1 |
/* |
---|
2 |
* Copyright (C) 2005-2008 by Pieter Palmers |
---|
3 |
* |
---|
4 |
* This file is part of FFADO |
---|
5 |
* FFADO = Free Firewire (pro-)audio drivers for linux |
---|
6 |
* |
---|
7 |
* FFADO is based upon FreeBoB. |
---|
8 |
* |
---|
9 |
* This program is free software: you can redistribute it and/or modify |
---|
10 |
* it under the terms of the GNU General Public License as published by |
---|
11 |
* the Free Software Foundation, either version 2 of the License, or |
---|
12 |
* (at your option) version 3 of the License. |
---|
13 |
* |
---|
14 |
* This program is distributed in the hope that it will be useful, |
---|
15 |
* but WITHOUT ANY WARRANTY; without even the implied warranty of |
---|
16 |
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
---|
17 |
* GNU General Public License for more details. |
---|
18 |
* |
---|
19 |
* You should have received a copy of the GNU General Public License |
---|
20 |
* along with this program. If not, see <http://www.gnu.org/licenses/>. |
---|
21 |
* |
---|
22 |
*/ |
---|
23 |
|
---|
24 |
#include "Watchdog.h" |
---|
25 |
#include "SystemTimeSource.h" |
---|
26 |
#include "PosixThread.h" |
---|
27 |
|
---|
28 |
#include "config.h" |
---|
29 |
|
---|
30 |
// needed for clock_nanosleep |
---|
31 |
#ifndef _GNU_SOURCE |
---|
32 |
#define _GNU_SOURCE |
---|
33 |
#endif |
---|
34 |
|
---|
35 |
#include <time.h> |
---|
36 |
|
---|
37 |
namespace Util { |
---|
38 |
|
---|
39 |
IMPL_DEBUG_MODULE( Watchdog, Watchdog, DEBUG_LEVEL_NORMAL ); |
---|
40 |
|
---|
41 |
// --- liveness check Thread --- // |
---|
42 |
Watchdog::WatchdogCheckTask::WatchdogCheckTask(Watchdog& parent, unsigned int interval_usecs) |
---|
43 |
: m_parent( parent ) |
---|
44 |
, m_interval( interval_usecs ) |
---|
45 |
, m_debugModule( parent.m_debugModule ) |
---|
46 |
{ |
---|
47 |
} |
---|
48 |
|
---|
49 |
bool |
---|
50 |
Watchdog::WatchdogCheckTask::Init() |
---|
51 |
{ |
---|
52 |
#ifdef DEBUG |
---|
53 |
m_last_loop_entry = 0; |
---|
54 |
m_successive_short_loops = 0; |
---|
55 |
#endif |
---|
56 |
return true; |
---|
57 |
} |
---|
58 |
|
---|
59 |
bool |
---|
60 |
Watchdog::WatchdogCheckTask::Execute() |
---|
61 |
{ |
---|
62 |
SystemTimeSource::SleepUsecRelative(m_interval); |
---|
63 |
if(m_parent.getHartbeat()) { |
---|
64 |
debugOutput(DEBUG_LEVEL_VERY_VERBOSE, |
---|
65 |
"(%p) watchdog %p still alive\n", this, &m_parent); |
---|
66 |
m_parent.clearHartbeat(); |
---|
67 |
} else { |
---|
68 |
debugWarning("(%p) watchdog %p died\n", this, &m_parent); |
---|
69 |
// set all watched threads to non-rt scheduling |
---|
70 |
m_parent.rescheduleThreads(); |
---|
71 |
} |
---|
72 |
|
---|
73 |
#ifdef DEBUG |
---|
74 |
uint64_t now = Util::SystemTimeSource::getCurrentTimeAsUsecs(); |
---|
75 |
int diff = now - m_last_loop_entry; |
---|
76 |
if(diff < 100) { |
---|
77 |
debugOutputExtreme(DEBUG_LEVEL_VERY_VERBOSE, |
---|
78 |
"(%p) short loop detected (%d usec), cnt: %d\n", |
---|
79 |
this, diff, m_successive_short_loops); |
---|
80 |
m_successive_short_loops++; |
---|
81 |
if(m_successive_short_loops > 100) { |
---|
82 |
debugError("Shutting down runaway thread\n"); |
---|
83 |
return false; |
---|
84 |
} |
---|
85 |
} else { |
---|
86 |
// reset the counter |
---|
87 |
m_successive_short_loops = 0; |
---|
88 |
} |
---|
89 |
m_last_loop_entry = now; |
---|
90 |
#endif |
---|
91 |
|
---|
92 |
return true; |
---|
93 |
} |
---|
94 |
|
---|
95 |
// --- hartbeat Thread --- // |
---|
96 |
|
---|
97 |
Watchdog::WatchdogHartbeatTask::WatchdogHartbeatTask(Watchdog& parent, unsigned int interval_usecs) |
---|
98 |
: m_parent( parent ) |
---|
99 |
, m_interval( interval_usecs ) |
---|
100 |
, m_debugModule( parent.m_debugModule ) |
---|
101 |
{ |
---|
102 |
} |
---|
103 |
|
---|
104 |
bool |
---|
105 |
Watchdog::WatchdogHartbeatTask::Init() |
---|
106 |
{ |
---|
107 |
#ifdef DEBUG |
---|
108 |
m_last_loop_entry = 0; |
---|
109 |
m_successive_short_loops = 0; |
---|
110 |
#endif |
---|
111 |
return true; |
---|
112 |
} |
---|
113 |
|
---|
114 |
bool |
---|
115 |
Watchdog::WatchdogHartbeatTask::Execute() |
---|
116 |
{ |
---|
117 |
SystemTimeSource::SleepUsecRelative(m_interval); |
---|
118 |
debugOutput(DEBUG_LEVEL_VERY_VERBOSE, |
---|
119 |
"(%p) watchdog %p hartbeat\n", this, &m_parent); |
---|
120 |
m_parent.setHartbeat(); |
---|
121 |
|
---|
122 |
#ifdef DEBUG |
---|
123 |
uint64_t now = Util::SystemTimeSource::getCurrentTimeAsUsecs(); |
---|
124 |
int diff = now - m_last_loop_entry; |
---|
125 |
if(diff < 100) { |
---|
126 |
debugOutputExtreme(DEBUG_LEVEL_VERY_VERBOSE, |
---|
127 |
"(%p) short loop detected (%d usec), cnt: %d\n", |
---|
128 |
this, diff, m_successive_short_loops); |
---|
129 |
m_successive_short_loops++; |
---|
130 |
if(m_successive_short_loops > 100) { |
---|
131 |
debugError("Shutting down runaway thread\n"); |
---|
132 |
return false; |
---|
133 |
} |
---|
134 |
} else { |
---|
135 |
// reset the counter |
---|
136 |
m_successive_short_loops = 0; |
---|
137 |
} |
---|
138 |
m_last_loop_entry = now; |
---|
139 |
#endif |
---|
140 |
|
---|
141 |
return true; |
---|
142 |
} |
---|
143 |
|
---|
144 |
// the actual watchdog class |
---|
145 |
Watchdog::Watchdog() |
---|
146 |
: m_hartbeat( true ) |
---|
147 |
, m_check_interval( WATCHDOG_DEFAULT_CHECK_INTERVAL_USECS ) |
---|
148 |
, m_realtime( WATCHDOG_DEFAULT_RUN_REALTIME ) |
---|
149 |
, m_priority( WATCHDOG_DEFAULT_PRIORITY ) |
---|
150 |
, m_CheckThread( NULL ) |
---|
151 |
, m_HartbeatThread( NULL ) |
---|
152 |
, m_CheckTask( NULL ) |
---|
153 |
, m_HartbeatTask( NULL ) |
---|
154 |
{ |
---|
155 |
} |
---|
156 |
|
---|
157 |
Watchdog::Watchdog(unsigned int interval_usec, bool realtime, unsigned int priority) |
---|
158 |
: m_hartbeat( true ) |
---|
159 |
, m_check_interval( interval_usec ) |
---|
160 |
, m_realtime( realtime ) |
---|
161 |
, m_priority( priority ) |
---|
162 |
, m_CheckThread( NULL ) |
---|
163 |
, m_HartbeatThread( NULL ) |
---|
164 |
, m_CheckTask( NULL ) |
---|
165 |
, m_HartbeatTask( NULL ) |
---|
166 |
{ |
---|
167 |
} |
---|
168 |
|
---|
169 |
Watchdog::~Watchdog() |
---|
170 |
{ |
---|
171 |
// kill threads instead of stoping them since |
---|
172 |
// they are sleeping |
---|
173 |
if (m_CheckThread) { |
---|
174 |
//m_CheckThread->Stop(); |
---|
175 |
m_CheckThread->Kill(); |
---|
176 |
delete m_CheckThread; |
---|
177 |
} |
---|
178 |
if (m_HartbeatThread) { |
---|
179 |
//m_HartbeatThread->Stop(); |
---|
180 |
m_HartbeatThread->Kill(); |
---|
181 |
delete m_HartbeatThread; |
---|
182 |
} |
---|
183 |
if (m_CheckTask) { |
---|
184 |
delete m_CheckTask; |
---|
185 |
} |
---|
186 |
if (m_HartbeatTask) { |
---|
187 |
delete m_HartbeatTask; |
---|
188 |
} |
---|
189 |
} |
---|
190 |
|
---|
191 |
void |
---|
192 |
Watchdog::setVerboseLevel(int i) |
---|
193 |
{ |
---|
194 |
setDebugLevel(i); |
---|
195 |
} |
---|
196 |
|
---|
197 |
bool |
---|
198 |
Watchdog::start() |
---|
199 |
{ |
---|
200 |
debugOutput( DEBUG_LEVEL_VERBOSE, "(%p) Starting watchdog...\n", this); |
---|
201 |
debugOutput( DEBUG_LEVEL_VERBOSE, "Create hartbeat task/thread for %p...\n", this); |
---|
202 |
m_HartbeatTask = new WatchdogHartbeatTask( *this, m_check_interval/2 ); |
---|
203 |
if(!m_HartbeatTask) { |
---|
204 |
debugFatal("No hartbeat task\n"); |
---|
205 |
return false; |
---|
206 |
} |
---|
207 |
m_HartbeatThread = new Util::PosixThread(m_HartbeatTask, "WDGHBT", false, |
---|
208 |
0, PTHREAD_CANCEL_ASYNCHRONOUS); |
---|
209 |
if(!m_HartbeatThread) { |
---|
210 |
debugFatal("No hartbeat thread\n"); |
---|
211 |
return false; |
---|
212 |
} |
---|
213 |
debugOutput( DEBUG_LEVEL_VERBOSE, |
---|
214 |
" hartbeat task: %p, thread %p...\n", |
---|
215 |
m_HartbeatTask, m_HartbeatThread); |
---|
216 |
|
---|
217 |
debugOutput( DEBUG_LEVEL_VERBOSE, "Create check task/thread for %p...\n", this); |
---|
218 |
m_CheckTask = new WatchdogCheckTask( *this, m_check_interval ); |
---|
219 |
if(!m_CheckTask) { |
---|
220 |
debugFatal("No check task\n"); |
---|
221 |
return false; |
---|
222 |
} |
---|
223 |
m_CheckThread = new Util::PosixThread(m_CheckTask,"WDGCHK", false, |
---|
224 |
0, PTHREAD_CANCEL_ASYNCHRONOUS); |
---|
225 |
if(!m_CheckThread) { |
---|
226 |
debugFatal("No check thread\n"); |
---|
227 |
return false; |
---|
228 |
} |
---|
229 |
debugOutput( DEBUG_LEVEL_VERBOSE, |
---|
230 |
" check task: %p, thread %p...\n", |
---|
231 |
m_CheckTask, m_CheckThread); |
---|
232 |
|
---|
233 |
// switch to realtime if necessary |
---|
234 |
if(m_realtime) { |
---|
235 |
if(!m_CheckThread->AcquireRealTime(m_priority)) { |
---|
236 |
debugWarning("(%p) Could not aquire realtime priotiry for watchdog thread.\n", this); |
---|
237 |
} |
---|
238 |
} |
---|
239 |
|
---|
240 |
// start threads |
---|
241 |
if (m_HartbeatThread->Start() != 0) { |
---|
242 |
debugFatal("Could not start hartbeat thread\n"); |
---|
243 |
return false; |
---|
244 |
} |
---|
245 |
if (m_CheckThread->Start() != 0) { |
---|
246 |
debugFatal("Could not start check thread\n"); |
---|
247 |
return false; |
---|
248 |
} |
---|
249 |
debugOutput( DEBUG_LEVEL_VERBOSE, "(%p) Watchdog running...\n", this); |
---|
250 |
return true; |
---|
251 |
} |
---|
252 |
|
---|
253 |
bool |
---|
254 |
Watchdog::setThreadParameters(bool rt, int priority) |
---|
255 |
{ |
---|
256 |
debugOutput( DEBUG_LEVEL_VERBOSE, "(%p) switch to: (rt=%d, prio=%d)...\n", this, rt, priority); |
---|
257 |
if (priority > THREAD_MAX_RTPRIO) priority = THREAD_MAX_RTPRIO; // cap the priority |
---|
258 |
m_realtime = rt; |
---|
259 |
m_priority = priority; |
---|
260 |
|
---|
261 |
if (m_CheckThread) { |
---|
262 |
if (m_realtime) { |
---|
263 |
m_CheckThread->AcquireRealTime(m_priority); |
---|
264 |
} else { |
---|
265 |
m_CheckThread->DropRealTime(); |
---|
266 |
} |
---|
267 |
} |
---|
268 |
return true; |
---|
269 |
} |
---|
270 |
|
---|
271 |
/** |
---|
272 |
* register a thread to the watchdog |
---|
273 |
* @param thread |
---|
274 |
* @return |
---|
275 |
*/ |
---|
276 |
bool |
---|
277 |
Watchdog::registerThread(Thread *thread) |
---|
278 |
{ |
---|
279 |
assert(thread); |
---|
280 |
debugOutput( DEBUG_LEVEL_VERBOSE, "(%p) Adding thread %p\n", |
---|
281 |
this, thread); |
---|
282 |
|
---|
283 |
for ( ThreadVectorIterator it = m_Threads.begin(); |
---|
284 |
it != m_Threads.end(); |
---|
285 |
++it ) |
---|
286 |
{ |
---|
287 |
if(*it == thread) { |
---|
288 |
debugError("Thread %p already registered with watchdog\n", thread); |
---|
289 |
return false; |
---|
290 |
} |
---|
291 |
} |
---|
292 |
m_Threads.push_back(thread); |
---|
293 |
return true; |
---|
294 |
} |
---|
295 |
|
---|
296 |
bool |
---|
297 |
Watchdog::unregisterThread(Thread *thread) |
---|
298 |
{ |
---|
299 |
assert(thread); |
---|
300 |
debugOutput( DEBUG_LEVEL_VERBOSE, "(%p) unregistering thread %p\n", this, thread); |
---|
301 |
|
---|
302 |
for ( ThreadVectorIterator it = m_Threads.begin(); |
---|
303 |
it != m_Threads.end(); |
---|
304 |
++it ) |
---|
305 |
{ |
---|
306 |
if(*it == thread) { |
---|
307 |
m_Threads.erase(it); |
---|
308 |
return true; |
---|
309 |
} |
---|
310 |
} |
---|
311 |
debugOutput( DEBUG_LEVEL_VERBOSE, "(%p) thread %p not found \n", this, thread); |
---|
312 |
return false; //not found |
---|
313 |
} |
---|
314 |
|
---|
315 |
void Watchdog::rescheduleThreads() |
---|
316 |
{ |
---|
317 |
debugOutput( DEBUG_LEVEL_VERBOSE, "(%p) rescheduling threads\n", this); |
---|
318 |
|
---|
319 |
for ( ThreadVectorIterator it = m_Threads.begin(); |
---|
320 |
it != m_Threads.end(); |
---|
321 |
++it ) |
---|
322 |
{ |
---|
323 |
(*it)->DropRealTime(); |
---|
324 |
} |
---|
325 |
} |
---|
326 |
|
---|
327 |
} // end of namespace Util |
---|