Merge pull request #203 from takeoverjp/feature-garbage-collect-inodeproc

Issue: #96 - Garbage collect inodeproc on each ui refresh.

doc/nethogs.8

@@ -19,6 +19,7 @@ nethogs \- Net top tool grouping bandwidth per process
 .RB [ "\-l" ]
 .RB [ "\-f filter" ]
 .RB [ "\-C" ]
+.RB [ "\-g period" ]
 .RI [device(s)]
 .SH DESCRIPTION
 NetHogs is a small 'net top' tool. Instead of breaking the traffic down per protocol or per subnet, like most such tools do, it groups bandwidth by process - and does not rely on a special kernel module to be loaded. So if there's suddenly a lot of network traffic, you can fire up NetHogs and immediately see which PID is causing this, and if it's some kind of spinning process, kill it.
@@ -61,6 +62,9 @@ monitor all devices, even loopback/stopped ones.
 \fB-C\fP
 capture TCP and UDP.
 .TP
+\fB-g\fP
+garbage collection period in number of refresh. default is 50.
+.TP
 \fB-f\fP
 EXPERIMENTAL: specify string pcap filter (like tcpdump). This may be removed or changed in a future version.
 .TP

src/inode2prog.cpp

@@ -20,6 +20,7 @@
  *
  */
 
+#include <algorithm>
 #include <cerrno>
 #include <climits>
 #include <cstdio>
@@ -30,6 +31,7 @@
 #include <fcntl.h>
 #include <iostream>
 #include <map>
+#include <set>
 #include <sstream>
 #include <string>
 #include <sys/stat.h>
@@ -224,6 +226,73 @@ void get_info_for_pid(const char *pid) {
   closedir(dir);
 }
 
+static quad_t get_ms() {
+  struct timespec ts;
+  clock_gettime(CLOCK_MONOTONIC, &ts);
+  return static_cast<quad_t>(ts.tv_sec) * 1000 + ts.tv_nsec / 1000000;
+}
+
+static void get_pids(std::set<pid_t> *pids) {
+  DIR *proc = opendir("/proc");
+  if (proc == 0) {
+    std::cerr << "Error reading /proc, needed to get pid set" << std::endl;
+    exit(1);
+  }
+  dirent *entry;
+
+  while ((entry = readdir(proc))) {
+    if (entry->d_type != DT_DIR)
+      continue;
+
+    if (!is_number(entry->d_name))
+      continue;
+
+    pids->insert(str2int(entry->d_name));
+  }
+  closedir(proc);
+}
+
+void garbage_collect_inodeproc() {
+  static quad_t last_ms = 0;
+  quad_t start_ms = 0;
+  if (bughuntmode) {
+    start_ms = get_ms();
+    if (last_ms) {
+      std::cout << "PERF: GC interval: " << start_ms - last_ms << "[ms]"
+                << std::endl;
+    }
+  }
+
+  std::set<pid_t> pids;
+  get_pids(&pids);
+  if (pids.size() == 0) {
+    return;
+  }
+
+  for (std::map<unsigned long, prg_node *>::iterator it = inodeproc.begin();
+       it != inodeproc.end();) {
+    if (!it->second || pids.find(it->second->pid) != pids.end()) {
+      ++it;
+      continue;
+    }
+
+    if (bughuntmode) {
+      std::cout << "GC prg_node (inode=" << it->first
+                << ", pid=" << it->second->pid
+                << ", cmdline=" << it->second->cmdline.c_str() << ")"
+                << std::endl;
+    }
+    delete it->second;
+    inodeproc.erase(it++);
+  }
+
+  if (bughuntmode) {
+    last_ms = get_ms();
+    std::cout << "PERF: GC proctime: " << last_ms - start_ms << "[ms]"
+              << std::endl;
+  }
+}
+
 /* updates the `inodeproc' inode-to-prg_node mapping
  * for all processes in /proc */
 void reread_mapping() {

src/inode2prog.h

@@ -41,4 +41,6 @@ void prg_cache_clear();
 // reread the inode-to-prg_node-mapping
 void reread_mapping();
 
+void garbage_collect_inodeproc();
+
 #endif

src/main.cpp

@@ -46,6 +46,8 @@ static void help(bool iserror) {
   output << "		-a : monitor all devices, even loopback/stopped "
             "ones.\n";
   output << "		-C : capture TCP and UDP.\n";
+  output << "		-g : garbage collection period in number of refresh. "
+            "default is 50.\n";
   output << "		-f : EXPERIMENTAL: specify string pcap filter (like "
             "tcpdump)."
             " This may be removed or changed in a future version.\n";
@@ -140,9 +142,10 @@ int main(int argc, char **argv) {
   int promisc = 0;
   bool all = false;
   char *filter = NULL;
+  int garbage_collection_period = 50;
 
   int opt;
-  while ((opt = getopt(argc, argv, "Vhbtpsd:v:c:laf:C")) != -1) {
+  while ((opt = getopt(argc, argv, "Vhbtpsd:v:c:laf:Cg:")) != -1) {
     switch (opt) {
     case 'V':
       versiondisplay();
@@ -184,6 +187,9 @@ int main(int argc, char **argv) {
     case 'C':
       catchall = true;
       break;
+    case 'g':
+      garbage_collection_period = (time_t)atoi(optarg);
+      break;
     default:
       help(true);
       exit(EXIT_FAILURE);
@@ -294,6 +300,7 @@ int main(int argc, char **argv) {
   struct dpargs *userdata = (dpargs *)malloc(sizeof(struct dpargs));
 
   // Main loop:
+  int refresh_count = 0;
   while (1) {
     bool packets_read = false;
 
@@ -322,6 +329,12 @@ int main(int argc, char **argv) {
         ui_tick();
       }
       do_refresh();
+      ++refresh_count;
+
+      if ((garbage_collection_period > 0) &&
+          (refresh_count % garbage_collection_period == 0)) {
+        garbage_collect_processes();
+      }
     }
 
     // if not packets, do a select() until next packet

src/process.cpp

@@ -424,3 +424,5 @@ void remove_timed_out_processes() {
     previousproc = curproc;
   }
 }
+
+void garbage_collect_processes() { garbage_collect_inodeproc(); }

src/process.h

@@ -145,4 +145,6 @@ void procclean();
 
 void remove_timed_out_processes();
 
+void garbage_collect_processes();
+
 #endif