1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312
|
/*
* Copyright (c) 2014 Christian Authmann
*/
#include "common/binarystream.h"
#include "common/constants.h"
#include "datatypes/foo.h"
#include "datatypes/bar.h"
#include <cstdlib>
#include <cstdio>
#include <string.h> // memset()
#include <map>
#include <atomic>
#include <iostream>
#include <fstream>
#include <stdexcept>
#include <chrono> // for sleeping
#include <thread>
#include <time.h>
#include <sys/types.h>
#include <sys/socket.h>
#include <sys/un.h>
#include <sys/wait.h>
#include <sys/stat.h>
#include <poll.h>
#include <signal.h>
/*
* This is an example server. It sets up an R environment and a listening socket, then waits
* for clients and fork()s.
* The actual communication with clients is handled by server/rinsideserver.cpp.
* This file only does initialization and process tracking.
*/
/*
* Since the server fork()s a lot, we would like to prepend the pid to each logged line
*/
#define LOG(...) {fprintf(stderr, "%d: ", getpid());fprintf(stderr, __VA_ARGS__);fprintf(stderr, "\n");}
/*
* If an R script gets stuck in an infinite loop, we need to stop it eventually. We thus define a global
* timeout after which it gets killed.
*
* It would be desirable to allow the client to specify the timeout. Unfortunately, only the fork()ed
* child process can communicate with the client, but the parent process needs to know about the timeout,
* making this more complicated than one might expect.
*/
const int TIMEOUT_SECONDS = 600;
#include <RcppCommon.h>
#include "datatypes/foo_rcpp_wrapper_declarations.h"
#include "datatypes/bar_rcpp_wrapper_declarations.h"
#include <Rcpp.h>
#include <RInside.h>
#if !defined(RINSIDE_CALLBACKS)
#error "RInside was not compiled with RINSIDE_CALLBACKS"
#endif
#include "datatypes/foo_rcpp_wrapper_definitions.h"
#include "datatypes/bar_rcpp_wrapper_definitions.h"
/*
* The RInsideServer must be included AFTER RInside and all wrappers are included
*/
#include "server/rinsideserver.h"
int cmpTimespec(const struct timespec &t1, const struct timespec &t2) {
if (t1.tv_sec < t2.tv_sec)
return -1;
if (t1.tv_sec > t2.tv_sec)
return 1;
if (t1.tv_nsec < t2.tv_nsec)
return -1;
if (t1.tv_nsec > t2.tv_nsec)
return 1;
return 0;
}
void signal_handler(int signum) {
LOG("Caught signal %d, exiting", signum);
exit(signum);
}
int main()
{
// register our custom types with the server
RInsideServer::registerDefaultTypes();
RInsideServer::registerType<Foo>();
RInsideServer::registerType<Bar>();
// Install signal handlers
int signals[] = {SIGHUP, SIGINT, 0};
for (int i=0;signals[i];i++) {
if (signal(signals[i], signal_handler) == SIG_ERR) {
perror("Cannot install signal handler");
exit(1);
}
else
printf("Signal handler for %d installed\n", signals[i]);
}
signal(SIGPIPE, SIG_IGN);
/*
* If R prints anything to the console, we must catch it.
* Instead of redirecting stdout (which we might want to use for diagnostics or logging), we
* use RInside's callbacks. They're marked experimental and aren't enabled by default, but in our
* tests, they worked just fine.
*/
RInsideCallbacks *Rcallbacks = new RInsideCallbacks();
// Initialize R environment
printf("...loading R\n");
RInside R;
R.set_callbacks( Rcallbacks );
printf("...loading packages\n");
try {
/*
* Loading packages is slow. We want to load all common packages once on
* server startup, before the fork()
*
* For example, sandboxR might be useful to restrict the damage an R script can do.
* See https://github.com/rapporter/sandboxR
*/
//R.parseEval("library(\"sandboxR\")");
}
catch (const std::exception &e) {
printf("error loading packages: %s\nR's output:\n%s", e.what(), Rcallbacks->getConsoleOutput().c_str());
exit(5);
}
Rcallbacks->resetConsoleOutput();
printf("R is ready\n");
// get rid of leftover sockets
unlink(ris_socket_address);
// create a fresh socket
int listen_fd = socket(AF_UNIX, SOCK_STREAM, 0);
if (listen_fd < 0) {
perror("socket() failed");
exit(1);
}
// bind socket
struct sockaddr_un server_addr;
memset((void *) &server_addr, 0, sizeof(server_addr));
server_addr.sun_family = AF_UNIX;
strcpy(server_addr.sun_path, ris_socket_address);
if (bind(listen_fd, (sockaddr *) &server_addr, sizeof(server_addr)) < 0) {
perror("bind() failed");
exit(1);
}
// adjust this for your own needs..
chmod(ris_socket_address, 0777);
/*
* We need to keep track of all the children to enforce timeouts. This map
* contains pids of all child processes and their end times.
*/
std::map<pid_t, timespec> running_clients;
printf("Socket started, listening..\n");
// Start listening and fork()
listen(listen_fd, 5);
while (true) {
/*
* Try to reap all child processes that exited on their own. Not only
* will this keep our running_clients map small, it will also allow the
* OS to remove any "zombie" processes.
*/
int status;
pid_t exited_pid;
while ((exited_pid = waitpid(-1, &status, WNOHANG)) > 0) {
LOG("Client %d no longer exists", (int) exited_pid);
running_clients.erase(exited_pid);
}
/*
* Now check if any children exceeded their timeout. Kill them.
*/
struct timespec current_t;
clock_gettime(CLOCK_MONOTONIC, ¤t_t);
for (auto it = running_clients.begin(); it != running_clients.end(); ) {
auto timeout_t = it->second;
if (cmpTimespec(timeout_t, current_t) < 0) {
auto timeouted_pid = it->first;
LOG("Client %d gets killed due to timeout", (int) timeouted_pid);
/*
* We kill the client using SIGHUP. Since we installed a signal handler, and signal handlers
* are kept during fork(), this should be enough to end it.
* That is, unless an R package removes the signal handler. In that case, we'd need to keep
* tracking the process and force a SIGKILL if it refuses to exit.
*/
if (kill(timeouted_pid, SIGHUP) < 0) {
perror("kill() failed");
++it;
}
else {
// the postincrement of the iterator is important to avoid using an invalid iterator
running_clients.erase(it++);
}
}
else {
++it;
}
}
/*
* Wait for new connections.
*
* We may want to limit the amount of clients running at the same time.
*/
if (running_clients.size() > 10) {
std::this_thread::sleep_for(std::chrono::milliseconds(5000));
continue;
}
struct pollfd pollfds[1];
pollfds[0].fd = listen_fd;
pollfds[0].events = POLLIN;
int poll_res = poll(pollfds, /* count = */ 1, /* timeout in ms = */ 5000);
if (poll_res < 0) {
perror("poll() failed");
exit(1);
}
/*
* If no new connection is made within 5 seconds, we repeat the loop and check
* for finished or timeouted children again.
*/
if (poll_res == 0)
continue;
if ((pollfds[0].revents & POLLIN) == 0)
continue;
struct sockaddr_un client_addr;
socklen_t client_addr_len = sizeof(client_addr);
int client_fd = accept(listen_fd, (struct sockaddr *) &client_addr, &client_addr_len);
if (client_fd < 0) {
if (errno == EAGAIN || errno == EWOULDBLOCK)
continue;
perror("accept() failed");
exit(1);
}
// fork
pid_t pid = fork();
if (pid < 0) {
perror("fork() failed");
exit(1);
}
if (pid == 0) {
/*
* This is the child process.
*
* If the child process needs to drop any privileges the server may have had,
* this is an excellent time to do so.
* Whether it's a chroot, seccomp-bpf or a MAC framework like SELinux or AppArmor.
*
* Note that neither is an excuse to run the parent process unrestricted; creating
* a new restricted user for the server seems wise.
*/
close(listen_fd);
LOG("Client starting");
auto start_c = clock();
struct timespec start_t;
clock_gettime(CLOCK_MONOTONIC, &start_t);
try {
BinaryStream stream(client_fd, client_fd);
RInsideServer ris(stream, R, *Rcallbacks);
ris.run();
}
catch (const std::exception &e) {
LOG("Exception: %s", e.what());
}
auto end_c = clock();
struct timespec end_t;
clock_gettime(CLOCK_MONOTONIC, &end_t);
double c = (double) (end_c - start_c) / CLOCKS_PER_SEC;
double t = (double) (end_t.tv_sec - start_t.tv_sec) + (double) (end_t.tv_nsec - start_t.tv_nsec) / 1000000000;
LOG("Client finished, %.3fs real, %.3fs CPU", t, c);
exit(0);
}
else {
// This is the parent process
close(client_fd);
struct timespec timeout_t;
clock_gettime(CLOCK_MONOTONIC, &timeout_t);
timeout_t.tv_sec += TIMEOUT_SECONDS;
running_clients[pid] = timeout_t;
}
}
}
|