[PATCH 1/3] powerpc/eeh: Dump PCI config space for all child devices

[PATCH 1/3] powerpc/eeh: Dump PCI config space for all child devices

[Gavin Shan]

The PEs can be organized as nested. Current implementation doesn't
dump PCI config space for subordinate devices of child PEs. However,
the frozen PE could be caused by those subordinate devices of its
child PEs.

The patch dumps PCI config space for all subordinate devices of the
problematic PE.

Signed-off-by: Gavin Shan <gwshan@linux.vnet.ibm.com>
---
 arch/powerpc/kernel/eeh.c | 35 ++++++++++++++++++++---------------
 1 file changed, 20 insertions(+), 15 deletions(-)

diff --git a/arch/powerpc/kernel/eeh.c b/arch/powerpc/kernel/eeh.c
index 6c88d781..03d75cb 100644
--- a/arch/powerpc/kernel/eeh.c
+++ b/arch/powerpc/kernel/eeh.c
@@ -117,7 +117,7 @@ static DEFINE_MUTEX(eeh_dev_mutex);
  * not dynamically alloced, so that it ends up in RMO where RTAS
  * can access it.
  */
-#define EEH_PCI_REGS_LOG_LEN 4096
+#define EEH_PCI_REGS_LOG_LEN 8192
 static unsigned char pci_regs_buf[EEH_PCI_REGS_LOG_LEN];
 
 /*
@@ -148,16 +148,12 @@ static int __init eeh_setup(char *str)
 }
 __setup("eeh=", eeh_setup);
 
-/**
- * eeh_gather_pci_data - Copy assorted PCI config space registers to buff
- * @edev: device to report data for
- * @buf: point to buffer in which to log
- * @len: amount of room in buffer
- *
- * This routine captures assorted PCI configuration space data,
- * and puts them into a buffer for RTAS error logging.
+/*
+ * This routine captures assorted PCI configuration space data
+ * for the indicated PCI device, and puts them into a buffer
+ * for RTAS error logging.
  */
-static size_t eeh_gather_pci_data(struct eeh_dev *edev, char *buf, size_t len)
+static size_t eeh_dump_dev_log(struct eeh_dev *edev, char *buf, size_t len)
 {
 	struct device_node *dn = eeh_dev_to_of_node(edev);
 	u32 cfg;
@@ -255,6 +251,19 @@ static size_t eeh_gather_pci_data(struct eeh_dev *edev, char *buf, size_t len)
 	return n;
 }
 
+static void *eeh_dump_pe_log(void *data, void *flag)
+{
+	struct eeh_pe *pe = data;
+	struct eeh_dev *edev, *tmp;
+	size_t *plen = flag;
+
+	eeh_pe_for_each_dev(pe, edev, tmp)
+		*plen += eeh_dump_dev_log(edev, pci_regs_buf + *plen,
+					  EEH_PCI_REGS_LOG_LEN - *plen);
+
+	return NULL;
+}
+
 /**
  * eeh_slot_error_detail - Generate combined log including driver log and error log
  * @pe: EEH PE
@@ -268,7 +277,6 @@ static size_t eeh_gather_pci_data(struct eeh_dev *edev, char *buf, size_t len)
 void eeh_slot_error_detail(struct eeh_pe *pe, int severity)
 {
 	size_t loglen = 0;
-	struct eeh_dev *edev, *tmp;
 
 	/*
 	 * When the PHB is fenced or dead, it's pointless to collect
@@ -286,10 +294,7 @@ void eeh_slot_error_detail(struct eeh_pe *pe, int severity)
 		eeh_pe_restore_bars(pe);
 
 		pci_regs_buf[0] = 0;
-		eeh_pe_for_each_dev(pe, edev, tmp) {
-			loglen += eeh_gather_pci_data(edev, pci_regs_buf + loglen,
-						      EEH_PCI_REGS_LOG_LEN - loglen);
-		}
+		eeh_pe_traverse(pe, eeh_dump_pe_log, &loglen);
 	}
 
 	eeh_ops->get_log(pe, severity, pci_regs_buf, loglen);
-- 
1.8.3.2

[PATCH 2/3] powerpc/powernv: Fetch frozen PE on top level

[Gavin Shan]

It should have been part of commit 1ad7a72c5 ("powerpc/eeh: Report
frozen parent PE prior to child PE"). There are 2 ways to report
EEH errors: proactively polling triggered by PCI config or IO
accesses, or interrupt driven event. We missed to report and handle
parent frozen PE prior to child frozen PE for the later case on
PowerNV platform.

Signed-off-by: Gavin Shan <gwshan@linux.vnet.ibm.com>
---
 arch/powerpc/platforms/powernv/eeh-ioda.c | 48 ++++++++++++++++++++++---------
 1 file changed, 34 insertions(+), 14 deletions(-)

diff --git a/arch/powerpc/platforms/powernv/eeh-ioda.c b/arch/powerpc/platforms/powernv/eeh-ioda.c
index f248586..57de63c 100644
--- a/arch/powerpc/platforms/powernv/eeh-ioda.c
+++ b/arch/powerpc/platforms/powernv/eeh-ioda.c
@@ -880,14 +880,12 @@ static int ioda_eeh_get_pe(struct pci_controller *hose,
 	 * the master PE because slave PE is invisible
 	 * to EEH core.
 	 */
-	if (phb->get_pe_state) {
-		pnv_pe = &phb->ioda.pe_array[pe_no];
-		if (pnv_pe->flags & PNV_IODA_PE_SLAVE) {
-			pnv_pe = pnv_pe->master;
-			WARN_ON(!pnv_pe ||
-				!(pnv_pe->flags & PNV_IODA_PE_MASTER));
-			pe_no = pnv_pe->pe_number;
-		}
+	pnv_pe = &phb->ioda.pe_array[pe_no];
+	if (pnv_pe->flags & PNV_IODA_PE_SLAVE) {
+		pnv_pe = pnv_pe->master;
+		WARN_ON(!pnv_pe ||
+			!(pnv_pe->flags & PNV_IODA_PE_MASTER));
+		pe_no = pnv_pe->pe_number;
 	}
 
 	/* Find the PE according to PE# */
@@ -898,15 +896,37 @@ static int ioda_eeh_get_pe(struct pci_controller *hose,
 	if (!dev_pe)
 		return -EEXIST;
 
-	/*
-	 * At this point, we're sure the compound PE should
-	 * be put into frozen state.
-	 */
+	/* Freeze the (compound) PE */
 	*pe = dev_pe;
-	if (phb->freeze_pe &&
-	    !(dev_pe->state & EEH_PE_ISOLATED))
+	if (!(dev_pe->state & EEH_PE_ISOLATED))
 		phb->freeze_pe(phb, pe_no);
 
+	/*
+	 * At this point, we're sure the (compound) PE should
+	 * have been frozen. However, we still need poke until
+	 * hitting the frozen PE on top level.
+	 */
+	dev_pe = dev_pe->parent;
+	while (dev_pe && !(dev_pe->type & EEH_PE_PHB)) {
+		int ret;
+		int active_flags = (EEH_STATE_MMIO_ACTIVE |
+				    EEH_STATE_DMA_ACTIVE);
+
+		ret = eeh_ops->get_state(dev_pe, NULL);
+		if (ret <= 0 || (ret & active_flags) == active_flags) {
+			dev_pe = dev_pe->parent;
+			continue;
+		}
+
+		/* Frozen parent PE */
+		*pe = dev_pe;
+		if (!(dev_pe->state & EEH_PE_ISOLATED))
+			phb->freeze_pe(phb, dev_pe->addr);
+
+		/* Next one */
+		dev_pe = dev_pe->parent;
+	}
+
 	return 0;
 }
 
-- 
1.8.3.2