x\n", data->busAgentStatus);
755 pr_info(" deviceStatus: %08x\n", data->deviceStatus);
756 pr_info(" slotStatus: %08x\n", data->slotStatus);
757 pr_info(" linkStatus: %08x\n", data->linkStatus);
758 pr_info(" devCmdStatus: %08x\n", data->devCmdStatus);
759 pr_info(" devSecStatus: %08x\n", data->devSecStatus);
761 pr_info(" rootErrorStatus: %08x\n", data->rootErrorStatus);
762 pr_info(" uncorrErrorStatus: %08x\n", data->uncorrErrorStatus);
763 pr_info(" corrErrorStatus: %08x\n", data->corrErrorStatus);
764 pr_info(" tlpHdr1: %08x\n", data->tlpHdr1);
765 pr_info(" tlpHdr2: %08x\n", data->tlpHdr2);
766 pr_info(" tlpHdr3: %08x\n", data->tlpHdr3);
767 pr_info(" tlpHdr4: %08x\n", data->tlpHdr4);
768 pr_info(" sourceId: %08x\n", data->sourceId);
769 pr_info(" errorClass: %016llx\n", data->errorClass);
770 pr_info(" correlator: %016llx\n", data->correlator);
771 pr_info(" nFir: %016llx\n", data->nFir);
772 pr_info(" nFirMask: %016llx\n", data->nFirMask);
773 pr_info(" nFirWOF: %016llx\n", data->nFirWOF);
774 pr_info(" PhbPlssr: %016llx\n", data->phbPlssr);
775 pr_info(" PhbCsr: %016llx\n", data->phbCsr);
776 pr_info(" lemFir: %016llx\n", data->lemFir);
777 pr_info(" lemErrorMask: %016llx\n", data->lemErrorMask);
778 pr_info(" lemWOF: %016llx\n", data->lemWOF);
779 pr_info(" phbErrorStatus: %016llx\n", data->phbErrorStatus);
780 pr_info(" phbFirstErrorStatus: %016llx\n", data->phbFirstErrorStatus);
781 pr_info(" phbErrorLog0: %016llx\n", data->phbErrorLog0);
782 pr_info(" phbErrorLog1: %016llx\n", data->phbErrorLog1);
783 pr_info(" mmioErrorStatus: %016llx\n", data->mmioErrorStatus);
784 pr_info(" mmioFirstErrorStatus: %016llx\n", data->mmioFirstErrorStatus);
785 pr_info(" mmioErrorLog0: %016llx\n", data->mmioErrorLog0);
786 pr_info(" mmioErrorLog1: %016llx\n", data->mmioErrorLog1);
787 pr_info(" dma0ErrorStatus: %016llx\n", data->dma0ErrorStatus);
788 pr_info(" dma0FirstErrorStatus: %016llx\n", data->dma0FirstErrorStatus);
789 pr_info(" dma0ErrorLog0: %016llx\n", data->dma0ErrorLog0);
790 pr_info(" dma0ErrorLog1: %016llx\n", data->dma0ErrorLog1);
791 pr_info(" dma1ErrorStatus: %016llx\n", data->dma1ErrorStatus);
792 pr_info(" dma1FirstErrorStatus: %016llx\n", data->dma1FirstErrorStatus);
793 pr_info(" dma1ErrorLog0: %016llx\n", data->dma1ErrorLog0);
794 pr_info(" dma1ErrorLog1: %016llx\n", data->dma1ErrorLog1);
796 for (i = 0; i < OPAL_PHB3_NUM_PEST_REGS; i++) {
797 if ((data->pestA[i] >> 63) == 0 &&
798 (data->pestB[i] >> 63) == 0)
801 pr_info(" PE[%3d] PESTA: %016llx\n", i, data->pestA[i]);
802 pr_info(" PESTB: %016llx\n", data->pestB[i]);
806 static void ioda_eeh_phb_diag(struct pci_controller *hose)
808 struct pnv_phb *phb = hose->private_data;
809 struct OpalIoPhbErrorCommon *common;
812 rc = opal_pci_get_phb_diag_data2(phb->opal_id, phb->diag.blob,
813 PNV_PCI_DIAG_BUF_SIZE);
814 if (rc != OPAL_SUCCESS) {
815 pr_warning("%s: Failed to get diag-data for PHB#%x (%ld)\n",
816 __func__, hose->global_number, rc);
820 common = (struct OpalIoPhbErrorCommon *)phb->diag.blob;
821 switch (common->ioType) {
822 case OPAL_PHB_ERROR_DATA_TYPE_P7IOC:
823 ioda_eeh_p7ioc_phb_diag(hose, common);
825 case OPAL_PHB_ERROR_DATA_TYPE_PHB3:
826 ioda_eeh_phb3_phb_diag(hose, common);
829 pr_warning("%s: Unrecognized I/O chip %d\n",
830 __func__, common->ioType);
834 static int ioda_eeh_get_phb_pe(struct pci_controller *hose,
837 struct eeh_pe *phb_pe;
839 phb_pe = eeh_phb_pe_get(hose);
841 pr_warning("%s Can't find PE for PHB#%d\n",
842 __func__, hose->global_number);
850 static int ioda_eeh_get_pe(struct pci_controller *hose,
851 u16 pe_no, struct eeh_pe **pe)
853 struct eeh_pe *phb_pe, *dev_pe;
856 /* Find the PHB PE */
857 if (ioda_eeh_get_phb_pe(hose, &phb_pe))
860 /* Find the PE according to PE# */
861 memset(&dev, 0, sizeof(struct eeh_dev));
863 dev.pe_config_addr = pe_no;
864 dev_pe = eeh_pe_get(&dev);
866 pr_warning("%s: Can't find PE for PHB#%x - PE#%x\n",
867 __func__, hose->global_number, pe_no);
876 * ioda_eeh_next_error - Retrieve next error for EEH core to handle
877 * @pe: The affected PE
879 * The function is expected to be called by EEH core while it gets
880 * special EEH event (without binding PE). The function calls to
881 * OPAL APIs for next error to handle. The informational error is
882 * handled internally by platform. However, the dead IOC, dead PHB,
883 * fenced PHB and frozen PE should be handled by EEH core eventually.
885 static int ioda_eeh_next_error(struct eeh_pe **pe)
887 struct pci_controller *hose, *tmp;
890 u16 err_type, severity;
895 * While running here, it's safe to purge the event queue.
896 * And we should keep the cached OPAL notifier event sychronized
897 * between the kernel and firmware.
899 eeh_remove_event(NULL);
900 opal_notifier_update_evt(OPAL_EVENT_PCI_ERROR, 0x0ul);
902 list_for_each_entry_safe(hose, tmp, &hose_list, list_node) {
904 * If the subordinate PCI buses of the PHB has been
905 * removed, we needn't take care of it any more.
907 phb = hose->private_data;
908 if (phb->eeh_state & PNV_EEH_STATE_REMOVED)
911 rc = opal_pci_next_error(phb->opal_id,
912 &frozen_pe_no, &err_type, &severity);
914 /* If OPAL API returns error, we needn't proceed */
915 if (rc != OPAL_SUCCESS) {
916 pr_devel("%s: Invalid return value on "
917 "PHB#%x (0x%lx) from opal_pci_next_error",
918 __func__, hose->global_number, rc);
922 /* If the PHB doesn't have error, stop processing */
923 if (err_type == OPAL_EEH_NO_ERROR ||
924 severity == OPAL_EEH_SEV_NO_ERROR) {
925 pr_devel("%s: No error found on PHB#%x\n",
926 __func__, hose->global_number);
931 * Processing the error. We're expecting the error with
932 * highest priority reported upon multiple errors on the
935 pr_devel("%s: Error (%d, %d, %llu) on PHB#%x\n",
936 __func__, err_type, severity,
937 frozen_pe_no, hose->global_number);
939 case OPAL_EEH_IOC_ERROR:
940 if (severity == OPAL_EEH_SEV_IOC_DEAD) {
941 list_for_each_entry_safe(hose, tmp,
942 &hose_list, list_node) {
943 phb = hose->private_data;
944 phb->eeh_state |= PNV_EEH_STATE_REMOVED;
947 pr_err("EEH: dead IOC detected\n");
950 } else if (severity == OPAL_EEH_SEV_INF) {
951 pr_info("EEH: IOC informative error "
953 ioda_eeh_hub_diag(hose);
957 case OPAL_EEH_PHB_ERROR:
958 if (severity == OPAL_EEH_SEV_PHB_DEAD) {
959 if (ioda_eeh_get_phb_pe(hose, pe))
962 pr_err("EEH: dead PHB#%x detected\n",
963 hose->global_number);
964 phb->eeh_state |= PNV_EEH_STATE_REMOVED;
967 } else if (severity == OPAL_EEH_SEV_PHB_FENCED) {
968 if (ioda_eeh_get_phb_pe(hose, pe))
971 pr_err("EEH: fenced PHB#%x detected\n",
972 hose->global_number);
975 } else if (severity == OPAL_EEH_SEV_INF) {
976 pr_info("EEH: PHB#%x informative error "
978 hose->global_number);
979 ioda_eeh_phb_diag(hose);
983 case OPAL_EEH_PE_ERROR:
984 if (ioda_eeh_get_pe(hose, frozen_pe_no, pe))
987 pr_err("EEH: Frozen PE#%x on PHB#%x detected\n",
988 (*pe)->addr, (*pe)->phb->global_number);
999 struct pnv_eeh_ops ioda_eeh_ops = {
1000 .post_init = ioda_eeh_post_init,
1001 .set_option = ioda_eeh_set_option,
1002 .get_state = ioda_eeh_get_state,
1003 .reset = ioda_eeh_reset,
1004 .get_log = ioda_eeh_get_log,
1005 .configure_bridge = ioda_eeh_configure_bridge,
1006 .next_error = ioda_eeh_next_error