improve list traversal performance and fix fd leaks

Author: Chris Morin

src/bootchart.c

@@ -41,7 +41,6 @@
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
-#include <sys/resource.h>
 #include <sys/uio.h>
 #include <time.h>
 #include <unistd.h>
@@ -323,7 +322,6 @@ int main(int argc, char *argv[]) {
         time_t t = 0;
         int r, samples;
         struct ps_struct *ps;
-        struct rlimit rlim;
         struct list_sample_data *head;
         struct sigaction sig = {
                 .sa_handler = signal_handler,
@@ -351,10 +349,6 @@ int main(int argc, char *argv[]) {
         }
         argv[0][0] = '@';
 
-        rlim.rlim_cur = 4096;
-        rlim.rlim_max = 4096;
-        (void) setrlimit(RLIMIT_NOFILE, &rlim);
-
         schfd = open("/proc/sys/kernel/sched_schedstats", O_WRONLY);
         if (schfd >= 0) {
                 write(schfd, "1\n", 2);
@@ -466,8 +460,8 @@ int main(int argc, char *argv[]) {
 
         /* do some cleanup, close fd's */
         ps = ps_first;
-        while (ps->next_ps) {
-                ps = ps->next_ps;
+        while (ps->next_running) {
+                ps = ps->next_running;
                 ps->schedstat = safe_close(ps->schedstat);
                 ps->sched = safe_close(ps->sched);
                 ps->smaps = safe_fclose(ps->smaps);

src/bootchart.h

@@ -62,10 +62,11 @@ struct list_sample_data {
 
 /* process info */
 struct ps_struct {
-        struct ps_struct *next_ps;    /* SLL pointer */
-        struct ps_struct *parent;     /* ppid ref */
-        struct ps_struct *children;   /* children */
-        struct ps_struct *next;       /* siblings */
+        struct ps_struct *next_ps;      /* SLL pointer */
+        struct ps_struct *next_running; /* currently running */
+        struct ps_struct *parent;       /* ppid ref */
+        struct ps_struct *children;     /* children */
+        struct ps_struct *next;         /* siblings */
 
         /* must match - otherwise it's a new process with same PID */
         char name[256];
@@ -78,6 +79,9 @@ struct ps_struct {
         int schedstat;
         FILE *smaps;
 
+        /* used to garbage collect running process list*/
+        bool still_running;
+
         /* pointers to first/last seen timestamps */
         struct ps_sched_struct *first;
         struct ps_sched_struct *last;

src/store.c

@@ -96,6 +96,29 @@ static int pid_cmdline_strscpy(int procfd, char *buffer, size_t buf_len, int pid
         return 0;
 }
 
+static void garbage_collect_dead_processes(struct ps_struct *ps_first) {
+        struct ps_struct *ps;
+        struct ps_struct *ps_next;
+
+        ps = ps_first;
+        while ((ps_next = ps->next_running)) {
+                if (!ps_next->still_running) {
+                        /* close the stream and fds */
+                        ps_next->schedstat = safe_close(ps_next->schedstat);
+                        ps_next->sched = safe_close(ps_next->sched);
+                        if (ps->smaps)
+                                fclose(ps->smaps);
+
+                        ps->next_running = ps_next->next_running;
+                } else {
+                        ps = ps_next;
+                }
+                /* this resets the flag for both running and dead processes*/
+                ps_next->still_running = false;
+        }
+}
+
+
 int log_sample(DIR *proc,
                int sample,
                struct ps_struct *ps_first,
@@ -221,8 +244,8 @@ int log_sample(DIR *proc,
                         continue;
 
                 ps = ps_first;
-                while (ps->next_ps) {
-                        ps = ps->next_ps;
+                while (ps->next_running) {
+                        ps = ps->next_running;
                         if (ps->pid == pid)
                                 break;
                 }
@@ -233,7 +256,7 @@ int log_sample(DIR *proc,
                         char t[32];
                         struct ps_struct *parent;
 
-                        ps->next_ps = new0(struct ps_struct, 1);
+                        ps->next_ps = ps->next_running = new0(struct ps_struct, 1);
                         if (!ps->next_ps)
                                 return log_oom();
 
@@ -371,13 +394,8 @@ int log_sample(DIR *proc,
                 }
 
                 s = pread(ps->schedstat, buf, sizeof(buf) - 1, 0);
-                if (s <= 0) {
-                        /* clean up our file descriptors - assume that the process exited */
-                        close(ps->schedstat);
-                        ps->schedstat = -1;
-                        ps->sched = safe_close(ps->sched);
+                if (s <= 0)
                         continue;
-                }
 
                 buf[s] = '\0';
 
@@ -533,17 +551,13 @@ int log_sample(DIR *proc,
                         }
 
                         s = pread(ps->sched, buf, sizeof(buf) - 1, 0);
-                        if (s <= 0) {
-                                /* clean up file descriptors */
-                                ps->sched = safe_close(ps->sched);
-                                ps->schedstat = safe_close(ps->schedstat);
-                                goto no_sched2;
-                        }
+                        if (s <= 0)
+                                continue;
 
                         buf[s] = '\0';
 
                         if (!sscanf(buf, "%s %*s %*s", key))
-                                goto no_sched2;
+                                continue;
 
                         strscpy(ps->name, sizeof(ps->name), key);
 
@@ -552,7 +566,10 @@ int log_sample(DIR *proc,
                         if (arg_show_cmdline)
                                 pid_cmdline_strscpy(procfd, ps->name, sizeof(ps->name), pid);
                 }
+                ps->still_running = true;
         }
 
+        garbage_collect_dead_processes(ps_first);
+
         return 0;
 }