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, Version 1.0 only
6 * (the "License"). You may not use this file except in compliance
7 * with the License.
8 *
9 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
10 * or http://www.opensolaris.org/os/licensing.
11 * See the License for the specific language governing permissions
12 * and limitations under the License.
13 *
14 * When distributing Covered Code, include this CDDL HEADER in each
15 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
16 * If applicable, add the following below this CDDL HEADER, with the
17 * fields enclosed by brackets "[]" replaced with your own identifying
18 * information: Portions Copyright [yyyy] [name of copyright owner]
19 *
20 * CDDL HEADER END
21 */
22
23 /*
24 * Copyright 2005 Sun Microsystems, Inc. All rights reserved.
25 * Use is subject to license terms.
26 */
27
28 #pragma ident "%Z%%M% %I% %E% SMI"
29
30 #include <cda.h>
31
32 #include <strings.h>
33 #include <errno.h>
34 #include <time.h>
35 #include <sys/utsname.h>
36 #include <sys/systeminfo.h>
37 #include <fm/fmd_api.h>
38 #include <sys/fm/protocol.h>
39
40 cda_t cda;
41
42 cda_stats_t cda_stats = {
43 { "dp_offs", FMD_TYPE_UINT64, "successful cpu offlines" },
44 { "dp_fails", FMD_TYPE_UINT64, "datapath faults unresolveable" },
45 { "cpu_supp", FMD_TYPE_UINT64, "cpu offlines suppressed" },
46 { "bad_flts", FMD_TYPE_UINT64, "invalid fault events received" },
47 { "nop_flts", FMD_TYPE_UINT64, "inapplicable fault events received" },
48 };
49
50 typedef struct cda_subscriber {
51 const char *subr_class;
52 const char *subr_sname;
53 uint_t subr_svers;
54 void (*subr_func)(fmd_hdl_t *, nvlist_t *, nvlist_t *, const char *);
55 } cda_subscriber_t;
56
57 static const cda_subscriber_t cda_subrs[] = {
58 { "fault.asic.*.dp", FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION,
59 cda_dp_retire },
60 { NULL, NULL, 0, NULL }
61 };
62
63 static const cda_subscriber_t *
cda_get_subr(fmd_hdl_t * hdl,nvlist_t * nvl,nvlist_t ** asrup)64 cda_get_subr(fmd_hdl_t *hdl, nvlist_t *nvl, nvlist_t **asrup)
65 {
66 const cda_subscriber_t *sp;
67 nvlist_t *asru;
68 char *scheme;
69 uint8_t version;
70
71 if (nvlist_lookup_nvlist(nvl, FM_FAULT_ASRU, &asru) != 0 ||
72 nvlist_lookup_string(asru, FM_FMRI_SCHEME, &scheme) != 0 ||
73 nvlist_lookup_uint8(asru, FM_VERSION, &version) != 0) {
74 cda_stats.bad_flts.fmds_value.ui64++;
75 return (NULL);
76 }
77
78 for (sp = cda_subrs; sp->subr_class != NULL; sp++) {
79 if (fmd_nvl_class_match(hdl, nvl, sp->subr_class) &&
80 strcmp(scheme, sp->subr_sname) == 0 &&
81 version <= sp->subr_svers) {
82 *asrup = asru;
83 return (sp);
84 }
85 }
86
87 cda_stats.nop_flts.fmds_value.ui64++;
88 return (NULL);
89 }
90
91 static void
cda_recv_list(fmd_hdl_t * hdl,nvlist_t * nvl)92 cda_recv_list(fmd_hdl_t *hdl, nvlist_t *nvl)
93 {
94 char *uuid = NULL;
95 nvlist_t **nva;
96 uint_t nvc;
97 int err = 0;
98
99 err |= nvlist_lookup_string(nvl, FM_SUSPECT_UUID, &uuid);
100 err |= nvlist_lookup_nvlist_array(nvl, FM_SUSPECT_FAULT_LIST,
101 &nva, &nvc);
102 if (err != 0) {
103 cda_stats.bad_flts.fmds_value.ui64++;
104 return;
105 }
106
107 while (nvc-- != 0) {
108 nvlist_t *nvl = *nva++;
109 const cda_subscriber_t *subr;
110 nvlist_t *asru;
111
112 if (fmd_case_uuclosed(hdl, uuid))
113 break;
114
115 if ((subr = cda_get_subr(hdl, nvl, &asru)) == NULL)
116 continue;
117
118 if (subr->subr_func != NULL)
119 subr->subr_func(hdl, nvl, asru, uuid);
120 }
121 }
122
123 static void
cda_recv_one(fmd_hdl_t * hdl,nvlist_t * nvl)124 cda_recv_one(fmd_hdl_t *hdl, nvlist_t *nvl)
125 {
126 const cda_subscriber_t *subr;
127 nvlist_t *asru;
128
129 if ((subr = cda_get_subr(hdl, nvl, &asru)) == NULL)
130 return;
131
132 if (subr->subr_func != NULL)
133 subr->subr_func(hdl, nvl, asru, NULL);
134 }
135
136 /*ARGSUSED*/
137 static void
cda_recv(fmd_hdl_t * hdl,fmd_event_t * ep,nvlist_t * nvl,const char * class)138 cda_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl, const char *class)
139 {
140 fmd_hdl_debug(hdl, "received %s\n", class);
141
142 if (strcmp(class, FM_LIST_SUSPECT_CLASS) == 0)
143 cda_recv_list(hdl, nvl);
144 else
145 cda_recv_one(hdl, nvl);
146 }
147
148 static const fmd_hdl_ops_t fmd_ops = {
149 cda_recv, /* fmdo_recv */
150 NULL, /* fmdo_timeout */
151 NULL, /* fmdo_close */
152 NULL, /* fmdo_stats */
153 NULL, /* fmdo_gc */
154 };
155
156 static const fmd_prop_t fmd_props[] = {
157 { "cpu_tries", FMD_TYPE_UINT32, "10" },
158 { "cpu_delay", FMD_TYPE_TIME, "1sec" },
159 { "cpu_offline_enable", FMD_TYPE_BOOL, "true" },
160 { "cpu_forced_offline", FMD_TYPE_BOOL, "true" },
161 { NULL, 0, NULL }
162 };
163
164 static const fmd_hdl_info_t fmd_info = {
165 "Datapath Retire Agent", CDA_VERSION, &fmd_ops, fmd_props
166 };
167
168 static int
cda_platform_check_support(fmd_hdl_t * hdl)169 cda_platform_check_support(fmd_hdl_t *hdl)
170 {
171 char buf[SYS_NMLN];
172
173 if (sysinfo(SI_PLATFORM, buf, sizeof (buf)) == -1) {
174 fmd_hdl_debug(hdl, "sysinfo failed");
175 return (0);
176 }
177
178 if (strcmp(buf, "SUNW,Sun-Fire-15000") == 0 ||
179 strcmp(buf, "SUNW,Sun-Fire") == 0 ||
180 strcmp(buf, "SUNW,Netra-T12") == 0)
181 return (1);
182 else
183 return (0);
184 }
185
186 void
_fmd_init(fmd_hdl_t * hdl)187 _fmd_init(fmd_hdl_t *hdl)
188 {
189 hrtime_t nsec;
190
191 if (fmd_hdl_register(hdl, FMD_API_VERSION, &fmd_info) != 0)
192 return; /* invalid data in configuration file */
193
194 if (cda_platform_check_support(hdl) == 0) {
195 fmd_hdl_debug(hdl, "platform not supported");
196 fmd_hdl_unregister(hdl);
197 return;
198 }
199
200 fmd_hdl_subscribe(hdl, "fault.asic.*.dp");
201
202 (void) fmd_stat_create(hdl, FMD_STAT_NOALLOC, sizeof (cda_stats) /
203 sizeof (fmd_stat_t), (fmd_stat_t *)&cda_stats);
204
205 cda.cda_cpu_tries = fmd_prop_get_int32(hdl, "cpu_tries");
206
207 nsec = fmd_prop_get_int64(hdl, "cpu_delay");
208 cda.cda_cpu_delay.tv_sec = nsec / NANOSEC;
209 cda.cda_cpu_delay.tv_nsec = nsec % NANOSEC;
210
211 cda.cda_cpu_dooffline = fmd_prop_get_int32(hdl,
212 "cpu_offline_enable");
213 cda.cda_cpu_forcedoffline = fmd_prop_get_int32(hdl,
214 "cpu_forced_offline");
215 }
216
217 /*ARGSUSED*/
218 void
_fmd_fini(fmd_hdl_t * hdl)219 _fmd_fini(fmd_hdl_t *hdl)
220 {
221 }
222