Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
powerpc/powernv: Dump PHB diag-data immediately
The PHB diag-data is important to help locating the root cause for
EEH errors such as frozen PE or fenced PHB. However, the EEH core
enables IO path by clearing part of HW registers before collecting
this data causing it to be corrupted.
This patch fixes this by dumping the PHB diag-data immediately when
frozen/fenced state on PE or PHB is detected for the first time in
eeh_ops::get_state() or next_error() backend.
Signed-off-by: Gavin Shan <shangw@linux.vnet.ibm.com>
CC: <stable@vger.kernel.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
authored by
Gavin Shan
and committed by
Benjamin Herrenschmidt
94716604
573ebfa6
+43
-53
+43
-53
arch/powerpc/platforms/powernv/eeh-ioda.c
+43
-53
arch/powerpc/platforms/powernv/eeh-ioda.c
···
114
114
ioda_eeh_inbB_dbgfs_set, "0x%llx\n");
115
115
#endif /* CONFIG_DEBUG_FS */
116
116
117
+
117
118
/**
118
119
* ioda_eeh_post_init - Chip dependent post initialization
119
120
* @hose: PCI controller
···
222
221
return ret;
223
222
}
224
223
224
+
static void ioda_eeh_phb_diag(struct pci_controller *hose)
225
+
{
226
+
struct pnv_phb *phb = hose->private_data;
227
+
long rc;
228
+
229
+
rc = opal_pci_get_phb_diag_data2(phb->opal_id, phb->diag.blob,
230
+
PNV_PCI_DIAG_BUF_SIZE);
231
+
if (rc != OPAL_SUCCESS) {
232
+
pr_warning("%s: Failed to get diag-data for PHB#%x (%ld)\n",
233
+
__func__, hose->global_number, rc);
234
+
return;
235
+
}
236
+
237
+
pnv_pci_dump_phb_diag_data(hose, phb->diag.blob);
238
+
}
239
+
225
240
/**
226
241
* ioda_eeh_get_state - Retrieve the state of PE
227
242
* @pe: EEH PE
···
289
272
result |= EEH_STATE_DMA_ACTIVE;
290
273
result |= EEH_STATE_MMIO_ENABLED;
291
274
result |= EEH_STATE_DMA_ENABLED;
275
+
} else if (!(pe->state & EEH_PE_ISOLATED)) {
276
+
eeh_pe_state_mark(pe, EEH_PE_ISOLATED);
277
+
ioda_eeh_phb_diag(hose);
292
278
}
293
279
294
280
return result;
···
333
313
pr_warning("%s: Unexpected EEH status 0x%x "
334
314
"on PHB#%x-PE#%x\n",
335
315
__func__, fstate, hose->global_number, pe_no);
316
+
}
317
+
318
+
/* Dump PHB diag-data for frozen PE */
319
+
if (result != EEH_STATE_NOT_SUPPORT &&
320
+
(result & (EEH_STATE_MMIO_ACTIVE | EEH_STATE_DMA_ACTIVE)) !=
321
+
(EEH_STATE_MMIO_ACTIVE | EEH_STATE_DMA_ACTIVE) &&
322
+
!(pe->state & EEH_PE_ISOLATED)) {
323
+
eeh_pe_state_mark(pe, EEH_PE_ISOLATED);
324
+
ioda_eeh_phb_diag(hose);
336
325
}
337
326
338
327
return result;
···
559
530
}
560
531
561
532
/**
562
-
* ioda_eeh_get_log - Retrieve error log
563
-
* @pe: EEH PE
564
-
* @severity: Severity level of the log
565
-
* @drv_log: buffer to store the log
566
-
* @len: space of the log buffer
567
-
*
568
-
* The function is used to retrieve error log from P7IOC.
569
-
*/
570
-
static int ioda_eeh_get_log(struct eeh_pe *pe, int severity,
571
-
char *drv_log, unsigned long len)
572
-
{
573
-
s64 ret;
574
-
unsigned long flags;
575
-
struct pci_controller *hose = pe->phb;
576
-
struct pnv_phb *phb = hose->private_data;
577
-
578
-
spin_lock_irqsave(&phb->lock, flags);
579
-
580
-
ret = opal_pci_get_phb_diag_data2(phb->opal_id,
581
-
phb->diag.blob, PNV_PCI_DIAG_BUF_SIZE);
582
-
if (ret) {
583
-
spin_unlock_irqrestore(&phb->lock, flags);
584
-
pr_warning("%s: Can't get log for PHB#%x-PE#%x (%lld)\n",
585
-
__func__, hose->global_number, pe->addr, ret);
586
-
return -EIO;
587
-
}
588
-
589
-
/* The PHB diag-data is always indicative */
590
-
pnv_pci_dump_phb_diag_data(hose, phb->diag.blob);
591
-
592
-
spin_unlock_irqrestore(&phb->lock, flags);
593
-
594
-
return 0;
595
-
}
596
-
597
-
/**
598
533
* ioda_eeh_configure_bridge - Configure the PCI bridges for the indicated PE
599
534
* @pe: EEH PE
600
535
*
···
637
644
pr_warning("%s: Invalid type of HUB#%llx diag-data (%d)\n",
638
645
__func__, phb->hub_id, data->type);
639
646
}
640
-
}
641
-
642
-
static void ioda_eeh_phb_diag(struct pci_controller *hose)
643
-
{
644
-
struct pnv_phb *phb = hose->private_data;
645
-
long rc;
646
-
647
-
rc = opal_pci_get_phb_diag_data2(phb->opal_id, phb->diag.blob,
648
-
PNV_PCI_DIAG_BUF_SIZE);
649
-
if (rc != OPAL_SUCCESS) {
650
-
pr_warning("%s: Failed to get diag-data for PHB#%x (%ld)\n",
651
-
__func__, hose->global_number, rc);
652
-
return;
653
-
}
654
-
655
-
pnv_pci_dump_phb_diag_data(hose, phb->diag.blob);
656
647
}
657
648
658
649
static int ioda_eeh_get_phb_pe(struct pci_controller *hose,
···
812
835
}
813
836
814
837
/*
838
+
* EEH core will try recover from fenced PHB or
839
+
* frozen PE. In the time for frozen PE, EEH core
840
+
* enable IO path for that before collecting logs,
841
+
* but it ruins the site. So we have to dump the
842
+
* log in advance here.
843
+
*/
844
+
if ((ret == EEH_NEXT_ERR_FROZEN_PE ||
845
+
ret == EEH_NEXT_ERR_FENCED_PHB) &&
846
+
!((*pe)->state & EEH_PE_ISOLATED)) {
847
+
eeh_pe_state_mark(*pe, EEH_PE_ISOLATED);
848
+
ioda_eeh_phb_diag(hose);
849
+
}
850
+
851
+
/*
815
852
* If we have no errors on the specific PHB or only
816
853
* informative error there, we continue poking it.
817
854
* Otherwise, we need actions to be taken by upper
···
843
852
.set_option = ioda_eeh_set_option,
844
853
.get_state = ioda_eeh_get_state,
845
854
.reset = ioda_eeh_reset,
846
-
.get_log = ioda_eeh_get_log,
847
855
.configure_bridge = ioda_eeh_configure_bridge,
848
856
.next_error = ioda_eeh_next_error
849
857
};