1*f6e214c7SGavin Maltby /*
2*f6e214c7SGavin Maltby * CDDL HEADER START
3*f6e214c7SGavin Maltby *
4*f6e214c7SGavin Maltby * The contents of this file are subject to the terms of the
5*f6e214c7SGavin Maltby * Common Development and Distribution License (the "License").
6*f6e214c7SGavin Maltby * You may not use this file except in compliance with the License.
7*f6e214c7SGavin Maltby *
8*f6e214c7SGavin Maltby * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9*f6e214c7SGavin Maltby * or http://www.opensolaris.org/os/licensing.
10*f6e214c7SGavin Maltby * See the License for the specific language governing permissions
11*f6e214c7SGavin Maltby * and limitations under the License.
12*f6e214c7SGavin Maltby *
13*f6e214c7SGavin Maltby * When distributing Covered Code, include this CDDL HEADER in each
14*f6e214c7SGavin Maltby * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15*f6e214c7SGavin Maltby * If applicable, add the following below this CDDL HEADER, with the
16*f6e214c7SGavin Maltby * fields enclosed by brackets "[]" replaced with your own identifying
17*f6e214c7SGavin Maltby * information: Portions Copyright [yyyy] [name of copyright owner]
18*f6e214c7SGavin Maltby *
19*f6e214c7SGavin Maltby * CDDL HEADER END
20*f6e214c7SGavin Maltby */
21*f6e214c7SGavin Maltby
22*f6e214c7SGavin Maltby /*
23*f6e214c7SGavin Maltby * Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
24*f6e214c7SGavin Maltby */
25*f6e214c7SGavin Maltby
26*f6e214c7SGavin Maltby /*
27*f6e214c7SGavin Maltby * SMF software-response subsidiary
28*f6e214c7SGavin Maltby */
29*f6e214c7SGavin Maltby
30*f6e214c7SGavin Maltby #include <strings.h>
31*f6e214c7SGavin Maltby #include <fm/libtopo.h>
32*f6e214c7SGavin Maltby #include <libscf.h>
33*f6e214c7SGavin Maltby #include <sys/fm/protocol.h>
34*f6e214c7SGavin Maltby #include <fm/fmd_fmri.h>
35*f6e214c7SGavin Maltby
36*f6e214c7SGavin Maltby #include "../../common/sw.h"
37*f6e214c7SGavin Maltby #include "smf.h"
38*f6e214c7SGavin Maltby
39*f6e214c7SGavin Maltby static struct {
40*f6e214c7SGavin Maltby fmd_stat_t swrp_smf_repairs;
41*f6e214c7SGavin Maltby fmd_stat_t swrp_smf_clears;
42*f6e214c7SGavin Maltby fmd_stat_t swrp_smf_closed;
43*f6e214c7SGavin Maltby fmd_stat_t swrp_smf_wrongclass;
44*f6e214c7SGavin Maltby fmd_stat_t swrp_smf_badlist;
45*f6e214c7SGavin Maltby fmd_stat_t swrp_smf_badresource;
46*f6e214c7SGavin Maltby fmd_stat_t swrp_smf_badclrevent;
47*f6e214c7SGavin Maltby fmd_stat_t swrp_smf_noloop;
48*f6e214c7SGavin Maltby fmd_stat_t swrp_smf_suppressed;
49*f6e214c7SGavin Maltby fmd_stat_t swrp_smf_cachefull;
50*f6e214c7SGavin Maltby } swrp_smf_stats = {
51*f6e214c7SGavin Maltby { "swrp_smf_repairs", FMD_TYPE_UINT64,
52*f6e214c7SGavin Maltby "repair events received for propogation to SMF" },
53*f6e214c7SGavin Maltby { "swrp_smf_clears", FMD_TYPE_UINT64,
54*f6e214c7SGavin Maltby "notifications from SMF of exiting maint state" },
55*f6e214c7SGavin Maltby { "swrp_smf_closed", FMD_TYPE_UINT64,
56*f6e214c7SGavin Maltby "cases closed" },
57*f6e214c7SGavin Maltby { "swrp_smf_wrongclass", FMD_TYPE_UINT64,
58*f6e214c7SGavin Maltby "unexpected event class received" },
59*f6e214c7SGavin Maltby { "swrp_smf_badlist", FMD_TYPE_UINT64,
60*f6e214c7SGavin Maltby "list event with invalid structure" },
61*f6e214c7SGavin Maltby { "swrp_smf_badresource", FMD_TYPE_UINT64,
62*f6e214c7SGavin Maltby "list.repaired with smf fault but bad svc fmri" },
63*f6e214c7SGavin Maltby { "swrp_smf_badclrevent", FMD_TYPE_UINT64,
64*f6e214c7SGavin Maltby "maint clear event from SMF malformed" },
65*f6e214c7SGavin Maltby { "swrp_smf_noloop", FMD_TYPE_UINT64,
66*f6e214c7SGavin Maltby "avoidance of smf->fmd->smf repairs propogations" },
67*f6e214c7SGavin Maltby { "swrp_smf_suppressed", FMD_TYPE_UINT64,
68*f6e214c7SGavin Maltby "not propogated to smf because no longer in maint" },
69*f6e214c7SGavin Maltby { "swrp_smf_cachefull", FMD_TYPE_UINT64,
70*f6e214c7SGavin Maltby "uuid cache full" },
71*f6e214c7SGavin Maltby };
72*f6e214c7SGavin Maltby
73*f6e214c7SGavin Maltby #define BUMPSTAT(stat) swrp_smf_stats.stat.fmds_value.ui64++
74*f6e214c7SGavin Maltby
75*f6e214c7SGavin Maltby #define CACHE_NENT_INC 16
76*f6e214c7SGavin Maltby #define CACHE_NENT_MAX 128
77*f6e214c7SGavin Maltby
78*f6e214c7SGavin Maltby struct smf_uuid_cache_ent {
79*f6e214c7SGavin Maltby char uuid[37];
80*f6e214c7SGavin Maltby char fmristr[90];
81*f6e214c7SGavin Maltby uint8_t mark;
82*f6e214c7SGavin Maltby };
83*f6e214c7SGavin Maltby
84*f6e214c7SGavin Maltby #define CACHE_VERSION 1
85*f6e214c7SGavin Maltby
86*f6e214c7SGavin Maltby struct smf_uuid_cache {
87*f6e214c7SGavin Maltby uint32_t version; /* Version */
88*f6e214c7SGavin Maltby uint32_t nentries; /* Real size of array below */
89*f6e214c7SGavin Maltby struct smf_uuid_cache_ent entry[1]; /* Cache entries */
90*f6e214c7SGavin Maltby };
91*f6e214c7SGavin Maltby
92*f6e214c7SGavin Maltby static struct smf_uuid_cache *uuid_cache;
93*f6e214c7SGavin Maltby
94*f6e214c7SGavin Maltby #define UUID_CACHE_BUFNAME "uuid_cache"
95*f6e214c7SGavin Maltby
96*f6e214c7SGavin Maltby static void
uuid_cache_grow(fmd_hdl_t * hdl)97*f6e214c7SGavin Maltby uuid_cache_grow(fmd_hdl_t *hdl)
98*f6e214c7SGavin Maltby {
99*f6e214c7SGavin Maltby struct smf_uuid_cache *newcache;
100*f6e214c7SGavin Maltby size_t newsz;
101*f6e214c7SGavin Maltby uint32_t n;
102*f6e214c7SGavin Maltby
103*f6e214c7SGavin Maltby n = (uuid_cache == NULL ? 0 : uuid_cache->nentries) + CACHE_NENT_INC;
104*f6e214c7SGavin Maltby newsz = sizeof (struct smf_uuid_cache) + (n - 1) *
105*f6e214c7SGavin Maltby sizeof (struct smf_uuid_cache_ent);
106*f6e214c7SGavin Maltby
107*f6e214c7SGavin Maltby newcache = fmd_hdl_zalloc(hdl, newsz, FMD_SLEEP);
108*f6e214c7SGavin Maltby newcache->version = CACHE_VERSION;
109*f6e214c7SGavin Maltby newcache->nentries = n;
110*f6e214c7SGavin Maltby
111*f6e214c7SGavin Maltby if (uuid_cache != NULL) {
112*f6e214c7SGavin Maltby uint32_t oldn = uuid_cache->nentries;
113*f6e214c7SGavin Maltby size_t oldsz = sizeof (struct smf_uuid_cache) +
114*f6e214c7SGavin Maltby (oldn - 1) * sizeof (struct smf_uuid_cache_ent);
115*f6e214c7SGavin Maltby
116*f6e214c7SGavin Maltby bcopy(&uuid_cache->entry[0], &newcache->entry[0], oldsz);
117*f6e214c7SGavin Maltby fmd_hdl_free(hdl, uuid_cache, oldsz);
118*f6e214c7SGavin Maltby fmd_buf_destroy(hdl, NULL, UUID_CACHE_BUFNAME);
119*f6e214c7SGavin Maltby }
120*f6e214c7SGavin Maltby
121*f6e214c7SGavin Maltby uuid_cache = newcache;
122*f6e214c7SGavin Maltby fmd_buf_create(hdl, NULL, UUID_CACHE_BUFNAME, newsz);
123*f6e214c7SGavin Maltby }
124*f6e214c7SGavin Maltby
125*f6e214c7SGavin Maltby static void
uuid_cache_persist(fmd_hdl_t * hdl)126*f6e214c7SGavin Maltby uuid_cache_persist(fmd_hdl_t *hdl)
127*f6e214c7SGavin Maltby {
128*f6e214c7SGavin Maltby size_t sz = sizeof (struct smf_uuid_cache) +
129*f6e214c7SGavin Maltby (uuid_cache->nentries - 1) * sizeof (struct smf_uuid_cache_ent);
130*f6e214c7SGavin Maltby
131*f6e214c7SGavin Maltby fmd_buf_write(hdl, NULL, UUID_CACHE_BUFNAME, uuid_cache, sz);
132*f6e214c7SGavin Maltby }
133*f6e214c7SGavin Maltby
134*f6e214c7SGavin Maltby /*
135*f6e214c7SGavin Maltby * Garbage-collect the uuid cache. Any cases that are already resolved
136*f6e214c7SGavin Maltby * we do not need an entry for. If a case is not resolved but the
137*f6e214c7SGavin Maltby * service involved in that case is no longer in maintenance state
138*f6e214c7SGavin Maltby * then we've lost sync somehow, so repair the asru (which will
139*f6e214c7SGavin Maltby * also resolve the case).
140*f6e214c7SGavin Maltby */
141*f6e214c7SGavin Maltby static void
uuid_cache_gc(fmd_hdl_t * hdl)142*f6e214c7SGavin Maltby uuid_cache_gc(fmd_hdl_t *hdl)
143*f6e214c7SGavin Maltby {
144*f6e214c7SGavin Maltby struct smf_uuid_cache_ent *entp;
145*f6e214c7SGavin Maltby topo_hdl_t *thp = NULL;
146*f6e214c7SGavin Maltby nvlist_t *svcfmri;
147*f6e214c7SGavin Maltby char *svcname;
148*f6e214c7SGavin Maltby int err, i;
149*f6e214c7SGavin Maltby
150*f6e214c7SGavin Maltby for (i = 0; i < uuid_cache->nentries; i++) {
151*f6e214c7SGavin Maltby entp = &uuid_cache->entry[i];
152*f6e214c7SGavin Maltby
153*f6e214c7SGavin Maltby if (entp->uuid[0] == '\0')
154*f6e214c7SGavin Maltby continue;
155*f6e214c7SGavin Maltby
156*f6e214c7SGavin Maltby if (fmd_case_uuisresolved(hdl, entp->uuid)) {
157*f6e214c7SGavin Maltby bzero(entp->uuid, sizeof (entp->uuid));
158*f6e214c7SGavin Maltby bzero(entp->fmristr, sizeof (entp->fmristr));
159*f6e214c7SGavin Maltby entp->mark = 0;
160*f6e214c7SGavin Maltby } else {
161*f6e214c7SGavin Maltby if (thp == NULL)
162*f6e214c7SGavin Maltby thp = fmd_hdl_topo_hold(hdl, TOPO_VERSION);
163*f6e214c7SGavin Maltby
164*f6e214c7SGavin Maltby if (topo_fmri_str2nvl(thp, entp->fmristr, &svcfmri,
165*f6e214c7SGavin Maltby &err) != 0) {
166*f6e214c7SGavin Maltby fmd_hdl_error(hdl, "str2nvl failed for %s\n",
167*f6e214c7SGavin Maltby entp->fmristr);
168*f6e214c7SGavin Maltby continue;
169*f6e214c7SGavin Maltby }
170*f6e214c7SGavin Maltby
171*f6e214c7SGavin Maltby if (fmd_nvl_fmri_service_state(hdl, svcfmri) !=
172*f6e214c7SGavin Maltby FMD_SERVICE_STATE_UNUSABLE) {
173*f6e214c7SGavin Maltby svcname = sw_smf_svcfmri2shortstr(hdl, svcfmri);
174*f6e214c7SGavin Maltby (void) fmd_repair_asru(hdl, entp->fmristr);
175*f6e214c7SGavin Maltby fmd_hdl_strfree(hdl, svcname);
176*f6e214c7SGavin Maltby }
177*f6e214c7SGavin Maltby
178*f6e214c7SGavin Maltby nvlist_free(svcfmri);
179*f6e214c7SGavin Maltby }
180*f6e214c7SGavin Maltby }
181*f6e214c7SGavin Maltby
182*f6e214c7SGavin Maltby if (thp)
183*f6e214c7SGavin Maltby fmd_hdl_topo_rele(hdl, thp);
184*f6e214c7SGavin Maltby
185*f6e214c7SGavin Maltby uuid_cache_persist(hdl);
186*f6e214c7SGavin Maltby }
187*f6e214c7SGavin Maltby
188*f6e214c7SGavin Maltby static void
uuid_cache_restore(fmd_hdl_t * hdl)189*f6e214c7SGavin Maltby uuid_cache_restore(fmd_hdl_t *hdl)
190*f6e214c7SGavin Maltby {
191*f6e214c7SGavin Maltby size_t sz = fmd_buf_size(hdl, NULL, UUID_CACHE_BUFNAME);
192*f6e214c7SGavin Maltby
193*f6e214c7SGavin Maltby if (sz == 0)
194*f6e214c7SGavin Maltby return;
195*f6e214c7SGavin Maltby
196*f6e214c7SGavin Maltby uuid_cache = fmd_hdl_alloc(hdl, sz, FMD_SLEEP);
197*f6e214c7SGavin Maltby fmd_buf_read(hdl, NULL, UUID_CACHE_BUFNAME, uuid_cache, sz);
198*f6e214c7SGavin Maltby
199*f6e214c7SGavin Maltby /*
200*f6e214c7SGavin Maltby * Garbage collect now, not just for tidiness but also to help
201*f6e214c7SGavin Maltby * fmd and smf state stay in sync at module startup.
202*f6e214c7SGavin Maltby */
203*f6e214c7SGavin Maltby uuid_cache_gc(hdl);
204*f6e214c7SGavin Maltby }
205*f6e214c7SGavin Maltby
206*f6e214c7SGavin Maltby /*
207*f6e214c7SGavin Maltby * Add the UUID of an SMF maintenance defect case to our cache and
208*f6e214c7SGavin Maltby * record the associated full svc FMRI string for the case.
209*f6e214c7SGavin Maltby */
210*f6e214c7SGavin Maltby static void
swrp_smf_cache_add(fmd_hdl_t * hdl,char * uuid,char * fmristr)211*f6e214c7SGavin Maltby swrp_smf_cache_add(fmd_hdl_t *hdl, char *uuid, char *fmristr)
212*f6e214c7SGavin Maltby {
213*f6e214c7SGavin Maltby struct smf_uuid_cache_ent *entp = NULL;
214*f6e214c7SGavin Maltby int gced = 0;
215*f6e214c7SGavin Maltby int i;
216*f6e214c7SGavin Maltby
217*f6e214c7SGavin Maltby if (uuid_cache == NULL)
218*f6e214c7SGavin Maltby uuid_cache_grow(hdl);
219*f6e214c7SGavin Maltby
220*f6e214c7SGavin Maltby /*
221*f6e214c7SGavin Maltby * If we somehow already have an entry for this uuid then
222*f6e214c7SGavin Maltby * return leaving it undisturbed.
223*f6e214c7SGavin Maltby */
224*f6e214c7SGavin Maltby for (i = 0; i < uuid_cache->nentries; i++) {
225*f6e214c7SGavin Maltby if (strcmp(uuid, uuid_cache->entry[i].uuid) == 0)
226*f6e214c7SGavin Maltby return;
227*f6e214c7SGavin Maltby }
228*f6e214c7SGavin Maltby
229*f6e214c7SGavin Maltby scan:
230*f6e214c7SGavin Maltby for (i = 0; i < uuid_cache->nentries; i++) {
231*f6e214c7SGavin Maltby if (uuid_cache->entry[i].uuid[0] == '\0') {
232*f6e214c7SGavin Maltby entp = &uuid_cache->entry[i];
233*f6e214c7SGavin Maltby break;
234*f6e214c7SGavin Maltby }
235*f6e214c7SGavin Maltby }
236*f6e214c7SGavin Maltby
237*f6e214c7SGavin Maltby if (entp == NULL) {
238*f6e214c7SGavin Maltby uint32_t oldn = uuid_cache->nentries;
239*f6e214c7SGavin Maltby
240*f6e214c7SGavin Maltby /*
241*f6e214c7SGavin Maltby * Before growing the cache we try again after first
242*f6e214c7SGavin Maltby * garbage-collecting the existing cache for any cases
243*f6e214c7SGavin Maltby * that are confirmed as resolved.
244*f6e214c7SGavin Maltby */
245*f6e214c7SGavin Maltby if (!gced) {
246*f6e214c7SGavin Maltby uuid_cache_gc(hdl);
247*f6e214c7SGavin Maltby gced = 1;
248*f6e214c7SGavin Maltby goto scan;
249*f6e214c7SGavin Maltby }
250*f6e214c7SGavin Maltby
251*f6e214c7SGavin Maltby if (oldn < CACHE_NENT_MAX) {
252*f6e214c7SGavin Maltby uuid_cache_grow(hdl);
253*f6e214c7SGavin Maltby entp = &uuid_cache->entry[oldn];
254*f6e214c7SGavin Maltby } else {
255*f6e214c7SGavin Maltby BUMPSTAT(swrp_smf_cachefull);
256*f6e214c7SGavin Maltby return;
257*f6e214c7SGavin Maltby }
258*f6e214c7SGavin Maltby }
259*f6e214c7SGavin Maltby
260*f6e214c7SGavin Maltby (void) strncpy(entp->uuid, uuid, sizeof (entp->uuid));
261*f6e214c7SGavin Maltby (void) strncpy(entp->fmristr, fmristr, sizeof (entp->fmristr));
262*f6e214c7SGavin Maltby uuid_cache_persist(hdl);
263*f6e214c7SGavin Maltby }
264*f6e214c7SGavin Maltby
265*f6e214c7SGavin Maltby /*
266*f6e214c7SGavin Maltby * Mark cache entry/entries as resolved - if they match in either uuid
267*f6e214c7SGavin Maltby * (if not NULL) or fmristr (if not NULL) mark as resolved. Return 1 iff
268*f6e214c7SGavin Maltby * an entry that matched on uuid was already marked, otherwise (entry
269*f6e214c7SGavin Maltby * matched on either, matched on uuid but not marked, not found).
270*f6e214c7SGavin Maltby */
271*f6e214c7SGavin Maltby static int
swrp_smf_cache_mark(fmd_hdl_t * hdl,char * uuid,char * fmristr)272*f6e214c7SGavin Maltby swrp_smf_cache_mark(fmd_hdl_t *hdl, char *uuid, char *fmristr)
273*f6e214c7SGavin Maltby {
274*f6e214c7SGavin Maltby int dirty = 0;
275*f6e214c7SGavin Maltby int rv = 0;
276*f6e214c7SGavin Maltby int i;
277*f6e214c7SGavin Maltby
278*f6e214c7SGavin Maltby if (uuid_cache == NULL)
279*f6e214c7SGavin Maltby return (0);
280*f6e214c7SGavin Maltby
281*f6e214c7SGavin Maltby for (i = 0; i < uuid_cache->nentries; i++) {
282*f6e214c7SGavin Maltby struct smf_uuid_cache_ent *entp = &uuid_cache->entry[i];
283*f6e214c7SGavin Maltby
284*f6e214c7SGavin Maltby if (entp->uuid[0] == '\0')
285*f6e214c7SGavin Maltby continue;
286*f6e214c7SGavin Maltby
287*f6e214c7SGavin Maltby if (uuid && strcmp(uuid, entp->uuid) == 0) {
288*f6e214c7SGavin Maltby if (entp->mark)
289*f6e214c7SGavin Maltby rv = 1;
290*f6e214c7SGavin Maltby entp->mark = 1;
291*f6e214c7SGavin Maltby dirty++;
292*f6e214c7SGavin Maltby } else if (fmristr && strcmp(fmristr, entp->fmristr) == 0) {
293*f6e214c7SGavin Maltby entp->mark = 1;
294*f6e214c7SGavin Maltby dirty++;
295*f6e214c7SGavin Maltby }
296*f6e214c7SGavin Maltby }
297*f6e214c7SGavin Maltby
298*f6e214c7SGavin Maltby if (dirty)
299*f6e214c7SGavin Maltby uuid_cache_persist(hdl);
300*f6e214c7SGavin Maltby
301*f6e214c7SGavin Maltby return (rv);
302*f6e214c7SGavin Maltby }
303*f6e214c7SGavin Maltby
304*f6e214c7SGavin Maltby /*
305*f6e214c7SGavin Maltby * We will receive list events for cases we are not interested in. Test
306*f6e214c7SGavin Maltby * that this list has exactly one suspect and that it matches the maintenance
307*f6e214c7SGavin Maltby * defect. Return the defect to the caller in the second argument,
308*f6e214c7SGavin Maltby * and the defect resource element in the third arg.
309*f6e214c7SGavin Maltby */
310*f6e214c7SGavin Maltby static int
suspect_is_maint_defect(fmd_hdl_t * hdl,nvlist_t * nvl,nvlist_t ** defectnvl,nvlist_t ** rsrcnvl)311*f6e214c7SGavin Maltby suspect_is_maint_defect(fmd_hdl_t *hdl, nvlist_t *nvl,
312*f6e214c7SGavin Maltby nvlist_t **defectnvl, nvlist_t **rsrcnvl)
313*f6e214c7SGavin Maltby {
314*f6e214c7SGavin Maltby nvlist_t **faults;
315*f6e214c7SGavin Maltby uint_t nfaults;
316*f6e214c7SGavin Maltby
317*f6e214c7SGavin Maltby if (nvlist_lookup_nvlist_array(nvl, FM_SUSPECT_FAULT_LIST,
318*f6e214c7SGavin Maltby &faults, &nfaults) != 0) {
319*f6e214c7SGavin Maltby BUMPSTAT(swrp_smf_badlist);
320*f6e214c7SGavin Maltby return (0);
321*f6e214c7SGavin Maltby }
322*f6e214c7SGavin Maltby
323*f6e214c7SGavin Maltby if (nfaults != 1 ||
324*f6e214c7SGavin Maltby !fmd_nvl_class_match(hdl, faults[0], SW_SMF_MAINT_DEFECT))
325*f6e214c7SGavin Maltby return (0);
326*f6e214c7SGavin Maltby
327*f6e214c7SGavin Maltby if (nvlist_lookup_nvlist(faults[0], FM_FAULT_RESOURCE, rsrcnvl) != 0) {
328*f6e214c7SGavin Maltby BUMPSTAT(swrp_smf_badlist);
329*f6e214c7SGavin Maltby return (0);
330*f6e214c7SGavin Maltby }
331*f6e214c7SGavin Maltby
332*f6e214c7SGavin Maltby *defectnvl = faults[0];
333*f6e214c7SGavin Maltby
334*f6e214c7SGavin Maltby return (1);
335*f6e214c7SGavin Maltby }
336*f6e214c7SGavin Maltby
337*f6e214c7SGavin Maltby /*
338*f6e214c7SGavin Maltby * Received newly-diagnosed list.suspect events that are for the
339*f6e214c7SGavin Maltby * maintenane defect we diagnose. Close the case (the resource was already
340*f6e214c7SGavin Maltby * isolated by SMF) after cachng the case UUID.
341*f6e214c7SGavin Maltby */
342*f6e214c7SGavin Maltby /*ARGSUSED*/
343*f6e214c7SGavin Maltby static void
swrp_smf_cacheuuid(fmd_hdl_t * hdl,fmd_event_t * ep,nvlist_t * nvl,const char * class,void * arg)344*f6e214c7SGavin Maltby swrp_smf_cacheuuid(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl,
345*f6e214c7SGavin Maltby const char *class, void *arg)
346*f6e214c7SGavin Maltby {
347*f6e214c7SGavin Maltby nvlist_t *defect, *rsrc;
348*f6e214c7SGavin Maltby char *fmristr, *uuid;
349*f6e214c7SGavin Maltby
350*f6e214c7SGavin Maltby if (nvlist_lookup_string(nvl, FM_SUSPECT_UUID, &uuid) != 0) {
351*f6e214c7SGavin Maltby BUMPSTAT(swrp_smf_badlist);
352*f6e214c7SGavin Maltby return;
353*f6e214c7SGavin Maltby }
354*f6e214c7SGavin Maltby
355*f6e214c7SGavin Maltby if (!suspect_is_maint_defect(hdl, nvl, &defect, &rsrc))
356*f6e214c7SGavin Maltby return;
357*f6e214c7SGavin Maltby
358*f6e214c7SGavin Maltby if ((fmristr = sw_smf_svcfmri2str(hdl, rsrc)) == NULL) {
359*f6e214c7SGavin Maltby BUMPSTAT(swrp_smf_badlist);
360*f6e214c7SGavin Maltby return;
361*f6e214c7SGavin Maltby }
362*f6e214c7SGavin Maltby
363*f6e214c7SGavin Maltby swrp_smf_cache_add(hdl, uuid, fmristr);
364*f6e214c7SGavin Maltby fmd_hdl_strfree(hdl, fmristr);
365*f6e214c7SGavin Maltby
366*f6e214c7SGavin Maltby if (!fmd_case_uuclosed(hdl, uuid)) {
367*f6e214c7SGavin Maltby fmd_case_uuclose(hdl, uuid);
368*f6e214c7SGavin Maltby BUMPSTAT(swrp_smf_closed);
369*f6e214c7SGavin Maltby }
370*f6e214c7SGavin Maltby }
371*f6e214c7SGavin Maltby
372*f6e214c7SGavin Maltby /*ARGSUSED*/
373*f6e214c7SGavin Maltby static void
swrp_smf2fmd(fmd_hdl_t * hdl,fmd_event_t * ep,nvlist_t * nvl,const char * class,void * arg)374*f6e214c7SGavin Maltby swrp_smf2fmd(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl,
375*f6e214c7SGavin Maltby const char *class, void *arg)
376*f6e214c7SGavin Maltby {
377*f6e214c7SGavin Maltby nvlist_t *attr, *fmri;
378*f6e214c7SGavin Maltby char *fromstate;
379*f6e214c7SGavin Maltby char *fmristr;
380*f6e214c7SGavin Maltby
381*f6e214c7SGavin Maltby if (!fmd_nvl_class_match(hdl, nvl, TRANCLASS("*"))) {
382*f6e214c7SGavin Maltby BUMPSTAT(swrp_smf_wrongclass);
383*f6e214c7SGavin Maltby return;
384*f6e214c7SGavin Maltby }
385*f6e214c7SGavin Maltby
386*f6e214c7SGavin Maltby if (nvlist_lookup_nvlist(nvl, FM_IREPORT_ATTRIBUTES, &attr) != 0 ||
387*f6e214c7SGavin Maltby nvlist_lookup_string(attr, "from-state", &fromstate) != 0) {
388*f6e214c7SGavin Maltby BUMPSTAT(swrp_smf_badclrevent);
389*f6e214c7SGavin Maltby return;
390*f6e214c7SGavin Maltby }
391*f6e214c7SGavin Maltby
392*f6e214c7SGavin Maltby /*
393*f6e214c7SGavin Maltby * Filter those not describing a transition out of maintenance.
394*f6e214c7SGavin Maltby */
395*f6e214c7SGavin Maltby if (strcmp(fromstate, "maintenance") != 0)
396*f6e214c7SGavin Maltby return;
397*f6e214c7SGavin Maltby
398*f6e214c7SGavin Maltby if (nvlist_lookup_nvlist(attr, "svc", &fmri) != 0) {
399*f6e214c7SGavin Maltby BUMPSTAT(swrp_smf_badclrevent);
400*f6e214c7SGavin Maltby return;
401*f6e214c7SGavin Maltby }
402*f6e214c7SGavin Maltby
403*f6e214c7SGavin Maltby if ((fmristr = sw_smf_svcfmri2str(hdl, fmri)) == NULL) {
404*f6e214c7SGavin Maltby BUMPSTAT(swrp_smf_badclrevent);
405*f6e214c7SGavin Maltby return;
406*f6e214c7SGavin Maltby }
407*f6e214c7SGavin Maltby
408*f6e214c7SGavin Maltby /*
409*f6e214c7SGavin Maltby * Mark any UUID for a case against this service as resolved
410*f6e214c7SGavin Maltby * in our cache. When we fmd_repair_asru below fmd will emit
411*f6e214c7SGavin Maltby * a list.repaired as a result, and our handling of that event
412*f6e214c7SGavin Maltby * must not propogate the repair towards SMF (since the repair
413*f6e214c7SGavin Maltby * was initiated via SMF itself and not via fmadm).
414*f6e214c7SGavin Maltby */
415*f6e214c7SGavin Maltby (void) swrp_smf_cache_mark(hdl, NULL, fmristr);
416*f6e214c7SGavin Maltby
417*f6e214c7SGavin Maltby (void) fmd_repair_asru(hdl, fmristr);
418*f6e214c7SGavin Maltby fmd_hdl_strfree(hdl, fmristr);
419*f6e214c7SGavin Maltby BUMPSTAT(swrp_smf_clears);
420*f6e214c7SGavin Maltby }
421*f6e214c7SGavin Maltby
422*f6e214c7SGavin Maltby /*ARGSUSED*/
423*f6e214c7SGavin Maltby static void
swrp_fmd2smf(fmd_hdl_t * hdl,fmd_event_t * ep,nvlist_t * nvl,const char * class,void * arg)424*f6e214c7SGavin Maltby swrp_fmd2smf(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl,
425*f6e214c7SGavin Maltby const char *class, void *arg)
426*f6e214c7SGavin Maltby {
427*f6e214c7SGavin Maltby char *fmristr, *shrtfmristr;
428*f6e214c7SGavin Maltby nvlist_t *defect, *rsrc;
429*f6e214c7SGavin Maltby char *uuid;
430*f6e214c7SGavin Maltby int already;
431*f6e214c7SGavin Maltby
432*f6e214c7SGavin Maltby if (strcmp(class, FM_LIST_REPAIRED_CLASS) != 0) {
433*f6e214c7SGavin Maltby BUMPSTAT(swrp_smf_wrongclass);
434*f6e214c7SGavin Maltby return;
435*f6e214c7SGavin Maltby }
436*f6e214c7SGavin Maltby
437*f6e214c7SGavin Maltby if (nvlist_lookup_string(nvl, FM_SUSPECT_UUID, &uuid) != 0) {
438*f6e214c7SGavin Maltby BUMPSTAT(swrp_smf_badlist);
439*f6e214c7SGavin Maltby return;
440*f6e214c7SGavin Maltby }
441*f6e214c7SGavin Maltby
442*f6e214c7SGavin Maltby if (!suspect_is_maint_defect(hdl, nvl, &defect, &rsrc))
443*f6e214c7SGavin Maltby return;
444*f6e214c7SGavin Maltby
445*f6e214c7SGavin Maltby if ((fmristr = sw_smf_svcfmri2str(hdl, rsrc)) == NULL) {
446*f6e214c7SGavin Maltby BUMPSTAT(swrp_smf_badresource);
447*f6e214c7SGavin Maltby return;
448*f6e214c7SGavin Maltby }
449*f6e214c7SGavin Maltby
450*f6e214c7SGavin Maltby already = swrp_smf_cache_mark(hdl, uuid, fmristr);
451*f6e214c7SGavin Maltby fmd_hdl_strfree(hdl, fmristr);
452*f6e214c7SGavin Maltby
453*f6e214c7SGavin Maltby /*
454*f6e214c7SGavin Maltby * If the cache already had a marked entry for this UUID then
455*f6e214c7SGavin Maltby * this is a list.repaired arising from a SMF-initiated maintenance
456*f6e214c7SGavin Maltby * clear (propogated with fmd_repair_asru above which then results
457*f6e214c7SGavin Maltby * in a list.repaired) and so we should not propogate the repair
458*f6e214c7SGavin Maltby * back towards SMF. But do still force the case to RESOLVED state in
459*f6e214c7SGavin Maltby * case fmd is unable to confirm the service no longer in maintenance
460*f6e214c7SGavin Maltby * state (it may have failed again) so that a new case can be opened.
461*f6e214c7SGavin Maltby */
462*f6e214c7SGavin Maltby fmd_case_uuresolved(hdl, uuid);
463*f6e214c7SGavin Maltby if (already) {
464*f6e214c7SGavin Maltby BUMPSTAT(swrp_smf_noloop);
465*f6e214c7SGavin Maltby return;
466*f6e214c7SGavin Maltby }
467*f6e214c7SGavin Maltby
468*f6e214c7SGavin Maltby /*
469*f6e214c7SGavin Maltby * Only propogate to SMF if we can see that service still
470*f6e214c7SGavin Maltby * in maintenance state. We're not synchronized with SMF
471*f6e214c7SGavin Maltby * and this state could change at any time, but if we can
472*f6e214c7SGavin Maltby * see it's not in maintenance state then things are obviously
473*f6e214c7SGavin Maltby * moving (e.g., external svcadm active) so we don't poke
474*f6e214c7SGavin Maltby * at SMF otherwise we confuse things or duplicate operations.
475*f6e214c7SGavin Maltby */
476*f6e214c7SGavin Maltby
477*f6e214c7SGavin Maltby if (fmd_nvl_fmri_service_state(hdl, rsrc) ==
478*f6e214c7SGavin Maltby FMD_SERVICE_STATE_UNUSABLE) {
479*f6e214c7SGavin Maltby shrtfmristr = sw_smf_svcfmri2shortstr(hdl, rsrc);
480*f6e214c7SGavin Maltby
481*f6e214c7SGavin Maltby if (shrtfmristr != NULL) {
482*f6e214c7SGavin Maltby (void) smf_restore_instance(shrtfmristr);
483*f6e214c7SGavin Maltby fmd_hdl_strfree(hdl, shrtfmristr);
484*f6e214c7SGavin Maltby BUMPSTAT(swrp_smf_repairs);
485*f6e214c7SGavin Maltby } else {
486*f6e214c7SGavin Maltby BUMPSTAT(swrp_smf_badresource);
487*f6e214c7SGavin Maltby }
488*f6e214c7SGavin Maltby } else {
489*f6e214c7SGavin Maltby BUMPSTAT(swrp_smf_suppressed);
490*f6e214c7SGavin Maltby }
491*f6e214c7SGavin Maltby }
492*f6e214c7SGavin Maltby
493*f6e214c7SGavin Maltby const struct sw_disp swrp_smf_disp[] = {
494*f6e214c7SGavin Maltby { TRANCLASS("*"), swrp_smf2fmd, NULL },
495*f6e214c7SGavin Maltby { FM_LIST_SUSPECT_CLASS, swrp_smf_cacheuuid, NULL },
496*f6e214c7SGavin Maltby { FM_LIST_REPAIRED_CLASS, swrp_fmd2smf, NULL },
497*f6e214c7SGavin Maltby { NULL, NULL, NULL }
498*f6e214c7SGavin Maltby };
499*f6e214c7SGavin Maltby
500*f6e214c7SGavin Maltby /*ARGSUSED*/
501*f6e214c7SGavin Maltby int
swrp_smf_init(fmd_hdl_t * hdl,id_t id,const struct sw_disp ** dpp,int * nelemp)502*f6e214c7SGavin Maltby swrp_smf_init(fmd_hdl_t *hdl, id_t id, const struct sw_disp **dpp, int *nelemp)
503*f6e214c7SGavin Maltby {
504*f6e214c7SGavin Maltby (void) fmd_stat_create(hdl, FMD_STAT_NOALLOC, sizeof (swrp_smf_stats) /
505*f6e214c7SGavin Maltby sizeof (fmd_stat_t), (fmd_stat_t *)&swrp_smf_stats);
506*f6e214c7SGavin Maltby
507*f6e214c7SGavin Maltby uuid_cache_restore(hdl);
508*f6e214c7SGavin Maltby
509*f6e214c7SGavin Maltby /*
510*f6e214c7SGavin Maltby * We need to subscribe to all SMF transition class events because
511*f6e214c7SGavin Maltby * we need to look inside the payload to see which events indicate
512*f6e214c7SGavin Maltby * a transition out of maintenance state.
513*f6e214c7SGavin Maltby */
514*f6e214c7SGavin Maltby fmd_hdl_subscribe(hdl, TRANCLASS("*"));
515*f6e214c7SGavin Maltby
516*f6e214c7SGavin Maltby /*
517*f6e214c7SGavin Maltby * Subscribe to the defect class diagnosed for maintenance events.
518*f6e214c7SGavin Maltby * The module will then receive list.suspect events including
519*f6e214c7SGavin Maltby * these defects, and in our dispatch table above we list routing
520*f6e214c7SGavin Maltby * for list.suspect.
521*f6e214c7SGavin Maltby */
522*f6e214c7SGavin Maltby fmd_hdl_subscribe(hdl, SW_SMF_MAINT_DEFECT);
523*f6e214c7SGavin Maltby
524*f6e214c7SGavin Maltby *dpp = &swrp_smf_disp[0];
525*f6e214c7SGavin Maltby *nelemp = sizeof (swrp_smf_disp) / sizeof (swrp_smf_disp[0]);
526*f6e214c7SGavin Maltby return (SW_SUB_INIT_SUCCESS);
527*f6e214c7SGavin Maltby }
528*f6e214c7SGavin Maltby
529*f6e214c7SGavin Maltby /*ARGSUSED*/
530*f6e214c7SGavin Maltby void
swrp_smf_fini(fmd_hdl_t * hdl)531*f6e214c7SGavin Maltby swrp_smf_fini(fmd_hdl_t *hdl)
532*f6e214c7SGavin Maltby {
533*f6e214c7SGavin Maltby }
534*f6e214c7SGavin Maltby
535*f6e214c7SGavin Maltby const struct sw_subinfo smf_response_info = {
536*f6e214c7SGavin Maltby "smf repair", /* swsub_name */
537*f6e214c7SGavin Maltby SW_CASE_NONE, /* swsub_casetype */
538*f6e214c7SGavin Maltby swrp_smf_init, /* swsub_init */
539*f6e214c7SGavin Maltby swrp_smf_fini, /* swsub_fini */
540*f6e214c7SGavin Maltby NULL, /* swsub_timeout */
541*f6e214c7SGavin Maltby NULL, /* swsub_case_close */
542*f6e214c7SGavin Maltby NULL, /* swsub_case_vrfy */
543*f6e214c7SGavin Maltby };
544