-rw-r--r-- | src/process.cc | 118 |
1 files changed, 107 insertions, 11 deletions
diff --git a/src/process.cc b/src/process.cc index 1ffac9f..8a5b5d2 100644 --- a/src/process.cc +++ b/src/process.cc | |||
@@ -1,189 +1,285 @@ | |||
1 | #include <stdio.h> | 1 | #include <stdio.h> |
2 | #include <sys/types.h> | 2 | #include <sys/types.h> |
3 | #include <unistd.h> | 3 | #include <unistd.h> |
4 | #include <signal.h> | 4 | #include <signal.h> |
5 | #include <pwd.h> | 5 | #include <pwd.h> |
6 | #include <grp.h> | 6 | #include <grp.h> |
7 | #include <dirent.h> | ||
7 | #include <sys/wait.h> | 8 | #include <sys/wait.h> |
8 | #include <syslog.h> | 9 | #include <syslog.h> |
9 | #include <errno.h> | 10 | #include <errno.h> |
10 | #include <iostream> | 11 | #include <iostream> |
11 | #include <fstream> | 12 | #include <fstream> |
13 | #include <sstream> | ||
12 | #include <stdexcept> | 14 | #include <stdexcept> |
15 | #include <string> | ||
16 | #include <vector> | ||
13 | using namespace std; | 17 | using namespace std; |
14 | #include "process.h" | 18 | #include "process.h" |
15 | #include "configuration.h" | 19 | #include "configuration.h" |
16 | 20 | ||
21 | static multimap<string,pid_t> procpids; | ||
22 | |||
23 | void process::check() const { | ||
24 | if(!pidfile.empty()) { | ||
25 | signal(0); | ||
26 | }else if(!process_name.empty()) { | ||
27 | if(procpids.empty()) | ||
28 | gather_proc_info(); | ||
29 | if(procpids.find(process_name)==procpids.end()) | ||
30 | throw runtime_error("no such process"); | ||
31 | } // XXX: or else? | ||
32 | } | ||
17 | void process::check(const string& id,configuration& config) { | 33 | void process::check(const string& id,configuration& config) { |
18 | try { | 34 | try { |
19 | signal(0); | 35 | check(); |
20 | patience = 0; | 36 | patience = 0; |
21 | }catch(exception& e) { | 37 | }catch(exception& e) { |
22 | if(patience>60) { // TODO: configurable | 38 | if(patience>60) { // TODO: configurable |
23 | patience = 0; | 39 | patience = 0; |
24 | }else{ | 40 | }else{ |
25 | if(patience<10) { // TODO: configurable | 41 | if(patience<10) { // TODO: configurable |
26 | syslog(LOG_NOTICE,"The process '%s' is down, trying to launch.",id.c_str()); | 42 | syslog(LOG_NOTICE,"The process '%s' is down, trying to launch.",id.c_str()); |
27 | do_notify(id,"Starting up", | 43 | do_notify(id,"Starting up", |
28 | "The named process seems to be down. Dudki will try\n" | 44 | "The named process seems to be down. Dudki will try\n" |
29 | "to revive it by running the specified command.\n", | 45 | "to revive it by running the specified command.\n", |
30 | config); | 46 | config); |
31 | try { | 47 | try { |
32 | launch(id,config); | 48 | launch(id,config); |
33 | }catch(exception& e) { | 49 | }catch(exception& e) { |
34 | syslog(LOG_ERR,"Error trying to launch process '%s': %s",id.c_str(),e.what()); | 50 | syslog(LOG_ERR,"Error trying to launch process '%s': %s",id.c_str(),e.what()); |
35 | } | 51 | } |
36 | }else if(patience==10){ // TODO: configurable like the above | 52 | }else if(patience==10){ // TODO: configurable like the above |
37 | syslog(LOG_NOTICE,"Giving up on process '%s' for a while",id.c_str()); | 53 | syslog(LOG_NOTICE,"Giving up on process '%s' for a while",id.c_str()); |
38 | do_notify(id,"Giving up", | 54 | do_notify(id,"Giving up", |
39 | "After a number of attempts to relaunch the named process\n" | 55 | "After a number of attempts to relaunch the named process\n" |
40 | "It still seems to be down. Dudki is giving up attempts\n" | 56 | "It still seems to be down. Dudki is giving up attempts\n" |
41 | "to revive the process for a while.\n", | 57 | "to revive the process for a while.\n", |
42 | config); | 58 | config); |
43 | } | 59 | } |
44 | patience++; | 60 | patience++; |
45 | } | 61 | } |
46 | } | 62 | } |
47 | } | 63 | } |
48 | 64 | ||
49 | void process::launch(const string& id,configuration& config) { | 65 | void process::launch(const string& id,configuration& config) { |
50 | uid_t uid = (uid_t)-1; | 66 | uid_t uid = (uid_t)-1; |
51 | gid_t gid = (gid_t)-1; | 67 | gid_t gid = (gid_t)-1; |
52 | if(!user.empty()) { | 68 | if(!user.empty()) { |
53 | struct passwd *ptmp = getpwnam(user.c_str()); | 69 | struct passwd *ptmp = getpwnam(user.c_str()); |
54 | if(ptmp) { | 70 | if(ptmp) { |
55 | uid = ptmp->pw_uid; | 71 | uid = ptmp->pw_uid; |
56 | gid = ptmp->pw_gid; | 72 | gid = ptmp->pw_gid; |
57 | }else{ | 73 | }else{ |
58 | errno=0; | 74 | errno=0; |
59 | uid = strtol(user.c_str(),NULL,0); | 75 | uid = strtol(user.c_str(),NULL,0); |
60 | if(errno) | 76 | if(errno) |
61 | throw runtime_error("Failed to resolve User value to uid"); | 77 | throw runtime_error("Failed to resolve User value to uid"); |
62 | } | 78 | } |
63 | } | 79 | } |
64 | if(!group.empty()) { | 80 | if(!group.empty()) { |
65 | struct group *gtmp = getgrnam(group.c_str()); | 81 | struct group *gtmp = getgrnam(group.c_str()); |
66 | if(gtmp) { | 82 | if(gtmp) { |
67 | gid = gtmp->gr_gid; | 83 | gid = gtmp->gr_gid; |
68 | }else{ | 84 | }else{ |
69 | errno = 0; | 85 | errno = 0; |
70 | gid = strtol(group.c_str(),NULL,0); | 86 | gid = strtol(group.c_str(),NULL,0); |
71 | if(errno) | 87 | if(errno) |
72 | throw runtime_error("Failed to reslove Group value to gid"); | 88 | throw runtime_error("Failed to reslove Group value to gid"); |
73 | } | 89 | } |
74 | } | 90 | } |
75 | pid_t p = fork(); | 91 | pid_t p = fork(); |
76 | if(p<0) | 92 | if(p<0) |
77 | throw runtime_error(string(__PRETTY_FUNCTION__)+": failed to fork()"); | 93 | throw runtime_error(string(__PRETTY_FUNCTION__)+": failed to fork()"); |
78 | if(!p) { | 94 | if(!p) { |
79 | // child | 95 | // child |
80 | try { | 96 | try { |
81 | setsid(); | 97 | setsid(); |
82 | if(user.empty()) { | 98 | if(user.empty()) { |
83 | if((getgid()!=gid) && setgid(gid)) | 99 | if((getgid()!=gid) && setgid(gid)) |
84 | throw runtime_error(string(__PRETTY_FUNCTION__)+": failed to setgid()"); | 100 | throw runtime_error(string(__PRETTY_FUNCTION__)+": failed to setgid()"); |
85 | }else{ | 101 | }else{ |
86 | if(initgroups(user.c_str(),gid)) | 102 | if(initgroups(user.c_str(),gid)) |
87 | throw runtime_error(string(__PRETTY_FUNCTION__)+": failed to initgroups()"); | 103 | throw runtime_error(string(__PRETTY_FUNCTION__)+": failed to initgroups()"); |
88 | } | 104 | } |
89 | if(!chroot.empty()) { | 105 | if(!chroot.empty()) { |
90 | if(::chroot(chroot.c_str())) | 106 | if(::chroot(chroot.c_str())) |
91 | throw runtime_error(string(__PRETTY_FUNCTION__)+": failed to chroot()"); | 107 | throw runtime_error(string(__PRETTY_FUNCTION__)+": failed to chroot()"); |
92 | } | 108 | } |
93 | if(!user.empty()) { | 109 | if(!user.empty()) { |
94 | if((getuid()!=uid) && setuid(uid)) | 110 | if((getuid()!=uid) && setuid(uid)) |
95 | throw runtime_error(string(__PRETTY_FUNCTION__)+": failed to setuid()"); | 111 | throw runtime_error(string(__PRETTY_FUNCTION__)+": failed to setuid()"); |
96 | } | 112 | } |
97 | char *argv[] = { "/bin/sh", "-c", (char*)restart_cmd.c_str(), NULL }; | 113 | char *argv[] = { "/bin/sh", "-c", (char*)restart_cmd.c_str(), NULL }; |
98 | close(0); close(1); close(2); | 114 | close(0); close(1); close(2); |
99 | execv("/bin/sh",argv); | 115 | execv("/bin/sh",argv); |
100 | }catch(exception& e) { | 116 | }catch(exception& e) { |
101 | syslog(LOG_ERR,"Error trying to launch process '%s': %s",id.c_str(),e.what()); | 117 | syslog(LOG_ERR,"Error trying to launch process '%s': %s",id.c_str(),e.what()); |
102 | } | 118 | } |
103 | _exit(-1); | 119 | _exit(-1); |
104 | } | 120 | } |
105 | // parent | 121 | // parent |
106 | int rv; | 122 | int rv; |
107 | if(waitpid(p,&rv,0)<0) | 123 | if(waitpid(p,&rv,0)<0) |
108 | throw runtime_error(string(__PRETTY_FUNCTION__)+": failed to waitpid()"); | 124 | throw runtime_error(string(__PRETTY_FUNCTION__)+": failed to waitpid()"); |
109 | } | 125 | } |
110 | 126 | ||
111 | void process::do_notify(const string& id,const string& event,const string& description,configuration& config) { | 127 | void process::do_notify(const string& id,const string& event,const string& description,configuration& config) { |
112 | string the_notify; | 128 | string the_notify; |
113 | if(!notify.empty()) | 129 | if(!notify.empty()) |
114 | the_notify=notify; | 130 | the_notify=notify; |
115 | else if(!config.notify.empty()) | 131 | else if(!config.notify.empty()) |
116 | the_notify=config.notify; | 132 | the_notify=config.notify; |
117 | else | 133 | else |
118 | return; | 134 | return; |
119 | try { | 135 | try { |
120 | string::size_type colon = the_notify.find(':'); | 136 | string::size_type colon = the_notify.find(':'); |
121 | if(colon==string::npos) | 137 | if(colon==string::npos) |
122 | throw runtime_error("invalid notify action specification"); | 138 | throw runtime_error("invalid notify action specification"); |
123 | string nschema = the_notify.substr(0,colon); | 139 | string nschema = the_notify.substr(0,colon); |
124 | string ntarget = the_notify.substr(colon+1); | 140 | string ntarget = the_notify.substr(colon+1); |
125 | if(nschema=="mailto") { | 141 | if(nschema=="mailto") { |
126 | notify_mailto(ntarget,id,event,description,config); | 142 | notify_mailto(ntarget,id,event,description,config); |
127 | }else | 143 | }else |
128 | throw runtime_error("unrecognized notification schema"); | 144 | throw runtime_error("unrecognized notification schema"); |
129 | }catch(exception& e) { | 145 | }catch(exception& e) { |
130 | syslog(LOG_ERR,"Notification error: %s",e.what()); | 146 | syslog(LOG_ERR,"Notification error: %s",e.what()); |
131 | } | 147 | } |
132 | } | 148 | } |
133 | 149 | ||
134 | void process::notify_mailto(const string& email,const string& id,const string& event,const string& description,configuration& config) { | 150 | void process::notify_mailto(const string& email,const string& id,const string& event,const string& description,configuration& config) { |
135 | int files[2]; | 151 | int files[2]; |
136 | if(pipe(files)) | 152 | if(pipe(files)) |
137 | throw runtime_error("Failed to pipe()"); | 153 | throw runtime_error("Failed to pipe()"); |
138 | pid_t pid = vfork(); | 154 | pid_t pid = vfork(); |
139 | if(pid==-1) { | 155 | if(pid==-1) { |
140 | close(files[0]); | 156 | close(files[0]); |
141 | close(files[1]); | 157 | close(files[1]); |
142 | throw runtime_error("Failed to vfork()"); | 158 | throw runtime_error("Failed to vfork()"); |
143 | } | 159 | } |
144 | if(!pid) { | 160 | if(!pid) { |
145 | // child | 161 | // child |
146 | if(dup2(files[0],0)!=0) | 162 | if(dup2(files[0],0)!=0) |
147 | _exit(-1); | 163 | _exit(-1); |
148 | close(1); | 164 | close(1); |
149 | close(files[0]); | 165 | close(files[0]); |
150 | close(files[1]); | 166 | close(files[1]); |
151 | execl("/usr/sbin/sendmail","usr/sbin/sendmail","-i",email.c_str(),NULL); | 167 | execl("/usr/sbin/sendmail","usr/sbin/sendmail","-i",email.c_str(),NULL); |
152 | _exit(-1); | 168 | _exit(-1); |
153 | } | 169 | } |
154 | // parent | 170 | // parent |
155 | close(files[0]); | 171 | close(files[0]); |
156 | FILE *mta = fdopen(files[1],"w"); | 172 | FILE *mta = fdopen(files[1],"w"); |
157 | for(headers_t::const_iterator i=mailto_headers.begin();i!=mailto_headers.end();++i) { | 173 | for(headers_t::const_iterator i=mailto_headers.begin();i!=mailto_headers.end();++i) { |
158 | fprintf(mta,"%s: %s\n",i->first.c_str(),i->second.c_str()); | 174 | fprintf(mta,"%s: %s\n",i->first.c_str(),i->second.c_str()); |
159 | } | 175 | } |
160 | for(headers_t::const_iterator i=config.mailto_headers.begin();i!=config.mailto_headers.end();++i) { | 176 | for(headers_t::const_iterator i=config.mailto_headers.begin();i!=config.mailto_headers.end();++i) { |
161 | if(mailto_headers.find(i->first)!=mailto_headers.end()) | 177 | if(mailto_headers.find(i->first)!=mailto_headers.end()) |
162 | continue; | 178 | continue; |
163 | fprintf(mta,"%s: %s\n",i->first.c_str(),i->second.c_str()); | 179 | fprintf(mta,"%s: %s\n",i->first.c_str(),i->second.c_str()); |
164 | } | 180 | } |
165 | fprintf(mta, | 181 | fprintf(mta, |
166 | "Subject: [%s] %s\n\n" | 182 | "Subject: [%s] %s\n\n" |
167 | "%s\n" | 183 | "%s\n" |
168 | "---\n" | 184 | "---\n" |
169 | "This message was sent automatically by the 'dudki' daemon\n", | 185 | "This message was sent automatically by the 'dudki' daemon\n", |
170 | id.c_str(), event.c_str(), | 186 | id.c_str(), event.c_str(), |
171 | description.c_str() ); | 187 | description.c_str() ); |
172 | fclose(mta); | 188 | fclose(mta); |
173 | int status; | 189 | int status; |
174 | waitpid(pid,&status,0); | 190 | waitpid(pid,&status,0); |
175 | // TODO: check the return code | 191 | // TODO: check the return code |
176 | } | 192 | } |
177 | 193 | ||
178 | void process::signal(int signum) const { | 194 | void process::signal(int signum) const { |
179 | ifstream pids(pidfile.c_str(),ios::in); | 195 | if(!pidfile.empty()) { |
180 | if(!pids) | 196 | ifstream pids(pidfile.c_str(),ios::in); |
181 | throw runtime_error("no pidfile found"); | 197 | if(!pids) |
182 | pid_t pid = 0; | 198 | throw runtime_error("no pidfile found"); |
183 | pids >> pid; | 199 | pid_t pid = 0; |
184 | pids.close(); | 200 | pids >> pid; |
185 | if(!pid) | 201 | pids.close(); |
186 | throw runtime_error("no pid in pidfile"); | 202 | if(!pid) |
187 | if(kill(pid,signum)) | 203 | throw runtime_error("no pid in pidfile"); |
188 | throw runtime_error("failed to signal process"); | 204 | if(kill(pid,signum)) |
205 | throw runtime_error("failed to signal process"); | ||
206 | }else if(!process_name.empty()) { | ||
207 | if(procpids.empty()) | ||
208 | gather_proc_info(); | ||
209 | pair<multimap<string,pid_t>::const_iterator,multimap<string,pid_t>::const_iterator> range = procpids.equal_range(process_name); | ||
210 | int count = 0; | ||
211 | for(multimap<string,pid_t>::const_iterator i=range.first;i!=range.second;++i) { | ||
212 | pid_t pid = i->second; | ||
213 | if(kill(i->second,signum)) | ||
214 | throw runtime_error("failed to signal process"); | ||
215 | count++; | ||
216 | } | ||
217 | if(!count) | ||
218 | throw runtime_error("no running instance detected"); | ||
219 | }else | ||
220 | throw runtime_error("nothing is known about the process"); | ||
221 | } | ||
222 | |||
223 | void process::prepare_herd() { | ||
224 | procpids.clear(); | ||
225 | } | ||
226 | void process::unprepare_herd() { | ||
227 | procpids.clear(); | ||
228 | } | ||
229 | void process::gather_proc_info() { | ||
230 | vector<pid_t> allpids; | ||
231 | DIR *pd = opendir("/proc"); | ||
232 | if(!pd) | ||
233 | throw runtime_error("failed to open /proc"); | ||
234 | struct dirent *pde; | ||
235 | pid_t selfpid = getpid(); | ||
236 | while(pde=readdir(pd)) { | ||
237 | errno=0; | ||
238 | pid_t pid = atoi(pde->d_name); | ||
239 | if((!pid) || pid==selfpid) | ||
240 | continue; | ||
241 | allpids.push_back(pid); | ||
242 | } | ||
243 | closedir(pd); | ||
244 | char s[256]; | ||
245 | procpids.clear(); | ||
246 | for(vector<pid_t>::const_iterator i=allpids.begin();i!=allpids.end();++i) { | ||
247 | int r = snprintf(s,sizeof(s),"/proc/%d/stat",*i); | ||
248 | if(r>=sizeof(s) || r<1) | ||
249 | continue; | ||
250 | string cmd; | ||
251 | ifstream ss(s,ios::in); | ||
252 | if(!ss) | ||
253 | continue; | ||
254 | getline(ss,cmd); | ||
255 | string::size_type op = cmd.find('('); | ||
256 | if(op==string::npos) | ||
257 | continue; | ||
258 | cmd.erase(0,op+1); | ||
259 | string::size_type cp = cmd.find(')'); | ||
260 | if(cp==string::npos) | ||
261 | continue; | ||
262 | cmd.erase(cp); | ||
263 | r = snprintf(s,sizeof(s),"/proc/%d/cmdline",*i); | ||
264 | if(r>=sizeof(s) || r<1) | ||
265 | continue; | ||
266 | ifstream cs(s,ios::binary); | ||
267 | if(!cs) | ||
268 | continue; | ||
269 | string command; | ||
270 | while(cs) { | ||
271 | string cl; | ||
272 | getline(cs,cl,(char)0); | ||
273 | string::size_type lsl = cl.rfind('/'); | ||
274 | if(lsl!=string::npos) | ||
275 | cl.erase(0,lsl+1); | ||
276 | if(cl.substr(0,cmd.length())==cmd) { | ||
277 | command = cl; | ||
278 | break; | ||
279 | } | ||
280 | } | ||
281 | procpids.insert(pair<string,pid_t>(cmd,*i)); | ||
282 | if((!command.empty()) && cmd!=command) | ||
283 | procpids.insert(pair<string,pid_t>(command,*i)); | ||
284 | } | ||
189 | } | 285 | } |