1 /*
2 * Copyright (c) 2007 The Regents of the University of California.
3 * Copyright (c) 2007-2009 Voltaire, Inc. All rights reserved.
4 * Copyright (c) 2009,2010 HNR Consulting. All rights reserved.
5 *
6 * This software is available to you under a choice of one of two
7 * licenses. You may choose to be licensed under the terms of the GNU
8 * General Public License (GPL) Version 2, available from the file
9 * COPYING in the main directory of this source tree, or the
10 * OpenIB.org BSD license below:
11 *
12 * Redistribution and use in source and binary forms, with or
13 * without modification, are permitted provided that the following
14 * conditions are met:
15 *
16 * - Redistributions of source code must retain the above
17 * copyright notice, this list of conditions and the following
18 * disclaimer.
19 *
20 * - Redistributions in binary form must reproduce the above
21 * copyright notice, this list of conditions and the following
22 * disclaimer in the documentation and/or other materials
23 * provided with the distribution.
24 *
25 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
26 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
27 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
28 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
29 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
30 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
31 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
32 * SOFTWARE.
33 *
34 */
35
36 #ifndef _OSM_PERFMGR_H_
37 #define _OSM_PERFMGR_H_
38
39 #if HAVE_CONFIG_H
40 # include <config.h>
41 #endif /* HAVE_CONFIG_H */
42
43 #ifdef ENABLE_OSM_PERF_MGR
44
45 #include <iba/ib_types.h>
46 #include <complib/cl_passivelock.h>
47 #include <complib/cl_event.h>
48 #include <complib/cl_timer.h>
49 #include <opensm/osm_subnet.h>
50 #include <opensm/osm_log.h>
51 #include <opensm/osm_perfmgr_db.h>
52 #include <opensm/osm_sm.h>
53 #include <opensm/osm_base.h>
54 #include <opensm/osm_event_plugin.h>
55
56 #ifdef __cplusplus
57 extern "C" {
58 #endif /* __cplusplus */
59
60 /****h* OpenSM/PerfMgr
61 * NAME
62 * PerfMgr
63 *
64 * DESCRIPTION
65 * Performance manager thread which takes care of polling the fabric for
66 * Port counters values.
67 *
68 * The PerfMgr object is thread safe.
69 *
70 * AUTHOR
71 * Ira Weiny, LLNL
72 *
73 *********/
74
75 #define OSM_PERFMGR_DEFAULT_SWEEP_TIME_S 180
76 #define OSM_PERFMGR_DEFAULT_DUMP_FILE "opensm_port_counters.log"
77 #define OSM_PERFMGR_DEFAULT_MAX_OUTSTANDING_QUERIES 500
78 #define OSM_PERFMGR_DEFAULT_XMIT_WAIT_THRESHOLD 0x0000FFFF
79
80 /****s* OpenSM: PerfMgr/osm_perfmgr_state_t */
81 typedef enum {
82 PERFMGR_STATE_DISABLE,
83 PERFMGR_STATE_ENABLED,
84 PERFMGR_STATE_NO_DB
85 } osm_perfmgr_state_t;
86
87 /****s* OpenSM: PerfMgr/osm_perfmgr_sweep_state_t */
88 typedef enum {
89 PERFMGR_SWEEP_SLEEP,
90 PERFMGR_SWEEP_ACTIVE,
91 PERFMGR_SWEEP_SUSPENDED,
92 PERFMGR_SWEEP_POST_PROCESSING
93 } osm_perfmgr_sweep_state_t;
94
95 typedef struct monitored_port {
96 uint16_t pkey_ix;
97 ib_net16_t orig_lid;
98 boolean_t redirection;
99 boolean_t valid;
100 /* Redirection fields from ClassPortInfo */
101 ib_gid_t gid;
102 ib_net16_t lid;
103 ib_net16_t pkey;
104 ib_net32_t qp;
105 /* ClassPortInfo fields */
106 boolean_t cpi_valid;
107 ib_net16_t cap_mask;
108 /* Remote end connected to */
109 boolean_t remote_valid;
110 uint64_t remote_guid;
111 char *remote_name;
112 uint8_t remote_port;
113 } monitored_port_t;
114
115 /* Node to store information about nodes being monitored */
116 typedef struct monitored_node {
117 cl_map_item_t map_item;
118 struct monitored_node *next;
119 uint64_t guid;
120 uint8_t node_type;
121 boolean_t esp0;
122 char *name;
123 uint32_t num_ports;
124 monitored_port_t port[1];
125 } monitored_node_t;
126
127 struct osm_opensm;
128
129 /****s* OpenSM: PerfMgr/osm_perfmgr_t
130 * This object should be treated as opaque and should
131 * be manipulated only through the provided functions.
132 */
133 typedef struct osm_perfmgr {
134 cl_timer_t sweep_timer;
135 struct osm_opensm *osm;
136 osm_subn_t *subn;
137 osm_sm_t *sm;
138 osm_log_t *log;
139 osm_mad_pool_t *mad_pool;
140 atomic32_t trans_id;
141 osm_vendor_t *vendor;
142 osm_bind_handle_t bind_handle;
143 cl_disp_reg_handle_t pc_disp_h;
144 osm_perfmgr_state_t state;
145 osm_perfmgr_sweep_state_t sweep_state;
146 cl_spinlock_t lock;
147 uint16_t sweep_time_s;
148 perfmgr_db_t *db;
149 atomic32_t outstanding_queries; /* this along with sig_query */
150 cl_event_t sig_query; /* will throttle our queries */
151 uint32_t max_outstanding_queries;
152 boolean_t ignore_cas;
153 cl_qmap_t monitored_map; /* map the nodes being tracked */
154 monitored_node_t *remove_list;
155 ib_net64_t port_guid;
156 int16_t local_port;
157 int rm_nodes;
158 boolean_t query_cpi;
159 boolean_t xmit_wait_log;
160 uint32_t xmit_wait_threshold;
161 } osm_perfmgr_t;
162 /*
163 * FIELDS
164 * subn
165 * Subnet object for this subnet.
166 *
167 * log
168 * Pointer to the log object.
169 *
170 * mad_pool
171 * Pointer to the MAD pool.
172 *
173 * mad_ctrl
174 * Mad Controller
175 *********/
176
177 /****f* OpenSM: Creation Functions */
178 void osm_perfmgr_shutdown(osm_perfmgr_t * p_perfmgr);
179 void osm_perfmgr_destroy(osm_perfmgr_t * p_perfmgr);
180
181 /****f* OpenSM: Inline accessor functions */
osm_perfmgr_set_state(osm_perfmgr_t * p_perfmgr,osm_perfmgr_state_t state)182 inline static void osm_perfmgr_set_state(osm_perfmgr_t * p_perfmgr,
183 osm_perfmgr_state_t state)
184 {
185 p_perfmgr->state = state;
186 if (state == PERFMGR_STATE_ENABLED) {
187 cl_timer_start(&p_perfmgr->sweep_timer, p_perfmgr->sweep_time_s * 1000);
188 } else {
189 cl_timer_stop(&p_perfmgr->sweep_timer);
190 }
191 }
192
osm_perfmgr_get_state(osm_perfmgr_t * perfmgr)193 inline static osm_perfmgr_state_t osm_perfmgr_get_state(osm_perfmgr_t * perfmgr)
194 {
195 return perfmgr->state;
196 }
197
osm_perfmgr_set_rm_nodes(osm_perfmgr_t * perfmgr,int rm_nodes)198 inline static void osm_perfmgr_set_rm_nodes(osm_perfmgr_t *perfmgr,
199 int rm_nodes)
200 {
201 perfmgr->rm_nodes = rm_nodes;
202 }
203
osm_perfmgr_get_rm_nodes(osm_perfmgr_t * perfmgr)204 inline static int osm_perfmgr_get_rm_nodes(osm_perfmgr_t *perfmgr)
205 {
206 return perfmgr->rm_nodes;
207 }
208
osm_perfmgr_set_query_cpi(osm_perfmgr_t * perfmgr,int query_cpi)209 inline static void osm_perfmgr_set_query_cpi(osm_perfmgr_t *perfmgr,
210 int query_cpi)
211 {
212 perfmgr->query_cpi = query_cpi;
213 }
214
osm_perfmgr_get_query_cpi(osm_perfmgr_t * perfmgr)215 inline static int osm_perfmgr_get_query_cpi(osm_perfmgr_t *perfmgr)
216 {
217 return perfmgr->query_cpi;
218 }
219
osm_perfmgr_get_state_str(osm_perfmgr_t * p_perfmgr)220 inline static const char *osm_perfmgr_get_state_str(osm_perfmgr_t * p_perfmgr)
221 {
222 switch (p_perfmgr->state) {
223 case PERFMGR_STATE_DISABLE:
224 return "Disabled";
225 break;
226 case PERFMGR_STATE_ENABLED:
227 return "Enabled";
228 break;
229 case PERFMGR_STATE_NO_DB:
230 return "No Database";
231 break;
232 }
233 return "UNKNOWN";
234 }
235
osm_perfmgr_get_sweep_state_str(osm_perfmgr_t * perfmgr)236 inline static const char *osm_perfmgr_get_sweep_state_str(osm_perfmgr_t * perfmgr)
237 {
238 switch (perfmgr->sweep_state) {
239 case PERFMGR_SWEEP_SLEEP:
240 return "Sleeping";
241 break;
242 case PERFMGR_SWEEP_ACTIVE:
243 return "Active";
244 break;
245 case PERFMGR_SWEEP_SUSPENDED:
246 return "Suspended";
247 break;
248 case PERFMGR_SWEEP_POST_PROCESSING:
249 return "PostProcessing";
250 break;
251 }
252 return "UNKNOWN";
253 }
254
osm_perfmgr_set_sweep_time_s(osm_perfmgr_t * p_perfmgr,uint16_t time_s)255 inline static void osm_perfmgr_set_sweep_time_s(osm_perfmgr_t * p_perfmgr,
256 uint16_t time_s)
257 {
258 p_perfmgr->sweep_time_s = time_s;
259 osm_sm_signal(p_perfmgr->sm, OSM_SIGNAL_PERFMGR_SWEEP);
260 }
261
osm_perfmgr_get_sweep_time_s(osm_perfmgr_t * p_perfmgr)262 inline static uint16_t osm_perfmgr_get_sweep_time_s(osm_perfmgr_t * p_perfmgr)
263 {
264 return p_perfmgr->sweep_time_s;
265 }
266
osm_perfmgr_delete_inactive(osm_perfmgr_t * pm)267 inline static unsigned osm_perfmgr_delete_inactive(osm_perfmgr_t * pm)
268 {
269 unsigned rc;
270 perfmgr_db_delete_inactive(pm->db, &rc);
271 return (rc);
272 }
273
274 void osm_perfmgr_clear_counters(osm_perfmgr_t * p_perfmgr);
275 void osm_perfmgr_dump_counters(osm_perfmgr_t * p_perfmgr,
276 perfmgr_db_dump_t dump_type);
277 void osm_perfmgr_print_counters(osm_perfmgr_t *pm, char *nodename, FILE *fp,
278 char *port, int err_only);
279 void osm_perfmgr_update_nodename(osm_perfmgr_t *pm, uint64_t node_guid,
280 char *nodename);
281
282 ib_api_status_t osm_perfmgr_bind(osm_perfmgr_t * p_perfmgr,
283 ib_net64_t port_guid);
284
285 void osm_perfmgr_process(osm_perfmgr_t * pm);
286
287 /****f* OpenSM: PerfMgr/osm_perfmgr_init */
288 ib_api_status_t osm_perfmgr_init(osm_perfmgr_t * perfmgr,
289 struct osm_opensm *osm,
290 const osm_subn_opt_t * p_opt);
291 /*
292 * PARAMETERS
293 * perfmgr
294 * [in] Pointer to an osm_perfmgr_t object to initialize.
295 *
296 * osm
297 * [in] Pointer to the OpenSM object.
298 *
299 * p_opt
300 * [in] Pointer to the subnet options structure.
301 *
302 * RETURN VALUES
303 * IB_SUCCESS if the PerfMgr object was initialized successfully.
304 *********/
305
306 #ifdef __cplusplus
307 }
308 #endif /* __cplusplus */
309
310 #endif /* ENABLE_OSM_PERF_MGR */
311
312 #endif /* _OSM_PERFMGR_H_ */
313