[PATCH] alpha: common ev6/ev7 machine check handler

Here is new ev6/ev7 machine check parser, which is a part of Jeff's
titan/marvel patchset.
Instead of fixing nautilus machine check handler, I've converted it
to use this new code.  Works fine.
Also, the CONFIG_VERBOSE_MCHECK is good for cia/pyxis - the detailed
error reporting is completely useless unless you have faulty CPU or RAM.

Ivan.

arch/alpha/Kconfig

@@ -522,6 +522,9 @@ config ALPHA_LARGE_VMALLOC
 
 	  Say N unless you know you need gobs and gobs of vmalloc space.
 
+config VERBOSE_MCHECK
+	bool "Verbose Machine Checks"
+
 source "drivers/pci/Kconfig"
 
 config HOTPLUG

arch/alpha/defconfig

@@ -59,6 +59,7 @@ CONFIG_ALPHA_GENERIC=y
 # CONFIG_ALPHA_TAKARA is not set
 # CONFIG_ALPHA_TITAN is not set
 # CONFIG_ALPHA_WILDFIRE is not set
+CONFIG_VERBOSE_MCHECK=y
 CONFIG_ISA=y
 CONFIG_EISA=y
 # CONFIG_SBUS is not set

arch/alpha/kernel/Makefile

@@ -10,7 +10,7 @@ export-objs := alpha_ksyms.o
 
 obj-y    := entry.o traps.o process.o init_task.o osf_sys.o irq.o \
 	    irq_alpha.o signal.o setup.o ptrace.o time.o semaphore.o \
-	    alpha_ksyms.o systbls.o
+	    alpha_ksyms.o systbls.o err_common.o
 
 #
 # FIXME!

arch/alpha/kernel/core_cia.c

@@ -755,6 +755,7 @@ cia_pci_clr_err(void)
 	*(vip)CIA_IOC_CIA_ERR;		/* re-read to force write.  */
 }
 
+#ifdef CONFIG_VERBOSE_MCHECK
 static void
 cia_decode_pci_error(struct el_CIA_sysdata_mcheck *cia, const char *msg)
 {
@@ -1022,13 +1023,13 @@ cia_decode_parity_error(struct el_CIA_sysdata_mcheck *cia)
 	printk(KERN_CRIT "  Command: %s, Parity bit: %d\n", cmd, par);
 	printk(KERN_CRIT "  Address: %#010lx, Mask: %#lx\n", addr, mask);
 }
+#endif
 
 static int
 cia_decode_mchk(unsigned long la_ptr)
 {
 	struct el_common *com;
 	struct el_CIA_sysdata_mcheck *cia;
-	int which;
 
 	com = (void *)la_ptr;
 	cia = (void *)(la_ptr + com->sys_offset);
@@ -1036,8 +1037,8 @@ cia_decode_mchk(unsigned long la_ptr)
 	if ((cia->cia_err & CIA_ERR_VALID) == 0)
 		return 0;
 
-	which = cia->cia_err & 0xfff;
-	switch (ffs(which) - 1) {
+#ifdef CONFIG_VERBOSE_MCHECK
+	switch (ffs(cia->cia_err & 0xfff) - 1) {
 	case 0: /* CIA_ERR_COR_ERR */
 		cia_decode_ecc_error(cia, "Corrected ECC error");
 		break;
@@ -1109,6 +1110,7 @@ cia_decode_mchk(unsigned long la_ptr)
 	if (cia->cia_err & CIA_ERR_LOST_IOA_TIMEOUT)
 		printk(KERN_CRIT "CIA lost machine check: "
 		       "I/O timeout\n");
+#endif
 
 	return 1;
 }

arch/alpha/kernel/err_common.c

+/*
+ *	linux/arch/alpha/kernel/err_common.c
+ *
+ *	Copyright (C) 2000 Jeff Wiedemeier (Compaq Computer Corporation)
+ *
+ *	Error handling code supporting Alpha systems
+ */
+
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/sched.h>
+
+#include <asm/io.h>
+#include <asm/hwrpb.h>
+#include <asm/smp.h>
+
+#include "err_impl.h"
+#include "proto.h"
+
+/*
+ * err_print_prefix -- error handling print routines should prefix
+ * all prints with this
+ */
+char *err_print_prefix = KERN_NOTICE;
+
+/*
+ * Forward references
+ */
+static void el_print_timestamp(union el_timestamp);
+static void el_process_subpackets(struct el_subpacket *, int);
+
+
+/*
+ * Generic
+ */
+void
+mchk_dump_mem(void *data, int length, char **annotation)
+{
+	unsigned long *ldata = data;
+	int i;
+	
+	for(i = 0; (i * sizeof(*ldata)) < length; i++) {
+		if (annotation && !annotation[i]) 
+			annotation = NULL;
+		printk("%s    %08x: %016lx    %s\n",
+		       err_print_prefix,
+		       (unsigned)(i * sizeof(*ldata)), ldata[i],
+		       annotation ? annotation[i] : "");
+	}
+}
+
+void
+mchk_dump_logout_frame(struct el_common *mchk_header)
+{
+	printk("%s  -- Frame Header --\n"
+	         "    Frame Size:   %d (0x%x) bytes\n"
+	         "    Flags:        %s%s\n"
+	         "    MCHK Code:    0x%x\n"
+	         "    Frame Rev:    %d\n"
+	         "    Proc Offset:  0x%08x\n"
+	         "    Sys Offset:   0x%08x\n"
+  	         "  -- Processor Region --\n",
+	       err_print_prefix, 
+	       mchk_header->size, mchk_header->size,
+	       mchk_header->retry ? "RETRY " : "", 
+  	         mchk_header->err2 ? "SECOND_ERR " : "",
+	       mchk_header->code,
+	       mchk_header->frame_rev,
+	       mchk_header->proc_offset,
+	       mchk_header->sys_offset);
+
+	mchk_dump_mem((void *)
+		      ((unsigned long)mchk_header + mchk_header->proc_offset),
+		      mchk_header->sys_offset - mchk_header->proc_offset,
+		      NULL);
+	
+	printk("%s  -- System Region --\n", err_print_prefix);
+	mchk_dump_mem((void *)
+		      ((unsigned long)mchk_header + mchk_header->sys_offset),
+		      mchk_header->size - mchk_header->sys_offset,
+		      NULL);
+	printk("%s  -- End of Frame --\n", err_print_prefix);
+}
+
+
+/*
+ * EV7 generic
+ */
+#if defined(CONFIG_ALPHA_GENERIC) || defined(CONFIG_ALPHA_EV7)
+
+void
+ev7_machine_check(u64 vector, u64 la_ptr, struct pt_regs *regs)
+{
+	/*
+	 * Sync the processor
+	 */
+	mb();
+	draina();
+
+	/*
+	 * Parse the logout frame without printing first. If the only error(s)
+	 * found are have a disposition of "dismiss", then just dismiss them
+	 * and don't print any message
+	 */
+	printk("%sEV7 MACHINE CHECK vector %lx\n", err_print_prefix, vector);
+
+	/* 
+	 * Release the logout frame 
+	 */
+	wrmces(0x7);
+	mb();
+}
+
+struct ev7_pal_subpacket {
+	union {
+		struct {
+			u32 mchk_code;
+			u32 subpacket_count;
+			u64 whami;
+			u64 rbox_whami;
+			u64 rbox_int;
+			u64 exc_addr;
+			union el_timestamp timestamp;
+			u64 halt_code;
+			u64 reserved;
+		} logout;
+	} by_type;
+};
+
+static char *el_ev7_processor_subpacket_annotation[] = {
+	"Subpacket Header",	"I_STAT",	"DC_STAT",
+	"C_ADDR",		"C_SYNDROME_1",	"C_SYNDROME_0",
+	"C_STAT",		"C_STS",	"MM_STAT",
+	"EXC_ADDR",		"IER_CM",	"ISUM",
+	"PAL_BASE",		"I_CTL",	"PROCESS_CONTEXT",
+	"CBOX_CTL",		"CBOX_STP_CTL",	"CBOX_ACC_CTL",
+	"CBOX_LCL_SET",		"CBOX_GLB_SET",	"BBOX_CTL",
+	"BBOX_ERR_STS",		"BBOX_ERR_IDX",	"CBOX_DDP_ERR_STS",
+	"BBOX_DAT_RMP",		NULL
+};
+
+static char *el_ev7_zbox_subpacket_annotation[] = {
+	"Subpacket Header", 	
+	"ZBOX(0): DRAM_ERR_STATUS_2 / DRAM_ERR_STATUS_1",
+	"ZBOX(0): DRAM_ERROR_CTL    / DRAM_ERR_STATUS_3",
+	"ZBOX(0): DIFT_TIMEOUT      / DRAM_ERR_ADR",
+	"ZBOX(0): FRC_ERR_ADR       / DRAM_MAPPER_CTL",
+	"ZBOX(0): reserved          / DIFT_ERR_STATUS",
+	"ZBOX(1): DRAM_ERR_STATUS_2 / DRAM_ERR_STATUS_1",
+	"ZBOX(1): DRAM_ERROR_CTL    / DRAM_ERR_STATUS_3",
+	"ZBOX(1): DIFT_TIMEOUT      / DRAM_ERR_ADR",
+	"ZBOX(1): FRC_ERR_ADR       / DRAM_MAPPER_CTL",
+	"ZBOX(1): reserved          / DIFT_ERR_STATUS",
+	"CBOX_CTL",		"CBOX_STP_CTL",
+	"ZBOX(0)_ERROR_PA",	"ZBOX(1)_ERROR_PA",
+	"ZBOX(0)_ORED_SYNDROME","ZBOX(1)_ORED_SYNDROME",
+	NULL
+};
+
+static char *el_ev7_rbox_subpacket_annotation[] = {
+	"Subpacket Header",	"RBOX_CFG",	"RBOX_N_CFG",
+	"RBOX_S_CFG",		"RBOX_E_CFG",	"RBOX_W_CFG",
+	"RBOX_N_ERR",		"RBOX_S_ERR",	"RBOX_E_ERR",
+	"RBOX_W_ERR",		"RBOX_IO_CFG",	"RBOX_IO_ERR",
+	"RBOX_L_ERR",		"RBOX_WHOAMI",	"RBOX_IMASL",
+	"RBOX_INTQ",		"RBOX_INT",	NULL
+};
+
+static char *el_ev7_io_subpacket_annotation[] = {
+	"Subpacket Header",	"IO_ASIC_REV",	"IO_SYS_REV",
+	"IO7_UPH",		"HPI_CTL",	"CRD_CTL",
+	"HEI_CTL",		"PO7_ERROR_SUM","PO7_UNCRR_SYM",
+	"PO7_CRRCT_SYM",	"PO7_UGBGE_SYM","PO7_ERR_PKT0",
+	"PO7_ERR_PKT1",		"reserved",	"reserved",
+	"PO0_ERR_SUM",		"PO0_TLB_ERR",	"PO0_SPL_COMPLT",
+	"PO0_TRANS_SUM",	"PO0_FIRST_ERR","PO0_MULT_ERR",
+	"DM CSR PH",		"DM CSR PH",	"DM CSR PH",
+	"DM CSR PH",		"reserved",
+	"PO1_ERR_SUM",		"PO1_TLB_ERR",	"PO1_SPL_COMPLT",
+	"PO1_TRANS_SUM",	"PO1_FIRST_ERR","PO1_MULT_ERR",
+	"DM CSR PH",		"DM CSR PH",	"DM CSR PH",
+	"DM CSR PH",		"reserved",
+	"PO2_ERR_SUM",		"PO2_TLB_ERR",	"PO2_SPL_COMPLT",
+	"PO2_TRANS_SUM",	"PO2_FIRST_ERR","PO2_MULT_ERR",
+	"DM CSR PH",		"DM CSR PH",	"DM CSR PH",
+	"DM CSR PH",		"reserved",
+	"PO3_ERR_SUM",		"PO3_TLB_ERR",	"PO3_SPL_COMPLT",
+	"PO3_TRANS_SUM",	"PO3_FIRST_ERR","PO3_MULT_ERR",
+	"DM CSR PH",		"DM CSR PH",	"DM CSR PH",
+	"DM CSR PH",		"reserved",	
+	NULL
+};
+	
+static struct el_subpacket_annotation el_ev7_pal_annotations[] = {
+	SUBPACKET_ANNOTATION(EL_CLASS__PAL,
+			     EL_TYPE__PAL__EV7_PROCESSOR,
+			     1,
+			     "EV7 Processor Subpacket",
+			     el_ev7_processor_subpacket_annotation),
+	SUBPACKET_ANNOTATION(EL_CLASS__PAL,
+			     EL_TYPE__PAL__EV7_ZBOX,
+			     1,
+			     "EV7 ZBOX Subpacket",
+			     el_ev7_zbox_subpacket_annotation),
+	SUBPACKET_ANNOTATION(EL_CLASS__PAL,
+			     EL_TYPE__PAL__EV7_RBOX,
+			     1,
+			     "EV7 RBOX Subpacket",
+			     el_ev7_rbox_subpacket_annotation),
+	SUBPACKET_ANNOTATION(EL_CLASS__PAL,
+			     EL_TYPE__PAL__EV7_IO,
+			     1,
+			     "EV7 IO Subpacket",
+			     el_ev7_io_subpacket_annotation)
+};
+
+static struct el_subpacket *
+ev7_process_pal_subpacket(struct el_subpacket *header)
+{
+	struct ev7_pal_subpacket *packet;
+
+	if (header->class != EL_CLASS__PAL) {
+		printk("%s  ** Unexpected header CLASS %d TYPE %d, aborting\n",
+		       err_print_prefix,
+		       header->class, header->type);
+		return NULL;
+	}
+
+	packet = (struct ev7_pal_subpacket *)header->by_type.raw.data_start;
+
+	switch(header->type) {
+	case EL_TYPE__PAL__LOGOUT_FRAME:
+		printk("%s*** MCHK occurred on LPID %ld (RBOX %lx)\n",
+		       err_print_prefix,
+		       packet->by_type.logout.whami, 
+		       packet->by_type.logout.rbox_whami);
+		el_print_timestamp(packet->by_type.logout.timestamp);
+		printk("%s  EXC_ADDR: %016lx\n"
+		         "  HALT_CODE: %lx\n", 
+		       err_print_prefix,
+		       packet->by_type.logout.exc_addr,
+		       packet->by_type.logout.halt_code);
+		el_process_subpackets(header,
+                                      packet->by_type.logout.subpacket_count);
+		break;
+	default:
+		printk("%s  ** PAL TYPE %d SUBPACKET\n", 
+		       err_print_prefix,
+		       header->type);
+		el_annotate_subpacket(header);
+		break;
+	}
+	
+	return (struct el_subpacket *)((unsigned long)header + header->length);
+}
+
+struct el_subpacket_handler ev7_pal_subpacket_handler =
+	SUBPACKET_HANDLER_INIT(EL_CLASS__PAL, ev7_process_pal_subpacket);
+
+void 
+ev7_register_error_handlers(void)
+{
+	int i;
+
+	for(i = 0;
+	    i<sizeof(el_ev7_pal_annotations)/sizeof(el_ev7_pal_annotations[1]);
+	    i++) {
+		cdl_register_subpacket_annotation(&el_ev7_pal_annotations[i]);
+	}	
+	cdl_register_subpacket_handler(&ev7_pal_subpacket_handler);
+}
+
+#endif /* CONFIG_ALPHA_GENERIC || CONFIG_ALPHA_EV7 */
+
+
+/*
+ * EV6 generic
+ */
+
+#if defined(CONFIG_ALPHA_GENERIC) || defined(CONFIG_ALPHA_EV6)
+
+static int
+ev6_parse_ibox(u64 i_stat, int print)
+{
+	int status = MCHK_DISPOSITION_REPORT;
+
+#define EV6__I_STAT__PAR	(1UL << 29)
+#define EV6__I_STAT__ERRMASK	(EV6__I_STAT__PAR)
+
+	if (!(i_stat & EV6__I_STAT__ERRMASK))
+		return MCHK_DISPOSITION_UNKNOWN_ERROR;
+
+	if (!print)
+		return status;
+
+	if (i_stat & EV6__I_STAT__PAR)
+		printk("%s    Icache parity error\n", err_print_prefix);
+
+	return status;
+}
+
+static int
+ev6_parse_mbox(u64 mm_stat, u64 d_stat, u64 c_stat, int print)
+{
+	int status = MCHK_DISPOSITION_REPORT;
+
+#define EV6__MM_STAT__DC_TAG_PERR	(1UL << 10)
+#define EV6__MM_STAT__ERRMASK		(EV6__MM_STAT__DC_TAG_PERR)
+#define EV6__D_STAT__TPERR_P0		(1UL << 0)
+#define EV6__D_STAT__TPERR_P1		(1UL << 1)
+#define EV6__D_STAT__ECC_ERR_ST		(1UL << 2)
+#define EV6__D_STAT__ECC_ERR_LD		(1UL << 3)
+#define EV6__D_STAT__SEO		(1UL << 4)
+#define EV6__D_STAT__ERRMASK		(EV6__D_STAT__TPERR_P0 |	\
+                                         EV6__D_STAT__TPERR_P1 | 	\
+                                         EV6__D_STAT__ECC_ERR_ST | 	\
+                                         EV6__D_STAT__ECC_ERR_LD | 	\
+                                         EV6__D_STAT__SEO)
+
+	if (!(d_stat & EV6__D_STAT__ERRMASK) && 
+	    !(mm_stat & EV6__MM_STAT__ERRMASK))
+		return MCHK_DISPOSITION_UNKNOWN_ERROR;
+
+	if (!print)
+		return status;
+
+	if (mm_stat & EV6__MM_STAT__DC_TAG_PERR)
+		printk("%s    Dcache tag parity error on probe\n",
+		       err_print_prefix);
+	if (d_stat & EV6__D_STAT__TPERR_P0)
+		printk("%s    Dcache tag parity error - pipe 0\n",
+		       err_print_prefix);
+	if (d_stat & EV6__D_STAT__TPERR_P1)
+		printk("%s    Dcache tag parity error - pipe 1\n",
+		       err_print_prefix);
+	if (d_stat & EV6__D_STAT__ECC_ERR_ST)
+		printk("%s    ECC error occurred on a store\n", 
+		       err_print_prefix);
+	if (d_stat & EV6__D_STAT__ECC_ERR_LD)
+		printk("%s    ECC error occurred on a %s load\n",
+		       err_print_prefix,
+		       c_stat ? "" : "speculative ");
+	if (d_stat & EV6__D_STAT__SEO)
+		printk("%s    Dcache second error\n", err_print_prefix);
+
+	return status;
+}
+
+static int
+ev6_parse_cbox(u64 c_addr, u64 c1_syn, u64 c2_syn, 
+	       u64 c_stat, u64 c_sts, int print)
+{
+	char *sourcename[] = { "UNKNOWN", "UNKNOWN", "UNKNOWN",
+			       "MEMORY", "BCACHE", "DCACHE", 
+			       "BCACHE PROBE", "BCACHE PROBE" };
+	char *streamname[] = { "D", "I" };
+	char *bitsname[] = { "SINGLE", "DOUBLE" };
+	int status = MCHK_DISPOSITION_REPORT;
+	int source = -1, stream = -1, bits = -1;
+
+#define EV6__C_STAT__BC_PERR		(0x01)
+#define EV6__C_STAT__DC_PERR		(0x02)
+#define EV6__C_STAT__DSTREAM_MEM_ERR	(0x03)
+#define EV6__C_STAT__DSTREAM_BC_ERR	(0x04)
+#define EV6__C_STAT__DSTREAM_DC_ERR	(0x05)
+#define EV6__C_STAT__PROBE_BC_ERR0	(0x06)	/* both 6 and 7 indicate... */
+#define EV6__C_STAT__PROBE_BC_ERR1	(0x07)	/* ...probe bc error.       */
+#define EV6__C_STAT__ISTREAM_MEM_ERR	(0x0B)
+#define EV6__C_STAT__ISTREAM_BC_ERR	(0x0C)
+#define EV6__C_STAT__DSTREAM_MEM_DBL	(0x13)
+#define EV6__C_STAT__DSTREAM_BC_DBL	(0x14)
+#define EV6__C_STAT__ISTREAM_MEM_DBL	(0x1B)
+#define EV6__C_STAT__ISTREAM_BC_DBL	(0x1C)
+#define EV6__C_STAT__SOURCE_MEMORY	(0x03)
+#define EV6__C_STAT__SOURCE_BCACHE	(0x04)
+#define EV6__C_STAT__SOURCE__S		(0)
+#define EV6__C_STAT__SOURCE__M 		(0x07)
+#define EV6__C_STAT__ISTREAM__S		(3)
+#define EV6__C_STAT__ISTREAM__M		(0x01)
+#define EV6__C_STAT__DOUBLE__S		(4)
+#define EV6__C_STAT__DOUBLE__M		(0x01)
+#define EV6__C_STAT__ERRMASK		(0x1F)
+#define EV6__C_STS__SHARED		(1 << 0)
+#define EV6__C_STS__DIRTY		(1 << 1)
+#define EV6__C_STS__VALID		(1 << 2)
+#define EV6__C_STS__PARITY		(1 << 3)
+
+	if (!(c_stat & EV6__C_STAT__ERRMASK))
+		return MCHK_DISPOSITION_UNKNOWN_ERROR;
+
+	if (!print)
+		return status;
+
+	source = EXTRACT(c_stat, EV6__C_STAT__SOURCE);
+	stream = EXTRACT(c_stat, EV6__C_STAT__ISTREAM);
+	bits = EXTRACT(c_stat, EV6__C_STAT__DOUBLE);
+
+	if (c_stat & EV6__C_STAT__BC_PERR) {
+		printk("%s    Bcache tag parity error\n", err_print_prefix);
+		source = -1;
+	}
+
+	if (c_stat & EV6__C_STAT__DC_PERR) {
+		printk("%s    Dcache tag parity error\n", err_print_prefix);
+		source = -1;
+	}
+
+	if (c_stat == EV6__C_STAT__PROBE_BC_ERR0 ||
+	    c_stat == EV6__C_STAT__PROBE_BC_ERR1) {
+		printk("%s    Bcache single-bit error on a probe hit\n",
+		       err_print_prefix);
+		source = -1;
+	}
+
+	if (source != -1) 
+		printk("%s    %s-STREAM %s-BIT ECC error from %s\n",
+		       err_print_prefix,
+		       streamname[stream], bitsname[bits], sourcename[source]);
+
+	printk("%s    Address: 0x%016lx\n"
+	         "    Syndrome[upper.lower]: %02lx.%02lx\n", 
+	       err_print_prefix,
+	       c_addr,
+	       c2_syn, c1_syn);
+
+	if (source == EV6__C_STAT__SOURCE_MEMORY ||
+	    source == EV6__C_STAT__SOURCE_BCACHE) 
+		printk("%s    Block status: %s%s%s%s\n",
+		       err_print_prefix,
+		       (c_sts & EV6__C_STS__SHARED) ? "SHARED " : "",
+		       (c_sts & EV6__C_STS__DIRTY)  ? "DIRTY "  : "",
+		       (c_sts & EV6__C_STS__VALID)  ? "VALID "  : "",
+		       (c_sts & EV6__C_STS__PARITY) ? "PARITY " : "");
+		
+	return status;
+}
+
+int
+ev6_process_logout_frame(struct el_common *mchk_header, int print)
+{
+	struct el_common_EV6_mcheck *ev6mchk = 
+		(struct el_common_EV6_mcheck *)mchk_header;
+	int status = MCHK_DISPOSITION_UNKNOWN_ERROR;
+
+	status |= ev6_parse_ibox(ev6mchk->I_STAT, print);
+	status |= ev6_parse_mbox(ev6mchk->MM_STAT, ev6mchk->DC_STAT, 
+				 ev6mchk->C_STAT, print);
+	status |= ev6_parse_cbox(ev6mchk->C_ADDR, ev6mchk->DC1_SYNDROME,
+				 ev6mchk->DC0_SYNDROME, ev6mchk->C_STAT,
+				 ev6mchk->C_STS, print);
+
+	if (!print)
+		return status;
+
+	if (status != MCHK_DISPOSITION_DISMISS) {
+		char *saved_err_prefix = err_print_prefix;
+
+		/*
+		 * Dump some additional information from the frame
+		 */
+		printk("%s    EXC_ADDR: 0x%016lx   IER_CM: 0x%016lx"
+		            "   ISUM: 0x%016lx\n"
+		         "    PAL_BASE: 0x%016lx   I_CTL:  0x%016lx"
+		            "   PCTX: 0x%016lx\n",
+		       err_print_prefix,
+		       ev6mchk->EXC_ADDR, ev6mchk->IER_CM, ev6mchk->ISUM,
+		       ev6mchk->PAL_BASE, ev6mchk->I_CTL, ev6mchk->PCTX);
+
+		if (status == MCHK_DISPOSITION_UNKNOWN_ERROR) {
+			printk("%s    UNKNOWN error, frame follows:\n",
+			       err_print_prefix);
+		} else {
+			/* had decode -- downgrade print level for frame */
+			err_print_prefix = KERN_NOTICE;
+		}
+
+		mchk_dump_logout_frame(mchk_header);
+
+		err_print_prefix = saved_err_prefix;
+	}
+
+	return status;
+}
+
+void
+ev6_machine_check(u64 vector, u64 la_ptr, struct pt_regs *regs)
+{
+	struct el_common *mchk_header = (struct el_common *)la_ptr;
+
+	/*
+	 * Sync the processor
+	 */
+	mb();
+	draina();
+
+	/*
+	 * Parse the logout frame without printing first. If the only error(s)
+	 * found are have a disposition of "dismiss", then just dismiss them
+	 * and don't print any message
+	 */
+	if (ev6_process_logout_frame(mchk_header, 0) != 
+	    MCHK_DISPOSITION_DISMISS) {
+		char *saved_err_prefix = err_print_prefix;
+		err_print_prefix = KERN_CRIT;
+
+		/*
+		 * Either a nondismissable error was detected or no
+		 * recognized error was detected  in the logout frame 
+		 * -- report the error in either case
+		 */
+		printk("%s*CPU %s Error (Vector 0x%x) reported on CPU %d:\n", 
+		       err_print_prefix,
+		       (vector == SCB_Q_PROCERR)?"Correctable":"Uncorrectable",
+		       (unsigned int)vector, (int)smp_processor_id());
+		
+		ev6_process_logout_frame(mchk_header, 1);
+		dik_show_regs(regs, NULL);
+
+		err_print_prefix = saved_err_prefix;
+	}
+
+	/* 
+	 * Release the logout frame 
+	 */
+	wrmces(0x7);
+	mb();
+}
+
+#endif /* CONFIG_ALPHA_GENERIC || CONFIG_ALPHA_EV6 */
+
+
+/*
+ * Console Data Log
+ */
+/* Data */
+static struct el_subpacket_handler *subpacket_handler_list = NULL;
+static struct el_subpacket_annotation *subpacket_annotation_list = NULL;
+
+static void
+el_print_timestamp(union el_timestamp timestamp)
+{
+	if (timestamp.as_int)
+		printk("%s  TIMESTAMP: %d/%d/%02d %d:%02d:%0d\n", 
+		       err_print_prefix,
+		       timestamp.b.month, timestamp.b.day,
+		       timestamp.b.year, timestamp.b.hour,
+		       timestamp.b.minute, timestamp.b.second);
+}
+
+static struct el_subpacket *
+el_process_header_subpacket(struct el_subpacket *header)
+{
+	union el_timestamp timestamp;
+	char *name = "UNKNOWN EVENT";
+	int packet_count = 0;
+	int length = 0;
+
+	if (header->class != EL_CLASS__HEADER) {
+		printk("%s** Unexpected header CLASS %d TYPE %d, aborting\n",
+		       err_print_prefix,
+		       header->class, header->type);
+		return NULL;
+	}
+
+	switch(header->type) {
+	case EL_TYPE__HEADER__SYSTEM_ERROR_FRAME:
+		name = "SYSTEM ERROR";
+		length = header->by_type.sys_err.frame_length;
+		packet_count = 
+			header->by_type.sys_err.frame_packet_count;
+		timestamp.as_int = 0;
+		break;
+	case EL_TYPE__HEADER__SYSTEM_EVENT_FRAME:
+		name = "SYSTEM EVENT";
+		length = header->by_type.sys_event.frame_length;
+		packet_count = 
+			header->by_type.sys_event.frame_packet_count;
+		timestamp = header->by_type.sys_event.timestamp;
+		break;
+	case EL_TYPE__HEADER__HALT_FRAME:
+		name = "ERROR HALT";
+		length = header->by_type.err_halt.frame_length;
+		packet_count = 
+			header->by_type.err_halt.frame_packet_count;
+		timestamp = header->by_type.err_halt.timestamp;
+		break;
+	case EL_TYPE__HEADER__LOGOUT_FRAME:
+		name = "LOGOUT FRAME";
+		length = header->by_type.logout_header.frame_length;
+		packet_count = 1;
+		timestamp.as_int = 0;
+		break;
+	default: /* Unknown */
+		printk("%s** Unknown header - CLASS %d TYPE %d, aborting\n",
+		       err_print_prefix,
+		       header->class, header->type);
+		return NULL;		
+	}
+
+	printk("%s*** %s:\n"
+	         "  CLASS %d, TYPE %d\n", 
+	       err_print_prefix,
+	       name,
+	       header->class, header->type);
+	el_print_timestamp(timestamp);
+	
+	/*
+	 * Process the subpackets
+	 */
+	el_process_subpackets(header, packet_count);
+
+	/* return the next header */
+	header = (struct el_subpacket *)
+		((unsigned long)header + header->length + length);
+	return header;
+}
+
+static void
+el_process_subpackets(struct el_subpacket *header, int packet_count)
+{
+	struct el_subpacket *subpacket;
+	int i;
+
+	subpacket = (struct el_subpacket *)
+		((unsigned long)header + header->length);
+
+	for(i = 0; subpacket && i < packet_count; i++) {
+		printk("%sPROCESSING SUBPACKET %d\n", err_print_prefix, i);
+		subpacket = el_process_subpacket(subpacket);
+	}
+}
+
+static struct el_subpacket *
+el_process_subpacket_reg(struct el_subpacket *header)
+{
+	struct el_subpacket *next = NULL;
+	struct el_subpacket_handler *h = subpacket_handler_list;
+
+	for(; h && h->class != header->class; h = h->next);
+	if (h) next = h->handler(header);
+
+	return next;
+}
+
+struct el_subpacket *
+el_process_subpacket(struct el_subpacket *header)
+{
+	struct el_subpacket *next = NULL;
+
+	switch(header->class) {
+	case EL_CLASS__TERMINATION:
+		/* Termination packet, there are no more */
+		break;
+	case EL_CLASS__HEADER: 
+		next = el_process_header_subpacket(header);
+		break;
+	default:
+		if (NULL == (next = el_process_subpacket_reg(header))) {
+			printk("%s** Unexpected header CLASS %d TYPE %d"
+			       " -- aborting.\n",
+			       err_print_prefix,
+			       header->class, header->type);
+		}
+		break;
+	}
+
+	return next;
+}
+
+void 
+el_annotate_subpacket(struct el_subpacket *header)
+{
+	struct el_subpacket_annotation *a;
+	char **annotation = NULL;
+
+	for(a = subpacket_annotation_list; a; a = a->next) {
+		if (a->class == header->class &&
+		    a->type == header->type &&
+		    a->revision == header->revision) {
+			/*
+			 * We found the annotation
+			 */
+			annotation = a->annotation;
+			printk("%s  %s\n", err_print_prefix, a->description);
+			break;
+		}
+	}
+
+	mchk_dump_mem(header, header->length, annotation);
+}
+
+static void __init
+cdl_process_console_data_log(int cpu, struct percpu_struct *pcpu)
+{
+	struct el_subpacket *header = (struct el_subpacket *)
+		(IDENT_ADDR | pcpu->console_data_log_pa);
+	int err;
+
+	printk("%s******* CONSOLE DATA LOG FOR CPU %d. *******\n"
+	         "*** Error(s) were logged on a previous boot\n",
+	       err_print_prefix, cpu);
+	
+	for(err = 0; header && (header->class != EL_CLASS__TERMINATION); err++)
+		header = el_process_subpacket(header);
+
+	/* let the console know it's ok to clear the error(s) at restart */
+	pcpu->console_data_log_pa = 0;
+
+	printk("%s*** %d total error(s) logged\n"
+	         "**** END OF CONSOLE DATA LOG FOR CPU %d ****\n", 
+	       err_print_prefix, err, cpu);
+}
+
+void __init
+cdl_check_console_data_log(void)
+{
+	struct percpu_struct *pcpu;
+	int cpu;
+
+	for(cpu = 0; cpu < hwrpb->nr_processors; cpu++) {
+		pcpu = (struct percpu_struct *)
+			((unsigned long)hwrpb + hwrpb->processor_offset 
+			 + cpu * hwrpb->processor_size);
+		if (pcpu->console_data_log_pa)
+			cdl_process_console_data_log(cpu, pcpu);
+	}
+
+}
+
+int __init
+cdl_register_subpacket_annotation(struct el_subpacket_annotation *new)
+{
+	struct el_subpacket_annotation *a = subpacket_annotation_list;
+
+	if (a == NULL) subpacket_annotation_list = new;
+	else {
+		for(; a->next != NULL; a = a->next) {
+			if ((a->class == new->class && a->type == new->type) ||
+			    a == new) {
+				printk("Attempted to re-register "
+				       "subpacket annotation\n");
+				return -EINVAL;
+			}
+		}
+		a->next = new;
+	}
+	new->next = NULL;
+
+	return 0;
+}
+
+int __init
+cdl_register_subpacket_handler(struct el_subpacket_handler *new)
+{
+	struct el_subpacket_handler *h = subpacket_handler_list;
+
+	if (h == NULL) subpacket_handler_list = new;
+	else {
+		for(; h->next != NULL; h = h->next) {
+			if (h->class == new->class || h == new) {
+				printk("Attempted to re-register "
+				       "subpacket handler\n");
+				return -EINVAL;
+			}
+		}
+		h->next = new;
+	}
+	new->next = NULL;
+
+	return 0;
+}
+

arch/alpha/kernel/err_impl.h

+/*
+ *	linux/arch/alpha/kernel/err_impl.h
+ *
+ *	Copyright (C) 2000 Jeff Wiedemeier (Compaq Computer Corporation)
+ *
+ *	Contains declarations and macros to support Alpha error handling
+ * 	implementations.
+ */
+
+/*
+ * SCB Vector definitions
+ */
+#define SCB_Q_SYSERR	0x620
+#define SCB_Q_PROCERR	0x630
+#define SCB_Q_SYSMCHK	0x660
+#define SCB_Q_PROCMCHK	0x670
+#define SCB_Q_SYSEVENT	0x680
+
+/*
+ * Disposition definitions for logout frame parser
+ */
+#define MCHK_DISPOSITION_UNKNOWN_ERROR		0x00
+#define MCHK_DISPOSITION_REPORT			0x01
+#define MCHK_DISPOSITION_DISMISS		0x02
+
+/*
+ * Error Log definitions
+ */
+/*
+ * Types
+ */
+
+#define EL_CLASS__TERMINATION		(0)
+#  define EL_TYPE__TERMINATION__TERMINATION		(0)
+#define EL_CLASS__HEADER		(5)
+#  define EL_TYPE__HEADER__SYSTEM_ERROR_FRAME		(1)
+#  define EL_TYPE__HEADER__SYSTEM_EVENT_FRAME		(2)
+#  define EL_TYPE__HEADER__HALT_FRAME			(3)
+#  define EL_TYPE__HEADER__LOGOUT_FRAME			(19)
+#define EL_CLASS__GENERAL_NOTIFICATION	(9)
+#define EL_CLASS__PCI_ERROR_FRAME	(11)
+#define EL_CLASS__REGATTA_FAMILY	(12)
+#  define EL_TYPE__REGATTA__PROCESSOR_ERROR_FRAME	(1)
+#  define EL_TYPE__REGATTA__SYSTEM_ERROR_FRAME		(2)
+#  define EL_TYPE__REGATTA__ENVIRONMENTAL_FRAME		(3)
+#  define EL_TYPE__REGATTA__TITAN_PCHIP0_EXTENDED	(8)
+#  define EL_TYPE__REGATTA__TITAN_PCHIP1_EXTENDED	(9)
+#  define EL_TYPE__REGATTA__TITAN_MEMORY_EXTENDED	(10)
+#  define EL_TYPE__REGATTA__PROCESSOR_DBL_ERROR_HALT	(11)
+#  define EL_TYPE__REGATTA__SYSTEM_DBL_ERROR_HALT	(12)
+#define EL_CLASS__PAL                   (14)
+#  define EL_TYPE__PAL__LOGOUT_FRAME                    (1)
+#  define EL_TYPE__PAL__EV7_PROCESSOR			(4)
+#  define EL_TYPE__PAL__EV7_ZBOX			(5)
+#  define EL_TYPE__PAL__EV7_RBOX			(6)
+#  define EL_TYPE__PAL__EV7_IO				(7)
+
+union el_timestamp {
+	struct {
+		u8 second;
+		u8 minute;
+		u8 hour;
+		u8 day;
+		u8 month;
+		u8 year;
+	} b;
+	u64 as_int;
+};
+
+struct el_subpacket {
+	u16 length;		/* length of header (in bytes)	*/
+	u16 class;		/* header class and type...   	*/
+	u16 type;		/* ...determine content     	*/
+	u16 revision;		/* header revision 		*/
+	union {
+		struct {	/* Class 5, Type 1 - System Error	*/
+			u32 frame_length;
+			u32 frame_packet_count;			
+		} sys_err;			
+		struct {	/* Class 5, Type 2 - System Event 	*/
+			union el_timestamp timestamp;
+			u32 frame_length;
+			u32 frame_packet_count;			
+		} sys_event;
+		struct {	/* Class 5, Type 3 - Double Error Halt	*/
+			u16 halt_code;
+			u16 reserved;
+			union el_timestamp timestamp;
+			u32 frame_length;
+			u32 frame_packet_count;
+		} err_halt;
+		struct {	/* Clasee 5, Type 19 - Logout Frame Header */
+			u32 frame_length;
+			u32 frame_flags;
+			u32 cpu_offset;	
+			u32 system_offset;
+		} logout_header;
+		struct {	/* Class 12 - Regatta			*/
+			u64 cpuid;
+			u64 data_start[1];
+		} regatta_frame;
+		struct {	/* Raw 				        */
+			u64 data_start[1];
+		} raw;
+	} by_type;
+};
+
+struct el_subpacket_annotation {
+	struct el_subpacket_annotation *next;
+	u16 class;
+	u16 type;
+	u16 revision;
+	char *description;
+	char **annotation;
+};
+#define SUBPACKET_ANNOTATION(c, t, r, d, a) {NULL, (c), (t), (r), (d), (a)}
+
+struct el_subpacket_handler {
+	struct el_subpacket_handler *next;
+	u16 class;
+	struct el_subpacket *(*handler)(struct el_subpacket *);
+};
+#define SUBPACKET_HANDLER_INIT(c, h) {NULL, (c), (h)}
+
+/*
+ * Extract a field from a register given it's name. defines
+ * for the LSB (__S - shift count) and bitmask (__M) are required
+ */
+#define EXTRACT(u, f) (((u) >> f##__S) & f##__M)
+
+/*
+ * err_common.c
+ */
+extern char *err_print_prefix;
+
+extern void mchk_dump_mem(void *, int, char **);
+extern void mchk_dump_logout_frame(struct el_common *);
+extern void ev7_register_error_handlers(void);
+extern void ev7_machine_check(u64, u64, struct pt_regs *);
+extern void ev6_register_error_handlers(void);
+extern int ev6_process_logout_frame(struct el_common *, int);
+extern void ev6_machine_check(u64, u64, struct pt_regs *);
+extern struct el_subpacket *el_process_subpacket(struct el_subpacket *);
+extern void el_annotate_subpacket(struct el_subpacket *);
+extern void cdl_check_console_data_log(void);
+extern int cdl_register_subpacket_annotation(struct el_subpacket_annotation *);
+extern int cdl_register_subpacket_handler(struct el_subpacket_handler *);
+

arch/alpha/kernel/irq_alpha.c

@@ -147,10 +147,10 @@ process_mcheck_info(unsigned long vector, unsigned long la_ptr,
 
 	mchk_header = (struct el_common *)la_ptr;
 
-	printk(KERN_CRIT "%s machine check: vector=0x%lx pc=0x%lx code=0x%lx\n",
+	printk(KERN_CRIT "%s machine check: vector=0x%lx pc=0x%lx code=0x%x\n",
 	       machine, vector, regs->pc, mchk_header->code);
 
-	switch ((unsigned int) mchk_header->code) {
+	switch (mchk_header->code) {
 	/* Machine check reasons.  Defined according to PALcode sources.  */
 	case 0x80: reason = "tag parity error"; break;
 	case 0x82: reason = "tag control parity error"; break;

arch/alpha/kernel/sys_nautilus.c

@@ -46,6 +46,7 @@
 #include <asm/tlbflush.h>
 
 #include "proto.h"
+#include "err_impl.h"
 #include "irq_impl.h"
 #include "pci_impl.h"
 #include "machvec_impl.h"
@@ -100,317 +101,6 @@ nautilus_kill_arch(int mode)
 	}
 }
 
-/* Machine check handler code
- *
- * Perform analysis of a machine check that was triggered by the EV6
- * CPU's fault-detection mechanism.
- */
-
-/* IPR structures for EV6, containing the necessary data for the
- * machine check handler to unpick the logout frame
- */
-
-/* I_STAT */
-
-#define EV6__I_STAT__PAR                ( 1 << 29 )
-
-/* MM_STAT */
-
-#define EV6__MM_STAT__DC_TAG_PERR       ( 1 << 10 )
-
-/* DC_STAT */
-
-#define EV6__DC_STAT__SEO               ( 1 << 4 )
-#define EV6__DC_STAT__ECC_ERR_LD        ( 1 << 3 )
-#define EV6__DC_STAT__ECC_ERR_ST        ( 1 << 2 )
-#define EV6__DC_STAT__TPERR_P1          ( 1 << 1 )
-#define EV6__DC_STAT__TPERR_P0          ( 1      )
-
-/* C_STAT */
-
-#define EV6__C_STAT__BC_PERR            ( 0x01 )
-#define EV6__C_STAT__DC_PERR            ( 0x02 )
-#define EV6__C_STAT__DSTREAM_MEM_ERR    ( 0x03 )
-#define EV6__C_STAT__DSTREAM_BC_ERR     ( 0x04 )
-#define EV6__C_STAT__DSTREAM_DC_ERR     ( 0x05 )
-#define EV6__C_STAT__PROBE_BC_ERR0      ( 0x06 )
-#define EV6__C_STAT__PROBE_BC_ERR1      ( 0x07 )
-#define EV6__C_STAT__ISTREAM_MEM_ERR    ( 0x0B )
-#define EV6__C_STAT__ISTREAM_BC_ERR     ( 0x0C )
-#define EV6__C_STAT__DSTREAM_MEM_DBL    ( 0x13 )
-#define EV6__C_STAT__DSTREAM_BC_DBL     ( 0x14 )
-#define EV6__C_STAT__ISTREAM_MEM_DBL    ( 0x1B )
-#define EV6__C_STAT__ISTREAM_BC_DBL     ( 0x1C )
-
-
-/* Take the two syndromes from the CBOX error chain and convert them
- * into a bit number.  */
-
-/* NOTE - since I don't know of any difference between C0 and C1 I
-   just ignore C1, since in all cases I've seen so far they are
-   identical.  */
-
-static const unsigned char ev6_bit_to_syndrome[72] =
-{
-	0xce, 0xcb, 0xd3, 0xd5, 0xd6, 0xd9, 0xda, 0xdc,     /* 0 */
-	0x23, 0x25, 0x26, 0x29, 0x2a, 0x2c, 0x31, 0x34,     /* 8 */
-	0x0e, 0x0b, 0x13, 0x15, 0x16, 0x19, 0x1a, 0x1c,     /* 16 */
-	0xe3, 0xe5, 0xe6, 0xe9, 0xea, 0xec, 0xf1, 0xf4,     /* 24 */
-	0x4f, 0x4a, 0x52, 0x54, 0x57, 0x58, 0x5b, 0x5d,     /* 32 */
-	0xa2, 0xa4, 0xa7, 0xa8, 0xab, 0xad, 0xb0, 0xb5,     /* 40 */
-	0x8f, 0x8a, 0x92, 0x94, 0x97, 0x98, 0x9b, 0x9d,     /* 48 */
-	0x62, 0x64, 0x67, 0x68, 0x6b, 0x6d, 0x70, 0x75,     /* 56 */
-	0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80      /* 64 */
-};
-
-
-static int ev6_syn2bit(unsigned long c0, unsigned long c1)
-{
-	int bit;
-
-	for (bit = 0; bit < 72; bit++)
-		if (ev6_bit_to_syndrome[bit] == c0)	return bit;
-	for (bit = 0; bit < 72; bit++)
-		if (ev6_bit_to_syndrome[bit] == c1)	return bit + 64;
-
-	return -1;                  /* not found */
-}
-
-
-/* Single bit ECC errors are categorized here.  */
-
-#if 0
-static const char *interr = "CPU internal error";
-static const char *slotb= "Slot-B error";
-static const char *membus= "Memory/EV6-bus error";
-#else
-static const char *interr = "";
-static const char *slotb = "";
-static const char *membus = "";
-#endif
-
-static void
-ev6_crd_interp(char *interp, struct el_common_EV6_mcheck * L)
-{
-	/* Icache data or tag parity error.  */
-	if (L->I_STAT & EV6__I_STAT__PAR) {
-		sprintf(interp, "%s: I_STAT[PAR]\n "
-			"Icache data or tag parity error", interr);
-		return;
-	}
-
-	/* Dcache tag parity error (on issue) (DFAULT).  */
-	if (L->MM_STAT & EV6__MM_STAT__DC_TAG_PERR) {
-		sprintf(interp, "%s: MM_STAT[DC_TAG_PERR]\n "
-			"Dcache tag parity error(on issue)", interr);
-		return;
-	}
-
-	/* Errors relating to D-stream set non-zero DC_STAT.
-	   Mask CRD bits.  */
-	switch (L->DC_STAT & (EV6__DC_STAT__ECC_ERR_ST
-			      | EV6__DC_STAT__ECC_ERR_LD)) {
-	case EV6__DC_STAT__ECC_ERR_ST:
-		/* Dcache single-bit ECC error on small store */
-		sprintf(interp, "%s: DC_STAT[ECC_ERR_ST]\n "
-			"Dcache single-bit ECC error on small store", interr);
-		return;
-
-	case EV6__DC_STAT__ECC_ERR_LD:
-		switch (L->C_STAT) {
-		case 0:
-			/* Dcache single-bit error on speculative load */
-			/* Bcache victim read on Dcache/Bcache miss */
-			sprintf(interp, "%s: DC_STAT[ECC_ERR_LD] C_STAT=0\n "
-				"Dcache single-bit ECC error on speculative load",
-				slotb);
-			return;
-
-		case EV6__C_STAT__DSTREAM_DC_ERR:
-			/* Dcache single bit error on load */
-			sprintf(interp, "%s: DC_STAT[ECC_ERR_LD] C_STAT[DSTREAM_DC_ERR]\n"
-				" Dcache single-bit ECC error on speculative load, bit %d",
-				interr, ev6_syn2bit(L->DC0_SYNDROME, L->DC1_SYNDROME));
-			return;
-
-		case EV6__C_STAT__DSTREAM_BC_ERR:
-			/* Bcache single-bit error on Dcache fill */
-			sprintf(interp, "%s: DC_STAT[ECC_ERR_LD] C_STAT[DSTREAM_BC_ERR]\n"
-				" Bcache single-bit error on Dcache fill, bit %d",
-				slotb, ev6_syn2bit(L->DC0_SYNDROME, L->DC1_SYNDROME));
-			return;
-
-		case EV6__C_STAT__DSTREAM_MEM_ERR:
-			/* Memory single-bit error on Dcache fill */
-			sprintf(interp, "%s (to Dcache): DC_STAT[ECC_ERR_LD] "
-				"C_STAT[DSTREAM_MEM_ERR]\n "
-				"Memory single-bit error on Dcache fill, "
-				"Address 0x%lX, bit %d",
-				membus, L->C_ADDR, ev6_syn2bit(L->DC0_SYNDROME,
-							       L->DC1_SYNDROME));
-			return;
-		}
-	}
-
-	/* I-stream, other misc errors go on C_STAT alone */
-	switch (L->C_STAT) {
-	case EV6__C_STAT__ISTREAM_BC_ERR:
-		/* Bcache single-bit error on Icache fill (also MCHK) */
-		sprintf(interp, "%s: C_STAT[ISTREAM_BC_ERR]\n "
-			"Bcache single-bit error on Icache fill, bit %d",
-			slotb, ev6_syn2bit(L->DC0_SYNDROME, L->DC1_SYNDROME));
-		return;
-
-	case EV6__C_STAT__ISTREAM_MEM_ERR:
-		/* Memory single-bit error on Icache fill (also MCHK) */
-		sprintf(interp, "%s : C_STATISTREAM_MEM_ERR]\n "
-			"Memory single-bit error on Icache fill "
-			"addr 0x%lX, bit %d",
-			membus, L->C_ADDR, ev6_syn2bit(L->DC0_SYNDROME,
-						       L->DC1_SYNDROME));
-		return;
-
-	case EV6__C_STAT__PROBE_BC_ERR0:
-	case EV6__C_STAT__PROBE_BC_ERR1:
-		/* Bcache single-bit error on a probe hit */
-		sprintf(interp, "%s: C_STAT[PROBE_BC_ERR]\n "
-			"Bcache single-bit error on a probe hit, "
-			"addr 0x%lx, bit %d",
-			slotb, L->C_ADDR, ev6_syn2bit(L->DC0_SYNDROME,
-						      L->DC1_SYNDROME));
-		return;
-	}
-}
-
-static void
-ev6_mchk_interp(char *interp, struct el_common_EV6_mcheck * L)
-{
-	/* Machine check errors described by DC_STAT */
-	switch (L->DC_STAT) {
-	case EV6__DC_STAT__TPERR_P0:
-	case EV6__DC_STAT__TPERR_P1:
-		/* Dcache tag parity error (on retry) */
-		sprintf(interp, "%s: DC_STAT[TPERR_P0|TPERR_P1]\n "
-			"Dcache tag parity error(on retry)", interr);
-		return;
-
-	case EV6__DC_STAT__SEO:
-		/* Dcache second error on store */
-		sprintf(interp, "%s: DC_STAT[SEO]\n "
-			"Dcache second error during mcheck", interr);
-		return;
-	}
-
-	/* Machine check errors described by C_STAT */
-	switch (L->C_STAT) {
-	case EV6__C_STAT__DC_PERR:
-		/* Dcache duplicate tag parity error */
-		sprintf(interp, "%s: C_STAT[DC_PERR]\n "
-			"Dcache duplicate tag parity error at 0x%lX",
-			interr, L->C_ADDR);
-		return;
-
-	case EV6__C_STAT__BC_PERR:
-		/* Bcache tag parity error */
-		sprintf(interp, "%s: C_STAT[BC_PERR]\n "
-			"Bcache tag parity error at 0x%lX",
-			slotb, L->C_ADDR);
-		return;
-
-	case EV6__C_STAT__ISTREAM_BC_ERR:
-		/* Bcache single-bit error on Icache fill (also CRD) */
-		sprintf(interp, "%s: C_STAT[ISTREAM_BC_ERR]\n "
-			"Bcache single-bit error on Icache fill 0x%lX bit %d",
-			slotb, L->C_ADDR,
-			ev6_syn2bit(L->DC0_SYNDROME, L->DC1_SYNDROME));
-		return;
-
-
-	case EV6__C_STAT__ISTREAM_MEM_ERR:
-		/* Memory single-bit error on Icache fill (also CRD) */
-		sprintf(interp, "%s: C_STAT[ISTREAM_MEM_ERR]\n "
-			"Memory single-bit error on Icache fill 0x%lX, bit %d",
-			membus, L->C_ADDR,
-			ev6_syn2bit(L->DC0_SYNDROME, L->DC1_SYNDROME));
-		return;
-
-
-	case EV6__C_STAT__ISTREAM_BC_DBL:
-		/* Bcache double-bit error on Icache fill */
-		sprintf(interp, "%s: C_STAT[ISTREAM_BC_DBL]\n "
-			"Bcache double-bit error on Icache fill at 0x%lX",
-			slotb, L->C_ADDR);
-		return;
-	case EV6__C_STAT__DSTREAM_BC_DBL:
-		/* Bcache double-bit error on Dcache fill */
-		sprintf(interp, "%s: C_STAT[DSTREAM_BC_DBL]\n "
-			"Bcache double-bit error on Dcache fill at 0x%lX",
-			slotb, L->C_ADDR);
-		return;
-
-	case EV6__C_STAT__ISTREAM_MEM_DBL:
-		/* Memory double-bit error on Icache fill */
-		sprintf(interp, "%s: C_STAT[ISTREAM_MEM_DBL]\n "
-			"Memory double-bit error on Icache fill at 0x%lX",
-			membus, L->C_ADDR);
-		return;
-
-	case EV6__C_STAT__DSTREAM_MEM_DBL:
-		/* Memory double-bit error on Dcache fill */
-		sprintf(interp, "%s: C_STAT[DSTREAM_MEM_DBL]\n "
-			"Memory double-bit error on Dcache fill at 0x%lX",
-			membus, L->C_ADDR);
-		return;
-	}
-}
-
-static void
-ev6_cpu_machine_check(unsigned long vector, struct el_common_EV6_mcheck *L,
-		      struct pt_regs *regs)
-{
-	char interp[80];
-
-	/* This is verbose and looks intimidating.  Should it be printed for
-	   corrected (CRD) machine checks? */
-
-	printk(KERN_CRIT "PALcode logout frame:  "
-	       "MCHK_Code       %d  "
-	       "MCHK_Frame_Rev  %d\n"
-	       "I_STAT  %016lx  "
-	       "DC_STAT %016lx  "
-	       "C_ADDR  %016lx\n"
-	       "SYND1   %016lx  "
-	       "SYND0   %016lx  "
-	       "C_STAT  %016lx\n"
-	       "C_STS   %016lx  "
-	       "RES     %016lx  "
-	       "EXC_ADDR%016lx\n"
-	       "IER_CM  %016lx  "
-	       "ISUM    %016lx  "
-	       "MM_STAT %016lx\n"
-	       "PALBASE %016lx  "
-	       "I_CTL   %016lx  "
-	       "PCTX    %016lx\n"
-	       "CPU registers: "
-	       "PC      %016lx  "
-	       "Return  %016lx\n",
-	       L->MCHK_Code, L->MCHK_Frame_Rev, L->I_STAT, L->DC_STAT,
-	       L->C_ADDR, L->DC1_SYNDROME, L->DC0_SYNDROME, L->C_STAT,
-	       L->C_STS, L->RESERVED0, L->EXC_ADDR, L->IER_CM, L->ISUM,
-	       L->MM_STAT, L->PAL_BASE, L->I_CTL, L->PCTX,
-	       regs->pc, regs->r26);
-
-	/* Attempt an interpretation on the meanings of the fields above.  */
-	sprintf(interp, "No interpretation available!" );
-	if (vector == SCB_Q_PROCERR)
-		ev6_crd_interp(interp, L);
-	else if (vector == SCB_Q_PROCMCHK)
-		ev6_mchk_interp(interp, L);
-
-	printk(KERN_CRIT "interpretation: %s\n\n", interp);
-}
-
-
 /* Perform analysis of a machine check that arrived from the system (NMI) */
 
 static void
@@ -430,7 +120,6 @@ nautilus_machine_check(unsigned long vector, unsigned long la_ptr,
 		       struct pt_regs *regs)
 {
 	char *mchk_class;
-	unsigned cpu_analysis=0, sys_analysis=0;
 
 	/* Now for some analysis.  Machine checks fall into two classes --
 	   those picked up by the system, and those picked up by the CPU.
@@ -463,39 +152,20 @@ nautilus_machine_check(unsigned long vector, unsigned long la_ptr,
 		return;
 	}
 
-	switch (vector) {
-	case SCB_Q_SYSERR:
-		mchk_class = "Correctable System Machine Check (NMI)";
-		sys_analysis = 1;
-		break;
-	case SCB_Q_SYSMCHK:
-		mchk_class = "Fatal System Machine Check (NMI)";
-		sys_analysis = 1;
-		break;
-
-	case SCB_Q_PROCERR:
-		mchk_class = "Correctable Processor Machine Check";
-		cpu_analysis = 1;
-		break;
-	case SCB_Q_PROCMCHK:
-		mchk_class = "Fatal Processor Machine Check";
-		cpu_analysis = 1;
-		break;
-
-	default:
-		mchk_class = "Unknown vector!";
-		break;
+	if (vector == SCB_Q_SYSERR)
+		mchk_class = "Correctable";
+	else if (vector == SCB_Q_SYSMCHK)
+		mchk_class = "Fatal";
+	else {
+		ev6_machine_check(vector, la_ptr, regs);
+		return;
 	}
 
-	printk(KERN_CRIT "NAUTILUS Machine check 0x%lx [%s]\n",
+	printk(KERN_CRIT "NAUTILUS Machine check 0x%lx "
+			 "[%s  System Machine Check (NMI)]\n",
 	       vector, mchk_class);
 
-	if (cpu_analysis)
-		ev6_cpu_machine_check(vector,
-				      (struct el_common_EV6_mcheck *)la_ptr,
-				      regs);
-	if (sys_analysis)
-		naut_sys_machine_check(vector, la_ptr, regs);
+	naut_sys_machine_check(vector, la_ptr, regs);
 
 	/* Tell the PALcode to clear the machine check */
 	draina();
@@ -504,7 +174,6 @@ nautilus_machine_check(unsigned long vector, unsigned long la_ptr,
 }
 
 
-
 /*
  * The System Vectors
  */

include/asm-alpha/hwrpb.h

@@ -109,6 +109,9 @@ struct percpu_struct {
 	unsigned long ipc_buffer[21];
 	unsigned long palcode_avail[16];
 	unsigned long compatibility;
+	unsigned long console_data_log_pa;
+	unsigned long console_data_log_length;
+	unsigned long bcache_info;
 };
 
 struct procdesc_struct {

include/asm-alpha/system.h

@@ -61,7 +61,8 @@ struct el_common {
 	int		retry	:  1;	/* retry flag */
 	unsigned int	proc_offset;	/* processor-specific offset */
 	unsigned int	sys_offset;	/* system-specific offset */
-	unsigned long	code;		/* machine check code */
+	unsigned int	code;		/* machine check code */
+	unsigned int	frame_rev;	/* frame revision */
 };
 
 /* Machine Check Frame for uncorrectable errors (Large format)
@@ -117,11 +118,11 @@ struct el_common_EV6_mcheck {
 	unsigned long DC0_SYNDROME;
 	unsigned long C_STAT;
 	unsigned long C_STS;
-	unsigned long RESERVED0;
+	unsigned long MM_STAT;
 	unsigned long EXC_ADDR;
 	unsigned long IER_CM;
 	unsigned long ISUM;
-	unsigned long MM_STAT;
+	unsigned long RESERVED0;
 	unsigned long PAL_BASE;
 	unsigned long I_CTL;
 	unsigned long PCTX;