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 /*
23 * Copyright 2010 Sun Microsystems, Inc. All rights reserved.
24 * Use is subject to license terms.
25 * Copyright 2012 Joyent, Inc. All rights reserved.
26 */
27
28 /*
29 * /dev/ipmi IPMI monitor
30 *
31 * The purpose of this module is to monitor the connection between the system
32 * and the service processor attached via /dev/ipmi0. The module assumes the SP
33 * supports the Sun OEM uptime IPMI command. If the BMC connection does not
34 * exist, or the uptime function is not implemented, then the module unloads
35 * without doing anything.
36 *
37 * When the module is first loaded, or a reset is detected, the module will
38 * generate the ESC_PLATFORM_SP_RESET sysevent as a system-wide notification to
39 * indicate that this event has occurred.
40 *
41 * Note that this event generation is not guaranteed to have a one-to-one
42 * correspondence with an SP reset. There is no persistence, so if fmd is
43 * restarted we will generate this event again. Thus the event only indicates
44 * the possibility that the SP has been reset. This could be enhanced using fmd
45 * checkpoints to have some persistent state to avoid this scenario. However,
46 * it currently serves the useful dual purpose of notifying consumers of system
47 * startup as well as SP reset through a single channel.
48 */
49
50 #include <errno.h>
51 #include <libipmi.h>
52 #include <libsysevent.h>
53 #include <string.h>
54 #include <fm/fmd_api.h>
55 #include <sys/sysevent/eventdefs.h>
56
57 typedef struct sp_monitor {
58 ipmi_handle_t *sm_hdl;
59 uint32_t sm_seconds;
60 uint32_t sm_generation;
61 hrtime_t sm_interval;
62 } sp_monitor_t;
63
64 static void
sp_post_sysevent(fmd_hdl_t * hdl)65 sp_post_sysevent(fmd_hdl_t *hdl)
66 {
67 sp_monitor_t *smp = fmd_hdl_getspecific(hdl);
68 sysevent_id_t eid;
69
70 fmd_hdl_debug(hdl, "SP reset detected, posting sysevent");
71
72 if (sysevent_post_event(EC_PLATFORM, ESC_PLATFORM_SP_RESET,
73 SUNW_VENDOR, "fmd", NULL, &eid) != 0) {
74 fmd_hdl_debug(hdl, "failed to send sysevent: %s",
75 strerror(errno));
76 /*
77 * We reset the seconds and generation so that the next time
78 * through we will try to post the sysevent again.
79 */
80 smp->sm_seconds = -1U;
81 smp->sm_generation = -1U;
82 }
83 }
84
85 /*ARGSUSED*/
86 static void
sp_timeout(fmd_hdl_t * hdl,id_t id,void * data)87 sp_timeout(fmd_hdl_t *hdl, id_t id, void *data)
88 {
89 sp_monitor_t *smp = fmd_hdl_getspecific(hdl);
90 uint32_t seconds, generation;
91
92 if (ipmi_sunoem_uptime(smp->sm_hdl, &seconds, &generation) != 0) {
93 /*
94 * Ignore uptime failures. We will generate the appropriate
95 * event when it comes back online.
96 */
97 fmd_hdl_debug(hdl, "failed to get uptime: %s",
98 ipmi_errmsg(smp->sm_hdl));
99 } else {
100 /*
101 * We want to catch cases where the generation number is
102 * explicitly reset, or when the SP configuration is reset after
103 * a reboot (and the generation number is 0). We also post a
104 * sysevent when the module initially loads, since we can't be
105 * sure if we missed a SP reset or not.
106 */
107 if (seconds < smp->sm_seconds ||
108 generation != smp->sm_generation ||
109 smp->sm_seconds == 0)
110 sp_post_sysevent(hdl);
111
112 smp->sm_seconds = seconds;
113 smp->sm_generation = generation;
114 }
115
116 (void) fmd_timer_install(hdl, NULL, NULL, smp->sm_interval);
117 }
118
119 static const fmd_hdl_ops_t fmd_ops = {
120 NULL, /* fmdo_recv */
121 sp_timeout, /* fmdo_timeout */
122 NULL, /* fmdo_close */
123 NULL, /* fmdo_stats */
124 NULL, /* fmdo_gc */
125 };
126
127 static const fmd_prop_t fmd_props[] = {
128 { "interval", FMD_TYPE_TIME, "60sec" },
129 { NULL, 0, NULL }
130 };
131
132 static const fmd_hdl_info_t fmd_info = {
133 "Service Processor Monitor", "1.0", &fmd_ops, fmd_props
134 };
135
136 void
_fmd_init(fmd_hdl_t * hdl)137 _fmd_init(fmd_hdl_t *hdl)
138 {
139 sp_monitor_t *smp;
140 int error;
141 char *msg;
142
143 if (fmd_hdl_register(hdl, FMD_API_VERSION, &fmd_info) != 0)
144 return;
145
146 smp = fmd_hdl_zalloc(hdl, sizeof (sp_monitor_t), FMD_SLEEP);
147 fmd_hdl_setspecific(hdl, smp);
148
149 if ((smp->sm_hdl = ipmi_open(&error, &msg, IPMI_TRANSPORT_BMC, NULL))
150 == NULL) {
151 /*
152 * If /dev/ipmi0 doesn't exist on the system, then unload the
153 * module without doing anything.
154 */
155 if (error != EIPMI_BMC_OPEN_FAILED)
156 fmd_hdl_abort(hdl, "failed to initialize IPMI "
157 "connection: %s\n", msg);
158 fmd_hdl_debug(hdl, "failed to load: no IPMI connection "
159 "present");
160 fmd_hdl_free(hdl, smp, sizeof (sp_monitor_t));
161 fmd_hdl_unregister(hdl);
162 return;
163 }
164
165 /*
166 * Attempt an initial uptime() call. If the IPMI command is
167 * unrecognized, then this is an unsupported platform and the module
168 * should be unloaded. Any other error is treated is transient failure.
169 */
170 if ((error = ipmi_sunoem_uptime(smp->sm_hdl, &smp->sm_seconds,
171 &smp->sm_generation)) != 0 &&
172 ipmi_errno(smp->sm_hdl) == EIPMI_INVALID_COMMAND) {
173 fmd_hdl_debug(hdl, "failed to load: uptime command "
174 "not supported");
175 ipmi_close(smp->sm_hdl);
176 fmd_hdl_free(hdl, smp, sizeof (sp_monitor_t));
177 fmd_hdl_unregister(hdl);
178 return;
179 }
180
181 smp->sm_interval = fmd_prop_get_int64(hdl, "interval");
182
183 if (error == 0)
184 fmd_hdl_debug(hdl, "successfully loaded, uptime = %u seconds "
185 "(generation %u)", smp->sm_seconds, smp->sm_generation);
186 else
187 fmd_hdl_debug(hdl, "successfully loaded, but uptime call "
188 "failed: %s", ipmi_errmsg(smp->sm_hdl));
189
190 /*
191 * Setup the recurring timer.
192 */
193 (void) fmd_timer_install(hdl, NULL, NULL, 0);
194 }
195
196 void
_fmd_fini(fmd_hdl_t * hdl)197 _fmd_fini(fmd_hdl_t *hdl)
198 {
199 sp_monitor_t *smp = fmd_hdl_getspecific(hdl);
200
201 if (smp) {
202 ipmi_close(smp->sm_hdl);
203 fmd_hdl_free(hdl, smp, sizeof (sp_monitor_t));
204 }
205 }
206