powerpc/powernv/pci: Add support for PHB4 diagnostics

As with P7IOC and PHB3, add kernel-side support for decoding and printing
diagnostic data for PHB4.

Signed-off-by: Russell Currey <ruscur@russell.cc>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

2 files changed, 178 insertions, 2 deletions

diff --git a/arch/powerpc/include/asm/opal-api.h b/arch/powerpc/include/asm/opal-api.h
index cb3e6242a78c..0b543f0f54f5 100644
--- a/arch/powerpc/include/asm/opal-api.h
+++ b/arch/powerpc/include/asm/opal-api.h
@@ -667,12 +667,14 @@ enum {
 
 enum {
 	OPAL_PHB_ERROR_DATA_TYPE_P7IOC = 1,
-	OPAL_PHB_ERROR_DATA_TYPE_PHB3 = 2
+	OPAL_PHB_ERROR_DATA_TYPE_PHB3 = 2,
+	OPAL_PHB_ERROR_DATA_TYPE_PHB4 = 3
 };
 
 enum {
 	OPAL_P7IOC_NUM_PEST_REGS = 128,
-	OPAL_PHB3_NUM_PEST_REGS = 256
+	OPAL_PHB3_NUM_PEST_REGS = 256,
+	OPAL_PHB4_NUM_PEST_REGS = 512
 };
 
 struct OpalIoPhbErrorCommon {
@@ -802,6 +804,75 @@ struct OpalIoPhb3ErrorData {
 	__be64 pestB[OPAL_PHB3_NUM_PEST_REGS];
 };
 
+struct OpalIoPhb4ErrorData {
+	struct OpalIoPhbErrorCommon common;
+
+	__be32 brdgCtl;
+
+	/* PHB4 cfg regs */
+	__be32 deviceStatus;
+	__be32 slotStatus;
+	__be32 linkStatus;
+	__be32 devCmdStatus;
+	__be32 devSecStatus;
+
+	/* cfg AER regs */
+	__be32 rootErrorStatus;
+	__be32 uncorrErrorStatus;
+	__be32 corrErrorStatus;
+	__be32 tlpHdr1;
+	__be32 tlpHdr2;
+	__be32 tlpHdr3;
+	__be32 tlpHdr4;
+	__be32 sourceId;
+
+	/* PHB4 ETU Error Regs */
+	__be64 nFir;				/* 000 */
+	__be64 nFirMask;			/* 003 */
+	__be64 nFirWOF;				/* 008 */
+	__be64 phbPlssr;			/* 120 */
+	__be64 phbCsr;				/* 110 */
+	__be64 lemFir;				/* C00 */
+	__be64 lemErrorMask;			/* C18 */
+	__be64 lemWOF;				/* C40 */
+	__be64 phbErrorStatus;			/* C80 */
+	__be64 phbFirstErrorStatus;		/* C88 */
+	__be64 phbErrorLog0;			/* CC0 */
+	__be64 phbErrorLog1;			/* CC8 */
+	__be64 phbTxeErrorStatus;		/* D00 */
+	__be64 phbTxeFirstErrorStatus;		/* D08 */
+	__be64 phbTxeErrorLog0;			/* D40 */
+	__be64 phbTxeErrorLog1;			/* D48 */
+	__be64 phbRxeArbErrorStatus;		/* D80 */
+	__be64 phbRxeArbFirstErrorStatus;	/* D88 */
+	__be64 phbRxeArbErrorLog0;		/* DC0 */
+	__be64 phbRxeArbErrorLog1;		/* DC8 */
+	__be64 phbRxeMrgErrorStatus;		/* E00 */
+	__be64 phbRxeMrgFirstErrorStatus;	/* E08 */
+	__be64 phbRxeMrgErrorLog0;		/* E40 */
+	__be64 phbRxeMrgErrorLog1;		/* E48 */
+	__be64 phbRxeTceErrorStatus;		/* E80 */
+	__be64 phbRxeTceFirstErrorStatus;	/* E88 */
+	__be64 phbRxeTceErrorLog0;		/* EC0 */
+	__be64 phbRxeTceErrorLog1;		/* EC8 */
+
+	/* PHB4 REGB Error Regs */
+	__be64 phbPblErrorStatus;		/* 1900 */
+	__be64 phbPblFirstErrorStatus;		/* 1908 */
+	__be64 phbPblErrorLog0;			/* 1940 */
+	__be64 phbPblErrorLog1;			/* 1948 */
+	__be64 phbPcieDlpErrorLog1;		/* 1AA0 */
+	__be64 phbPcieDlpErrorLog2;		/* 1AA8 */
+	__be64 phbPcieDlpErrorStatus;		/* 1AB0 */
+	__be64 phbRegbErrorStatus;		/* 1C00 */
+	__be64 phbRegbFirstErrorStatus;		/* 1C08 */
+	__be64 phbRegbErrorLog0;		/* 1C40 */
+	__be64 phbRegbErrorLog1;		/* 1C48 */
+
+	__be64 pestA[OPAL_PHB4_NUM_PEST_REGS];
+	__be64 pestB[OPAL_PHB4_NUM_PEST_REGS];
+};
+
 enum {
 	OPAL_REINIT_CPUS_HILE_BE	= (1 << 0),
 	OPAL_REINIT_CPUS_HILE_LE	= (1 << 1),
diff --git a/arch/powerpc/platforms/powernv/pci.c b/arch/powerpc/platforms/powernv/pci.c
index 209ad47a3383..7905d179d036 100644
--- a/arch/powerpc/platforms/powernv/pci.c
+++ b/arch/powerpc/platforms/powernv/pci.c
@@ -426,6 +426,108 @@ static void pnv_pci_dump_phb3_diag_data(struct pci_controller *hose,
 	pnv_pci_dump_pest(data->pestA, data->pestB, OPAL_PHB3_NUM_PEST_REGS);
 }
 
+static void pnv_pci_dump_phb4_diag_data(struct pci_controller *hose,
+					struct OpalIoPhbErrorCommon *common)
+{
+	struct OpalIoPhb4ErrorData *data;
+
+	data = (struct OpalIoPhb4ErrorData*)common;
+	pr_info("PHB4 PHB#%d Diag-data (Version: %d)\n",
+		hose->global_number, be32_to_cpu(common->version));
+	if (data->brdgCtl)
+		pr_info("brdgCtl:    %08x\n",
+			be32_to_cpu(data->brdgCtl));
+	if (data->deviceStatus || data->slotStatus   ||
+	    data->linkStatus   || data->devCmdStatus ||
+	    data->devSecStatus)
+		pr_info("RootSts:    %08x %08x %08x %08x %08x\n",
+			be32_to_cpu(data->deviceStatus),
+			be32_to_cpu(data->slotStatus),
+			be32_to_cpu(data->linkStatus),
+			be32_to_cpu(data->devCmdStatus),
+			be32_to_cpu(data->devSecStatus));
+	if (data->rootErrorStatus || data->uncorrErrorStatus ||
+	    data->corrErrorStatus)
+		pr_info("RootErrSts: %08x %08x %08x\n",
+			be32_to_cpu(data->rootErrorStatus),
+			be32_to_cpu(data->uncorrErrorStatus),
+			be32_to_cpu(data->corrErrorStatus));
+	if (data->tlpHdr1 || data->tlpHdr2 ||
+	    data->tlpHdr3 || data->tlpHdr4)
+		pr_info("RootErrLog: %08x %08x %08x %08x\n",
+			be32_to_cpu(data->tlpHdr1),
+			be32_to_cpu(data->tlpHdr2),
+			be32_to_cpu(data->tlpHdr3),
+			be32_to_cpu(data->tlpHdr4));
+	if (data->sourceId)
+		pr_info("sourceId:   %08x\n", be32_to_cpu(data->sourceId));
+	if (data->nFir)
+		pr_info("nFir:       %016llx %016llx %016llx\n",
+			be64_to_cpu(data->nFir),
+			be64_to_cpu(data->nFirMask),
+			be64_to_cpu(data->nFirWOF));
+	if (data->phbPlssr || data->phbCsr)
+		pr_info("PhbSts:     %016llx %016llx\n",
+			be64_to_cpu(data->phbPlssr),
+			be64_to_cpu(data->phbCsr));
+	if (data->lemFir)
+		pr_info("Lem:        %016llx %016llx %016llx\n",
+			be64_to_cpu(data->lemFir),
+			be64_to_cpu(data->lemErrorMask),
+			be64_to_cpu(data->lemWOF));
+	if (data->phbErrorStatus)
+		pr_info("PhbErr:     %016llx %016llx %016llx %016llx\n",
+			be64_to_cpu(data->phbErrorStatus),
+			be64_to_cpu(data->phbFirstErrorStatus),
+			be64_to_cpu(data->phbErrorLog0),
+			be64_to_cpu(data->phbErrorLog1));
+	if (data->phbTxeErrorStatus)
+		pr_info("PhbTxeErr:  %016llx %016llx %016llx %016llx\n",
+			be64_to_cpu(data->phbTxeErrorStatus),
+			be64_to_cpu(data->phbTxeFirstErrorStatus),
+			be64_to_cpu(data->phbTxeErrorLog0),
+			be64_to_cpu(data->phbTxeErrorLog1));
+	if (data->phbRxeArbErrorStatus)
+		pr_info("RxeArbErr:  %016llx %016llx %016llx %016llx\n",
+			be64_to_cpu(data->phbRxeArbErrorStatus),
+			be64_to_cpu(data->phbRxeArbFirstErrorStatus),
+			be64_to_cpu(data->phbRxeArbErrorLog0),
+			be64_to_cpu(data->phbRxeArbErrorLog1));
+	if (data->phbRxeMrgErrorStatus)
+		pr_info("RxeMrgErr:  %016llx %016llx %016llx %016llx\n",
+			be64_to_cpu(data->phbRxeMrgErrorStatus),
+			be64_to_cpu(data->phbRxeMrgFirstErrorStatus),
+			be64_to_cpu(data->phbRxeMrgErrorLog0),
+			be64_to_cpu(data->phbRxeMrgErrorLog1));
+	if (data->phbRxeTceErrorStatus)
+		pr_info("RxeTceErr:  %016llx %016llx %016llx %016llx\n",
+			be64_to_cpu(data->phbRxeTceErrorStatus),
+			be64_to_cpu(data->phbRxeTceFirstErrorStatus),
+			be64_to_cpu(data->phbRxeTceErrorLog0),
+			be64_to_cpu(data->phbRxeTceErrorLog1));
+
+	if (data->phbPblErrorStatus)
+		pr_info("PblErr:     %016llx %016llx %016llx %016llx\n",
+			be64_to_cpu(data->phbPblErrorStatus),
+			be64_to_cpu(data->phbPblFirstErrorStatus),
+			be64_to_cpu(data->phbPblErrorLog0),
+			be64_to_cpu(data->phbPblErrorLog1));
+	if (data->phbPcieDlpErrorStatus)
+		pr_info("PcieDlp:    %016llx %016llx %016llx\n",
+			be64_to_cpu(data->phbPcieDlpErrorLog1),
+			be64_to_cpu(data->phbPcieDlpErrorLog2),
+			be64_to_cpu(data->phbPcieDlpErrorStatus));
+	if (data->phbRegbErrorStatus)
+		pr_info("RegbErr:    %016llx %016llx %016llx %016llx\n",
+			be64_to_cpu(data->phbRegbErrorStatus),
+			be64_to_cpu(data->phbRegbFirstErrorStatus),
+			be64_to_cpu(data->phbRegbErrorLog0),
+			be64_to_cpu(data->phbRegbErrorLog1));
+
+
+	pnv_pci_dump_pest(data->pestA, data->pestB, OPAL_PHB4_NUM_PEST_REGS);
+}
+
 void pnv_pci_dump_phb_diag_data(struct pci_controller *hose,
 				unsigned char *log_buff)
 {
@@ -442,6 +544,9 @@ void pnv_pci_dump_phb_diag_data(struct pci_controller *hose,
 	case OPAL_PHB_ERROR_DATA_TYPE_PHB3:
 		pnv_pci_dump_phb3_diag_data(hose, common);
 		break;
+	case OPAL_PHB_ERROR_DATA_TYPE_PHB4:
+		pnv_pci_dump_phb4_diag_data(hose, common);
+		break;
 	default:
 		pr_warn("%s: Unrecognized ioType %d\n",
 			__func__, be32_to_cpu(common->ioType));