diff options
Diffstat (limited to 'linden/indra/llcommon/llheartbeat.cpp')
-rw-r--r-- | linden/indra/llcommon/llheartbeat.cpp | 165 |
1 files changed, 165 insertions, 0 deletions
diff --git a/linden/indra/llcommon/llheartbeat.cpp b/linden/indra/llcommon/llheartbeat.cpp new file mode 100644 index 0000000..13bcd46 --- /dev/null +++ b/linden/indra/llcommon/llheartbeat.cpp | |||
@@ -0,0 +1,165 @@ | |||
1 | /** | ||
2 | * @file llheartbeat.cpp | ||
3 | * @brief Class encapsulating logic for telling a watchdog that we live. | ||
4 | * | ||
5 | * $LicenseInfo:firstyear=2008&license=viewergpl$ | ||
6 | * | ||
7 | * Copyright (c) 2008, Linden Research, Inc. | ||
8 | * | ||
9 | * Second Life Viewer Source Code | ||
10 | * The source code in this file ("Source Code") is provided by Linden Lab | ||
11 | * to you under the terms of the GNU General Public License, version 2.0 | ||
12 | * ("GPL"), unless you have obtained a separate licensing agreement | ||
13 | * ("Other License"), formally executed by you and Linden Lab. Terms of | ||
14 | * the GPL can be found in doc/GPL-license.txt in this distribution, or | ||
15 | * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2 | ||
16 | * | ||
17 | * There are special exceptions to the terms and conditions of the GPL as | ||
18 | * it is applied to this Source Code. View the full text of the exception | ||
19 | * in the file doc/FLOSS-exception.txt in this software distribution, or | ||
20 | * online at http://secondlifegrid.net/programs/open_source/licensing/flossexception | ||
21 | * | ||
22 | * By copying, modifying or distributing this software, you acknowledge | ||
23 | * that you have read and understood your obligations described above, | ||
24 | * and agree to abide by those obligations. | ||
25 | * | ||
26 | * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO | ||
27 | * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY, | ||
28 | * COMPLETENESS OR PERFORMANCE. | ||
29 | * $/LicenseInfo$ | ||
30 | */ | ||
31 | |||
32 | #include <errno.h> | ||
33 | #include <signal.h> | ||
34 | |||
35 | #include "linden_common.h" | ||
36 | #include "llapp.h" | ||
37 | |||
38 | #include "llheartbeat.h" | ||
39 | |||
40 | LLHeartbeat::LLHeartbeat(F32 secs_between_heartbeat, | ||
41 | F32 aggressive_heartbeat_panic_secs, | ||
42 | F32 aggressive_heartbeat_max_blocking_secs) | ||
43 | : mSecsBetweenHeartbeat(secs_between_heartbeat), | ||
44 | mAggressiveHeartbeatPanicSecs(aggressive_heartbeat_panic_secs), | ||
45 | mAggressiveHeartbeatMaxBlockingSecs(aggressive_heartbeat_max_blocking_secs), | ||
46 | mSuppressed(false) | ||
47 | { | ||
48 | mBeatTimer.reset(); | ||
49 | mBeatTimer.setTimerExpirySec(mSecsBetweenHeartbeat); | ||
50 | mPanicTimer.reset(); | ||
51 | mPanicTimer.setTimerExpirySec(mAggressiveHeartbeatPanicSecs); | ||
52 | } | ||
53 | |||
54 | LLHeartbeat::~LLHeartbeat() | ||
55 | { | ||
56 | // do nothing. | ||
57 | } | ||
58 | |||
59 | void | ||
60 | LLHeartbeat::setSuppressed(bool is_suppressed) | ||
61 | { | ||
62 | mSuppressed = is_suppressed; | ||
63 | } | ||
64 | |||
65 | // returns 0 on success, -1 on permanent failure, 1 on temporary failure | ||
66 | int | ||
67 | LLHeartbeat::rawSend() | ||
68 | { | ||
69 | #if LL_WINDOWS | ||
70 | return 0; // Pretend we succeeded. | ||
71 | #else | ||
72 | if (mSuppressed) | ||
73 | return 0; // Pretend we succeeded. | ||
74 | |||
75 | union sigval dummy; | ||
76 | int result = sigqueue(getppid(), LL_HEARTBEAT_SIGNAL, dummy); | ||
77 | if (result == 0) | ||
78 | return 0; // success | ||
79 | |||
80 | int err = errno; | ||
81 | if (err == EAGAIN) | ||
82 | return 1; // failed to queue, try again | ||
83 | |||
84 | return -1; // other failure. | ||
85 | #endif | ||
86 | } | ||
87 | |||
88 | int | ||
89 | LLHeartbeat::rawSendWithTimeout(F32 timeout_sec) | ||
90 | { | ||
91 | int result = 0; | ||
92 | |||
93 | // Spin tightly until our heartbeat is digested by the watchdog | ||
94 | // or we time-out. We don't really want to sleep because our | ||
95 | // wake-up time might be undesirably synchronised to a hidden | ||
96 | // clock by the system's scheduler. | ||
97 | mTimeoutTimer.reset(); | ||
98 | mTimeoutTimer.setTimerExpirySec(timeout_sec); | ||
99 | do { | ||
100 | result = rawSend(); | ||
101 | //llinfos << " HEARTSENDc=" << result << llendl; | ||
102 | } while (result==1 && !mTimeoutTimer.hasExpired()); | ||
103 | |||
104 | return result; | ||
105 | } | ||
106 | |||
107 | bool | ||
108 | LLHeartbeat::send(F32 timeout_sec) | ||
109 | { | ||
110 | bool total_success = false; | ||
111 | int result = 1; | ||
112 | |||
113 | if (timeout_sec > 0.f) { | ||
114 | // force a spin until success or timeout | ||
115 | result = rawSendWithTimeout(timeout_sec); | ||
116 | } else { | ||
117 | if (mBeatTimer.hasExpired()) { | ||
118 | // zero-timeout; we don't care too much whether our | ||
119 | // heartbeat was digested. | ||
120 | result = rawSend(); | ||
121 | //llinfos << " HEARTSENDb=" << result << llendl; | ||
122 | } | ||
123 | } | ||
124 | |||
125 | if (result == -1) { | ||
126 | // big failure. | ||
127 | } else if (result == 0) { | ||
128 | total_success = true; | ||
129 | } else { | ||
130 | // need to retry at some point | ||
131 | } | ||
132 | |||
133 | if (total_success) { | ||
134 | mBeatTimer.reset(); | ||
135 | mBeatTimer.setTimerExpirySec(mSecsBetweenHeartbeat); | ||
136 | // reset the time until we start panicking about lost | ||
137 | // heartbeats again. | ||
138 | mPanicTimer.reset(); | ||
139 | mPanicTimer.setTimerExpirySec(mAggressiveHeartbeatPanicSecs); | ||
140 | } else { | ||
141 | // leave mBeatTimer as expired so we'll lazily poke the | ||
142 | // watchdog again next time through. | ||
143 | } | ||
144 | |||
145 | if (mPanicTimer.hasExpired()) { | ||
146 | // It's been ages since we successfully had a heartbeat | ||
147 | // digested by the watchdog. Sit here and spin a while | ||
148 | // in the hope that we can force it through. | ||
149 | llwarns << "Unable to deliver heartbeat to launcher for " << mPanicTimer.getElapsedTimeF32() << " seconds. Going to try very hard for up to " << mAggressiveHeartbeatMaxBlockingSecs << " seconds." << llendl; | ||
150 | result = rawSendWithTimeout(mAggressiveHeartbeatMaxBlockingSecs); | ||
151 | if (result == 0) { | ||
152 | total_success = true; | ||
153 | } else { | ||
154 | // we couldn't even force it through. That's bad, | ||
155 | // but we'll try again in a while. | ||
156 | llwarns << "Could not deliver heartbeat to launcher even after trying very hard for " << mAggressiveHeartbeatMaxBlockingSecs << " seconds." << llendl; | ||
157 | } | ||
158 | |||
159 | // in any case, reset the panic timer. | ||
160 | mPanicTimer.reset(); | ||
161 | mPanicTimer.setTimerExpirySec(mAggressiveHeartbeatPanicSecs); | ||
162 | } | ||
163 | |||
164 | return total_success; | ||
165 | } | ||