xref: /illumos-gate/usr/src/cmd/fm/eversholt/files/sparc/sun4u/psycho.esc (revision b7d3956b)

/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License (the "License").
 * You may not use this file except in compliance with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */
/*
 * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
 * Use is subject to license terms.
 */

#pragma ident	"%Z%%M%	%I%	%E% SMI"

#pragma dictionary "SUN4U"

#define	AGENT_ID_MASK	0x1f
#define	AGENT_ID_SHIFT	24

#define	HB_FIT		1000
#define HBUS_FIT	1000
#define PCI_BUS_FIT	500
#define PCI_DEV_FIT	1000
#define	CPU_FIT		500

#define	PCI_HB_DEV_PATH hostbridge/pcibus/pcidev[32]/pcifn[0]

fru hostbridge;
asru hostbridge;

event fault.io.psycho@hostbridge,
    FITrate=HB_FIT, FRU=hostbridge, ASRU=hostbridge;

event error.io.psy.ecc.thresh@hostbridge;
event ereport.io.psy.ecc.pue@hostbridge{within(5s)};
event ereport.io.psy.ecc.s-pue@hostbridge{within(5s)};
event ereport.io.psy.ecc.thresh@hostbridge;

/*
 * A faulty Psycho hostbridge may cause:
 *
 *  - pue:	the psycho to detect a PIO uncorrectable error, bad reader.
 *  - s-pue:	the psycho to detect a secondary PIO UE, bad reader.
 *  - ecc:	the SERD engine to gather enough PIO CEs to generate an ereport.
 */
prop fault.io.psycho@hostbridge (0)->
    ereport.io.psy.ecc.pue@hostbridge,
    ereport.io.psy.ecc.s-pue@hostbridge,
    error.io.psy.ecc.thresh@hostbridge;

engine serd.io.psycho.ecc@hostbridge,
    N=3, T=1day, method=persistent,
    trip=ereport.io.psy.ecc.thresh@hostbridge;

event upset.io.psycho@hostbridge,
    engine=serd.io.psycho.ecc@hostbridge;

event ereport.io.psy.ecc.pce@hostbridge{within(5s)};
event ereport.io.psy.ecc.s-pce@hostbridge{within(5s)};

prop error.io.psy.ecc.thresh@hostbridge (2)->
    ereport.io.psy.ecc.thresh@hostbridge,
    ereport.io.psy.ecc.pce@hostbridge;

/*
 * An upset Psycho may cause:
 *
 *  - pce:	the psycho to detect a PIO correctable error, bad reader.
 */
prop upset.io.psycho@hostbridge (0)->
    ereport.io.psy.ecc.pce@hostbridge;

event fault.io.hbus@hostbridge,
    FITrate=HBUS_FIT, FRU=hostbridge, ASRU=hostbridge;

/*
 * A faulty host bus may cause:
 *
 *  - pue:	a PIO uncorrectable error.
 *  - s-pue:	a secondary PIO UE.
 *  - ecc:	the SERD engine to gather enough PIO CEs to generate an ereport.
 */
prop fault.io.hbus@hostbridge (0)->
    ereport.io.psy.ecc.pue@hostbridge,
    ereport.io.psy.ecc.s-pue@hostbridge,
    error.io.psy.ecc.thresh@hostbridge;

fru cpu;

event fault.io.datapath@cpu, retire=0,
    FITrate=CPU_FIT, FRU=cpu;

event error.io.cpu.ecc.thresh@cpu;

/*
 * A faulty CPU may cause:
 *
 *  - pue:	a PIO uncorrectable error, where the captured Agentid matches
 *		the cpuid.
 *  - ecc:	the SERD engine for the CPU to fire due to PIO CEs from this
 *		CPU
 */
prop fault.io.datapath@cpu[cpuid] (0)->
    ereport.io.psy.ecc.pue@hostbridge
    {((payloadprop("ecc-afsr") >> AGENT_ID_SHIFT) & AGENT_ID_MASK) == cpuid};

prop fault.io.datapath@cpu (0)->
    error.io.cpu.ecc.thresh@cpu;

prop error.io.cpu.ecc.thresh@cpu (1)->
    ereport.io.psy.ecc.thresh@hostbridge<>;

prop error.io.cpu.ecc.thresh@cpu[cpuid] (1)->
    ereport.io.psy.ecc.pce@hostbridge<>
    {((payloadprop("ecc-afsr") >> AGENT_ID_SHIFT) & AGENT_ID_MASK) == cpuid};

asru pcibus/pcidev/pcifn;
fru pcibus/pcidev;

event fault.io.pci.device-interr@hostbridge/pcibus/pcidev/pcifn,
    FITrate=PCI_DEV_FIT, FRU=pcibus/pcidev, ASRU=pcibus/pcidev/pcifn;

event fault.io.pci.device-interr@pcibus/pcidev/pcifn,
    FITrate=PCI_DEV_FIT, FRU=pcibus/pcidev, ASRU=pcibus/pcidev/pcifn;

event error.io.psy.pbm.rl@hostbridge/pcibus/pcidev/pcifn;
event error.io.psy.pbm.rl@pcibus/pcidev/pcifn;
event error.io.psy.pbm.rl@pcibus/pcidev/pcifn/pcibus/pcidev/pcifn;
event error.io.psy.pbm.target-rl@pcibus/pcidev/pcifn;
event error.io.psy.pbm.target-rl@pcibus/pcidev/pcifn/pcibus/pcidev/pcifn;
event error.io.pci.ma-u@hostbridge/pcibus/pcidev/pcifn;
event error.io.pci.perr-pw-u@hostbridge/pcibus/pcidev/pcifn;
event error.io.pci.perr-dw-u@hostbridge/pcibus/pcidev/pcifn;
event error.io.pci.dpdata-dr-u@hostbridge/pcibus/pcidev/pcifn;
event error.io.pci.ta-u@hostbridge/pcibus/pcidev/pcifn;
event error.io.pci.serr-u@hostbridge/pcibus/pcidev/pcifn;
event error.io.pci.retry-to-d@hostbridge/pcibus/pcidev/pcifn;
event error.psy.cpu.berr@cpu;

event ereport.io.psy.sbh@hostbridge/pcibus/pcidev/pcifn{within(5s)};
event ereport.io.psy.pbm.rl@hostbridge/pcibus/pcidev/pcifn{within(5s)};
event ereport.io.psy.pbm.s-rl@hostbridge/pcibus/pcidev/pcifn{within(5s)};
event ereport.io.psy.pbm.s-ma@hostbridge/pcibus/pcidev/pcifn{within(5s)};
event ereport.io.psy.pbm.s-rta@hostbridge/pcibus/pcidev/pcifn{within(5s)};
event ereport.io.psy.pbm.s-mdpe@hostbridge/pcibus/pcidev/pcifn{within(5s)};
event ereport.io.psy.pbm.target-rl@pcibus/pcidev/pcifn{within(5s)};
event ereport.io.pci.rserr@hostbridge/pcibus/pcidev/pcifn{within(5s)};
event ereport.cpu.ultraSPARC-II.berr@cpu{within(5s)};

/*
 * A faulty PCI device may cause:
 *
 *  - rl:	it to retry a transaction beyond the specified limit.
 *  - sbh:	it to generate a streaming byte hole.
 *
 * For rl, there may be a target-rl ereport on a child device. There may also be
 * an associated dto - the retry-to-d error propagates into the pci.esc rules
 * to handle this.
 */

prop fault.io.pci.device-interr@pcibus/pcidev[fromdev]/pcifn (0)->
    error.io.psy.pbm.rl@pcibus/pcidev<todev>/pcifn {
	fromdev == todev && fromdev != 32 },
    error.io.psy.pbm.target-rl@pcibus/pcidev<todev>/pcifn {
	fromdev == todev && fromdev != 32 };

prop error.io.psy.pbm.rl@pcibus/pcidev/pcifn/pcibus/pcidev/pcifn (1)->
    error.io.psy.pbm.rl@pcibus/pcidev/pcifn;

prop error.io.psy.pbm.rl@hostbridge/pcibus/pcidev/pcifn (1)->
    ereport.io.psy.pbm.rl@PCI_HB_DEV_PATH,
    ereport.io.psy.pbm.s-rl@PCI_HB_DEV_PATH;

prop error.io.psy.pbm.target-rl@pcibus/pcidev/pcifn (1)->
    error.io.psy.pbm.target-rl@pcibus/pcidev/pcifn/pcibus<>/pcidev<>/pcifn<>;

prop error.io.psy.pbm.target-rl@pcibus/pcidev/pcifn (0)->
    ereport.io.psy.pbm.target-rl@pcibus/pcidev/pcifn;

prop error.io.psy.pbm.rl@hostbridge/pcibus/pcidev/pcifn (0)->
    error.io.pci.retry-to-d@hostbridge/pcibus/pcidev/pcifn;

prop error.io.psy.pbm.rl@hostbridge/pcibus/pcidev/pcifn (0)->
    error.psy.cpu.berr@cpu;

prop fault.io.pci.device-interr@hostbridge/pcibus/pcidev/pcifn (0)->
    ereport.io.psy.sbh@PCI_HB_DEV_PATH;

/*
 * Need to add the following psycho specific propagations to complete the PCI
 * fault tree. These are to allow propagations to secondary errors and cpu
 * bus errors, and to represent the way the chip raises rserr
 * on detection of SERR#
 */
prop error.io.pci.ma-u@hostbridge/pcibus/pcidev/pcifn (0)->
    ereport.io.psy.pbm.s-ma@PCI_HB_DEV_PATH;

prop error.io.pci.ta-u@hostbridge/pcibus/pcidev/pcifn (0)->
    ereport.io.psy.pbm.s-rta@PCI_HB_DEV_PATH;

prop error.io.pci.perr-pw-u@hostbridge/pcibus/pcidev/pcifn (0)->
    ereport.io.psy.pbm.s-mdpe@PCI_HB_DEV_PATH;

prop error.io.pci.perr-dw-u@hostbridge/pcibus/pcidev/pcifn (0)->
    ereport.io.psy.pbm.s-mdpe@PCI_HB_DEV_PATH;

prop error.io.pci.dpdata-dr-u@hostbridge/pcibus/pcidev/pcifn (0)->
    ereport.io.psy.pbm.s-mdpe@PCI_HB_DEV_PATH;

prop error.io.pci.ta-u@hostbridge/pcibus/pcidev/pcifn (0)->
    error.psy.cpu.berr@cpu;

prop error.io.pci.dpdata-dr-u@hostbridge/pcibus/pcidev/pcifn (0)->
    error.psy.cpu.berr@cpu;

prop error.io.pci.ma-u@hostbridge/pcibus/pcidev/pcifn (0)->
    error.psy.cpu.berr@cpu;

prop error.psy.cpu.berr@cpu (0)->
    ereport.cpu.ultraSPARC-II.berr@cpu;

prop error.io.pci.serr-u@hostbridge/pcibus/pcidev/pcifn (1)->
    ereport.io.pci.rserr@PCI_HB_DEV_PATH;

event ereport.io.psy.nodiag@hostbridge;

/*
 * Upset used to hide ereports that can not be currently diagnosed.
 */
engine serd.io.psy.nodiag@hostbridge,
    N=1000, T=1hour, method=persistent,
    trip=ereport.io.psy.nodiag@hostbridge;

event upset.io.psy.nodiag@hostbridge,
    engine=serd.io.psy.nodiag@hostbridge;

prop upset.io.psy.nodiag@hostbridge (0)->
    ereport.io.psy.ecc.s-pce@hostbridge,
    ereport.io.psy.nodiag@hostbridge;