usr/src/cmd/fm/modules/sun4/cpumem-diagnosis/cmd_mem.h

   1 /*
   2  * CDDL HEADER START
   3  *
   4  * The contents of this file are subject to the terms of the
   5  * Common Development and Distribution License (the "License").
   6  * You may not use this file except in compliance with the License.
   7  *
   8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
   9  * or http://www.opensolaris.org/os/licensing.
  10  * See the License for the specific language governing permissions
  11  * and limitations under the License.
  12  *
  13  * When distributing Covered Code, include this CDDL HEADER in each
  14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
  15  * If applicable, add the following below this CDDL HEADER, with the
  16  * fields enclosed by brackets "[]" replaced with your own identifying
  17  * information: Portions Copyright [yyyy] [name of copyright owner]
  18  *
  19  * CDDL HEADER END
  20  */
  21 /*
  22  * Copyright (c) 2004, 2010, Oracle and/or its affiliates. All rights reserved.
  23  */
  24
  25 #ifndef _CMD_MEM_H
  26 #define _CMD_MEM_H
  27
  28 /*
  29  * Support routines for managing state related to memory modules.
  30  *
  31  * Correctable errors generally cause changes to the DIMM-related state (see
  32  * cmd_dimm.c), whereas uncorrectable errors tend to use the bank-related
  33  * routines (see cmd_bank.c).  The primary exception to this division (though
  34  * it eventually devolves to one of the two) is the RxE/FRx pair emitted by
  35  * UltraSPARC-IIIi processors.  With these errors, a complete pair must be
  36  * received and matched before we know whether we're dealing with a CE or a UE.
  37  */
  38
  39 #include <cmd.h>
  40 #include <cmd_state.h>
  41 #include <cmd_fmri.h>
  42 #include <sys/errclassify.h>
  43 #include <cmd_cpu.h>
  44
  45 #ifdef __cplusplus
  46 extern "C" {
  47 #endif
  48
  49 #define CMD_MEM_F_FAULTING      0x1
  50
  51 /*
  52  * Used to store as-yet unmatched IOxEs, RxEs, and FRxs.  When a new IOxE,
  53  * RxE or FRx arrives, we traverse the cmd.cmd_iorxefrx list, looking for
  54  * matching entries.  Matching has a cpuid-based component, as well as a
  55  * temporal one.  We can compare the cpuids directly, using the cmd_iorxefrx_t
  56  * and the newly-received event. Temporal comparison isn't performed directly.
  57  * Instead, we ensure that entries in the iorxefrx list are removed when they
  58  * expire by means of timers. This frees the matching code from the need to
  59  * worry about time.
  60  */
  61 typedef struct cmd_iorxefrx {
  62         cmd_list_t rf_list;             /* List of cmd_iorxefrx_t's */
  63         cmd_errcl_t rf_errcl;           /* Error type (CMD_ERRCL_*) */
  64         uint_t rf_afsr_agentid;         /* Remote Agent ID (from AFSR) */
  65         uint_t rf_det_agentid;          /* Locat Agent ID (from detector) */
  66         id_t rf_expid;                  /* Timer ID for entry expiration */
  67         uint64_t rf_afar;               /* Valid for RxE only */
  68         uint8_t rf_afar_status;         /* Valid for RxE only */
  69         ce_dispact_t rf_type;           /* Valid for RxE only */
  70         uint16_t rf_synd;               /* Valid for FRx only */
  71         uint8_t rf_synd_status;         /* Valid for FRx only */
  72         uint64_t rf_afsr;               /* Valid for FRx only */
  73         uint64_t rf_disp;               /* Valid for RCE only */
  74 } cmd_iorxefrx_t;
  75
  76 typedef struct cmd_dimm cmd_dimm_t;
  77 typedef struct cmd_bank cmd_bank_t;
  78 #ifdef sun4v
  79 typedef struct cmd_branch cmd_branch_t;
  80 #endif
  81
  82 /*
  83  * Correctable and Uncorrectable memory errors
  84  *
  85  * CEs of "Unknown" or "Intermittent" classification are not used in diagnosis.
  86  *
  87  * "Persistent" CEs are added to per-DIMM SERD engines.  When the
  88  * engine for a given DIMM fires, the page corresponding to the CE that
  89  * caused the engine to fire is retired, and the SERD engine for that
  90  * DIMM is reset.
  91  *
  92  * "Possibly Persistent" CEs are at least Persistent and so are treated
  93  * as "Persistent" errors above, being added to the same SERD engines.
  94  *
  95  * "Leaky" CEs and "Sticky" CEs trigger immediate page retirement.
  96  *
  97  * "Possibly Sticky" CEs to which no valid partner test has been applied
  98  * are not used in diagnosis.  Where a valid partner test has been applied
  99  * but did not confirm "Sticky" status there is a _suggestion_ that the
 100  * original cpu may be a bad reader or writer or suffering from other
 101  * datapath issues.  To avoid retiring pages for such non-DIMM problems
 102  * these classifications are also not used in diagnosis.
 103  *
 104  * UEs immediately trigger page retirements, but do not affect the CE SERD
 105  * engines.  In addition, UEs are recorded in the UE caches of the detecting
 106  * CPUs.  When a page is to be retired, a fault.memory.page fault is
 107  * generated.
 108  *
 109  */
 110
 111 typedef cmd_evdisp_t cmd_xe_handler_f(fmd_hdl_t *, fmd_event_t *, nvlist_t *,
 112     const char *, uint64_t, uint8_t, uint16_t, uint8_t, ce_dispact_t, uint64_t,
 113     nvlist_t *);
 114
 115 extern ce_dispact_t cmd_mem_name2type(const char *, int);
 116 extern int cmd_synd2upos(uint16_t);
 117 extern cmd_evdisp_t cmd_ce(fmd_hdl_t *, fmd_event_t *, nvlist_t *,
 118     const char *, cmd_errcl_t);
 119 extern cmd_evdisp_t cmd_ue(fmd_hdl_t *, fmd_event_t *, nvlist_t *,
 120     const char *, cmd_errcl_t);
 121 extern cmd_evdisp_t cmd_ce_common(fmd_hdl_t *, fmd_event_t *, nvlist_t *,
 122     const char *, uint64_t, uint8_t, uint16_t, uint8_t,
 123     ce_dispact_t, uint64_t, nvlist_t *);
 124 extern cmd_evdisp_t cmd_ue_common(fmd_hdl_t *, fmd_event_t *, nvlist_t *,
 125     const char *, uint64_t, uint8_t, uint16_t, uint8_t,
 126     ce_dispact_t, uint64_t, nvlist_t *);
 127 extern cmd_evdisp_t cmd_mem_synd_check(fmd_hdl_t *, uint64_t, uint8_t,
 128     uint16_t, uint8_t, cmd_cpu_t *);
 129 extern void cmd_dimm_close(fmd_hdl_t *, void *);
 130 extern void cmd_bank_close(fmd_hdl_t *, void *);
 131 extern int cmd_same_datapath_dimms(cmd_dimm_t *, cmd_dimm_t *);
 132 extern void cmd_gen_datapath_fault(fmd_hdl_t *, cmd_dimm_t *, cmd_dimm_t *,
 133     uint16_t, nvlist_t *);
 134 extern void cmd_to_hashed_addr(uint64_t *, uint64_t, const char *);
 135
 136 #ifdef sun4u
 137 extern char *cmd_cpu_getfrustr_by_id(fmd_hdl_t *, uint32_t);
 138 #endif
 139
 140 #ifdef sun4v
 141 extern void cmd_branch_close(fmd_hdl_t *, void *);
 142 extern cmd_evdisp_t cmd_fb(fmd_hdl_t *, fmd_event_t *, nvlist_t *,
 143     const char *, cmd_errcl_t);
 144 extern cmd_evdisp_t cmd_fw_defect(fmd_hdl_t *, fmd_event_t *, nvlist_t *,
 145     const char *, cmd_errcl_t);
 146 extern cmd_evdisp_t cmd_fb_train(fmd_hdl_t *, fmd_event_t *, nvlist_t *,
 147     const char *, cmd_errcl_t);
 148 extern cmd_evdisp_t cmd_ue_train(fmd_hdl_t *, fmd_event_t *, nvlist_t *,
 149     const char *, cmd_errcl_t);
 150 #endif
 151
 152 /*
 153  * US-IIIi I/O, Remote and Foreign Read memory errors
 154  *
 155  * When one processor or I/O bridge attempts to read memory local to
 156  * another processor, one each of IOCE/IOUE/RCE/RUE and FRC/FRU will be
 157  * generated, depending on the type of error.  Both the IOxE/RxE and the FRx
 158  * are needed, as each contains data necessary to the diagnosis of the error.
 159  * Upon receipt of one of the errors, we wait until we receive the other.
 160  * When the pair has been successfully received and matched, a CE or UE,
 161  * as appropriate, is synthesized from the data in the matched ereports.
 162  * The synthesized ereports are handled by the normal CE and UE mechanisms.
 163  */
 164 extern cmd_evdisp_t cmd_frx(fmd_hdl_t *, fmd_event_t *, nvlist_t *,
 165     const char *, cmd_errcl_t);
 166 extern cmd_evdisp_t cmd_rxe(fmd_hdl_t *, fmd_event_t *, nvlist_t *,
 167     const char *, cmd_errcl_t);
 168 extern cmd_evdisp_t cmd_ioxe(fmd_hdl_t *, fmd_event_t *, nvlist_t *,
 169     const char *, cmd_errcl_t);
 170 extern cmd_evdisp_t cmd_ioxe_sec(fmd_hdl_t *, fmd_event_t *, nvlist_t *,
 171     const char *, cmd_errcl_t);
 172 extern cmd_evdisp_t cmd_rxefrx_common(fmd_hdl_t *hdl, fmd_event_t *ep,
 173     nvlist_t *nvl, const char *class, cmd_errcl_t clcode,
 174     cmd_errcl_t matchmask);
 175
 176 /*
 177  * A list of received IOxE/RxE/FRx ereports is maintained for correlation
 178  * purposes (see above).  These two routines manage the addition of new
 179  * ereports, and the retrieval of existing ones.  Pruning of the list is
 180  * handled automatically.
 181  */
 182 extern void cmd_iorxefrx_queue(fmd_hdl_t *, cmd_iorxefrx_t *);
 183 extern void cmd_iorxefrx_free(fmd_hdl_t *, cmd_iorxefrx_t *);
 184
 185 extern const char *cmd_fmri_get_unum(nvlist_t *);
 186 extern nvlist_t *cmd_mem_fmri_create(const char *, char **, size_t);
 187 extern nvlist_t *cmd_mem_fmri_derive(fmd_hdl_t *, uint64_t, uint64_t, uint16_t);
 188
 189 extern void cmd_mem_case_restore(fmd_hdl_t *, cmd_case_t *, fmd_case_t *,
 190     const char *, const char *);
 191 extern char *cmd_mem_serdnm_create(fmd_hdl_t *, const char *, const char *);
 192 extern char *cmd_page_serdnm_create(fmd_hdl_t *, const char *, uint64_t);
 193 extern char *cmd_mq_serdnm_create(fmd_hdl_t *, const char *, uint64_t,
 194     uint16_t, uint16_t);
 195 extern void cmd_mem_retirestat_create(fmd_hdl_t *, fmd_stat_t *, const char *,
 196     uint64_t, const char *);
 197 extern int cmd_mem_thresh_check(fmd_hdl_t *, uint_t);
 198 extern ulong_t cmd_mem_get_phys_pages(fmd_hdl_t *);
 199
 200 extern void cmd_mem_timeout(fmd_hdl_t *, id_t);
 201 extern void cmd_mem_gc(fmd_hdl_t *);
 202 extern void cmd_mem_fini(fmd_hdl_t *);
 203
 204 #ifdef __cplusplus
 205 }
 206 #endif
 207
 208 #endif /* _CMD_MEM_H */