/**
* collectd - src/processes.c
* Copyright (C) 2005 Lyonel Vincent
+ * Copyright (C) 2006 Florian Forster (Mach code)
*
* This program is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
* Florian octo Forster <octo at verplant.org>
**/
-#include "processes.h"
+#include "collectd.h"
+#include "common.h"
+#include "plugin.h"
+#include "utils_debug.h"
+#include "configfile.h"
+
+/* Include header files for the mach system, if they exist.. */
+#if HAVE_THREAD_INFO
+# if HAVE_MACH_MACH_INIT_H
+# include <mach/mach_init.h>
+# endif
+# if HAVE_MACH_HOST_PRIV_H
+# include <mach/host_priv.h>
+# endif
+# if HAVE_MACH_MACH_ERROR_H
+# include <mach/mach_error.h>
+# endif
+# if HAVE_MACH_MACH_HOST_H
+# include <mach/mach_host.h>
+# endif
+# if HAVE_MACH_MACH_PORT_H
+# include <mach/mach_port.h>
+# endif
+# if HAVE_MACH_MACH_TYPES_H
+# include <mach/mach_types.h>
+# endif
+# if HAVE_MACH_MESSAGE_H
+# include <mach/message.h>
+# endif
+# if HAVE_MACH_PROCESSOR_SET_H
+# include <mach/processor_set.h>
+# endif
+# if HAVE_MACH_TASK_H
+# include <mach/task.h>
+# endif
+# if HAVE_MACH_THREAD_ACT_H
+# include <mach/thread_act.h>
+# endif
+# if HAVE_MACH_VM_REGION_H
+# include <mach/vm_region.h>
+# endif
+# if HAVE_MACH_VM_MAP_H
+# include <mach/vm_map.h>
+# endif
+# if HAVE_MACH_VM_PROT_H
+# include <mach/vm_prot.h>
+# endif
+/* #endif HAVE_THREAD_INFO */
+
+#elif KERNEL_LINUX
+# if HAVE_LINUX_CONFIG_H
+# include <linux/config.h>
+# endif
+# ifndef CONFIG_HZ
+# define CONFIG_HZ 100
+# endif
+#endif /* KERNEL_LINUX */
-#if COLLECT_PROCESSES
#define MODULE_NAME "processes"
-#include "common.h"
-#include "plugin.h"
+#if HAVE_THREAD_INFO || KERNEL_LINUX
+# define PROCESSES_HAVE_READ 1
+#else
+# define PROCESSES_HAVE_READ 0
+#endif
-static char *ps_file = "processes.rrd";
+#define BUFSIZE 256
+
+static char *processes_file = "processes.rrd";
+static char *processes_ds_def[] =
+{
+ "DS:running:GAUGE:"COLLECTD_HEARTBEAT":0:65535",
+ "DS:sleeping:GAUGE:"COLLECTD_HEARTBEAT":0:65535",
+ "DS:zombies:GAUGE:"COLLECTD_HEARTBEAT":0:65535",
+ "DS:stopped:GAUGE:"COLLECTD_HEARTBEAT":0:65535",
+ "DS:paging:GAUGE:"COLLECTD_HEARTBEAT":0:65535",
+ "DS:blocked:GAUGE:"COLLECTD_HEARTBEAT":0:65535",
+ NULL
+};
+static int processes_ds_num = 6;
-static char *ds_def[] =
+static char *config_keys[] =
{
- "DS:running:GAUGE:25:0:65535",
- "DS:sleeping:GAUGE:25:0:65535",
- "DS:zombies:GAUGE:25:0:65535",
- "DS:stopped:GAUGE:25:0:65535",
- "DS:paging:GAUGE:25:0:65535",
- "DS:blocked:GAUGE:25:0:65535",
+ "CollectName",
NULL
};
-static int ds_num = 6;
+static int config_keys_num = 1;
-void ps_init (void)
+typedef struct procstat
{
+#define PROCSTAT_NAME_LEN 256
+ char name[PROCSTAT_NAME_LEN];
+ unsigned int num_proc;
+ unsigned int num_lwp;
+ unsigned long vmem_rss;
+ unsigned long vmem_minflt;
+ unsigned long vmem_majflt;
+ unsigned long long cpu_user;
+ unsigned long long cpu_system;
+ struct procstat *next;
+} procstat_t;
+
+static procstat_t *list_head_g = NULL;
+
+#if HAVE_THREAD_INFO
+static mach_port_t port_host_self;
+static mach_port_t port_task_self;
+
+static processor_set_name_array_t pset_list;
+static mach_msg_type_number_t pset_list_len;
+/* #endif HAVE_THREAD_INFO */
+
+#elif KERNEL_LINUX
+static long pagesize_g;
+#endif /* KERNEL_LINUX */
+
+static procstat_t *ps_list_append (procstat_t *list, const char *name)
+{
+ procstat_t *new;
+ procstat_t *ptr;
+
+ if ((new = (procstat_t *) malloc (sizeof (procstat_t))) == NULL)
+ return (NULL);
+ memset (new, 0, sizeof (procstat_t));
+ strncpy (new->name, name, PROCSTAT_NAME_LEN);
+
+ for (ptr = list; ptr != NULL; ptr = ptr->next)
+ if (ptr->next == NULL)
+ break;
+
+ if (ptr != NULL)
+ ptr->next = new;
+
+ return (new);
}
-void ps_write (char *host, char *inst, char *val)
+static void ps_list_add (procstat_t *list, procstat_t *entry)
{
- rrd_update_file (host, ps_file, val, ds_def, ds_num);
+ procstat_t *ptr;
+
+ ptr = list;
+ while ((ptr != NULL) && (strcmp (ptr->name, entry->name) != 0))
+ ptr = ptr->next;
+
+ if (ptr == NULL)
+ return;
+
+ ptr->num_proc += entry->num_proc;
+ ptr->num_lwp += entry->num_lwp;
+ ptr->vmem_rss += entry->vmem_rss;
+ ptr->vmem_minflt += entry->vmem_minflt;
+ ptr->vmem_majflt += entry->vmem_majflt;
+ ptr->cpu_user += entry->cpu_user;
+ ptr->cpu_system += entry->cpu_system;
}
-#define BUFSIZE 256
-void ps_submit (unsigned int running,
- unsigned int sleeping,
- unsigned int zombies,
- unsigned int stopped,
- unsigned int paging,
- unsigned int blocked)
+static void ps_list_reset (procstat_t *ps)
+{
+ while (ps != NULL)
+ {
+ ps->num_proc = 0;
+ ps->num_lwp = 0;
+ ps->vmem_rss = 0;
+ ps->vmem_minflt = 0;
+ ps->vmem_majflt = 0;
+ ps->cpu_user = 0;
+ ps->cpu_system = 0;
+ ps = ps->next;
+ }
+}
+
+static int ps_config (char *key, char *value)
+{
+ if (strcasecmp (key, "CollectName") == 0)
+ {
+ procstat_t *entry;
+
+ entry = ps_list_append (list_head_g, value);
+ if (entry == NULL)
+ {
+ syslog (LOG_ERR, "processes plugin: ps_list_append failed.");
+ return (1);
+ }
+ if (list_head_g == NULL)
+ list_head_g = entry;
+ }
+ else
+ {
+ return (-1);
+ }
+
+ return (0);
+}
+
+static void ps_init (void)
+{
+#if HAVE_THREAD_INFO
+ kern_return_t status;
+
+ port_host_self = mach_host_self ();
+ port_task_self = mach_task_self ();
+
+ if (pset_list != NULL)
+ {
+ vm_deallocate (port_task_self,
+ (vm_address_t) pset_list,
+ pset_list_len * sizeof (processor_set_t));
+ pset_list = NULL;
+ pset_list_len = 0;
+ }
+
+ if ((status = host_processor_sets (port_host_self,
+ &pset_list,
+ &pset_list_len)) != KERN_SUCCESS)
+ {
+ syslog (LOG_ERR, "host_processor_sets failed: %s\n",
+ mach_error_string (status));
+ pset_list = NULL;
+ pset_list_len = 0;
+ return;
+ }
+/* #endif HAVE_THREAD_INFO */
+
+#elif KERNEL_LINUX
+ pagesize_g = sysconf(_SC_PAGESIZE);
+#endif /* KERNEL_LINUX */
+
+ return;
+}
+
+static void ps_write (char *host, char *inst, char *val)
+{
+ rrd_update_file (host, processes_file, val,
+ processes_ds_def, processes_ds_num);
+}
+
+#if PROCESSES_HAVE_READ
+static void ps_submit (int running,
+ int sleeping,
+ int zombies,
+ int stopped,
+ int paging,
+ int blocked)
{
char buf[BUFSIZE];
- if (snprintf (buf, BUFSIZE, "%u:%u:%u:%u:%u:%u:%u",
+ if (snprintf (buf, BUFSIZE, "%u:%i:%i:%i:%i:%i:%i",
(unsigned int) curtime,
running, sleeping, zombies, stopped, paging,
blocked) >= BUFSIZE)
return;
+ DBG ("running = %i; sleeping = %i; zombies = %i; stopped = %i; paging = %i; blocked = %i;",
+ running, sleeping, zombies, stopped, paging, blocked);
+
plugin_submit (MODULE_NAME, "-", buf);
}
-void ps_read (void)
+static void ps_submit_proc (procstat_t *ps)
{
-#ifdef KERNEL_LINUX
- unsigned int running, sleeping, zombies, stopped, paging, blocked;
+ if (ps == NULL)
+ return;
- char buf[BUFSIZE];
- char filename[20]; /* need 17 bytes */
- char *fields[256];
+ DBG ("name = %s; num_proc = %i; num_lwp = %i; vmem_rss = %i; "
+ "vmem_minflt = %i; vmem_majflt = %i; "
+ "cpu_user = %i; cpu_system = %i;",
+ ps->name, ps->num_proc, ps->num_lwp, ps->vmem_rss,
+ ps->vmem_minflt, ps->vmem_majflt, ps->cpu_user,
+ ps->cpu_system);
+}
+#if KERNEL_LINUX
+static int *ps_read_tasks (int pid)
+{
+ int *list = NULL;
+ int list_size = 1; /* size of allocated space, in elements */
+ int list_len = 0; /* number of currently used elements */
+
+ char dirname[64];
+ DIR *dh;
struct dirent *ent;
- DIR *proc;
- FILE *fh;
- running = sleeping = zombies = stopped = paging = blocked = 0;
+ snprintf (dirname, 64, "/proc/%i/task", pid);
+ dirname[63] = '\0';
- if ((proc = opendir ("/proc")) == NULL)
+ if ((dh = opendir (dirname)) == NULL)
{
- syslog (LOG_ERR, "Cannot open `/proc': %s", strerror (errno));
- return;
+ syslog (LOG_NOTICE, "processes plugin: Failed to open directory `%s'",
+ dirname);
+ return (NULL);
}
- while ((ent = readdir (proc)) != NULL)
+ while ((ent = readdir (dh)) != NULL)
{
if (!isdigit (ent->d_name[0]))
continue;
- if (snprintf (filename, 20, "/proc/%s/stat", ent->d_name) >= 20)
- continue;
+ if ((list_len + 1) >= list_size)
+ {
+ int *new_ptr;
+ int new_size = 2 * list_size;
+ /* Comes in sizes: 2, 4, 8, 16, ... */
+
+ new_ptr = (int *) realloc (list, (size_t) (sizeof (int) * new_size));
+ if (new_ptr == NULL)
+ {
+ if (list != NULL)
+ free (list);
+ syslog (LOG_ERR, "processes plugin: "
+ "Failed to allocate more memory.");
+ return (NULL);
+ }
+
+ list = new_ptr;
+ list_size = new_size;
+
+ memset (list + list_len, 0, sizeof (int) * (list_size - list_len));
+ }
+
+ list[list_len] = atoi (ent->d_name);
+ if (list[list_len] != 0)
+ list_len++;
+ }
+
+ closedir (dh);
+
+ assert (list_len < list_size);
+ assert (list[list_len] == 0);
+
+ return (list);
+}
- if ((fh = fopen (filename, "r")) == NULL)
+int ps_read_process (int pid, procstat_t *ps, char *state)
+{
+ char filename[64];
+ char buffer[1024];
+ FILE *fh;
+
+ char *fields[64];
+ char fields_len;
+
+ int *tasks;
+ int i;
+
+ int ppid;
+ int name_len;
+
+ memset (ps, 0, sizeof (procstat_t));
+
+ snprintf (filename, 64, "/proc/%i/stat", pid);
+ filename[63] = '\0';
+
+ if ((fh = fopen (filename, "r")) == NULL)
+ return (-1);
+
+ if (fgets (buffer, 1024, fh) == NULL)
+ {
+ fclose (fh);
+ return (-1);
+ }
+
+ fclose (fh);
+
+ fields_len = strsplit (buffer, fields, 64);
+ if (fields_len < 24)
+ {
+ DBG ("`%s' has only %i fields..",
+ filename, fields_len);
+ return (-1);
+ }
+ else if (fields_len != 41)
+ {
+ DBG ("WARNING: (fields_len = %i) != 41", fields_len);
+ }
+
+ /* copy the name, strip brackets in the process */
+ name_len = strlen (fields[1]) - 2;
+ if ((fields[1][0] != '(') || (fields[1][name_len + 1] != ')'))
+ {
+ DBG ("No brackets found in process name: `%s'", fields[1]);
+ return (-1);
+ }
+ fields[1] = fields[1] + 1;
+ fields[1][name_len] = '\0';
+ strncpy (ps->name, fields[1], PROCSTAT_NAME_LEN);
+
+ ppid = atoi (fields[3]);
+
+ if ((tasks = ps_read_tasks (pid)) == NULL)
+ {
+ DBG ("ps_read_tasks (%i) failed.", pid);
+ return (-1);
+ }
+
+ *state = '\0';
+ ps->num_lwp = 0;
+ ps->num_proc = 1;
+ for (i = 0; tasks[i] != 0; i++)
+ ps->num_lwp++;
+
+ free (tasks);
+ tasks = NULL;
+
+ /* Leave the rest at zero if this is only an LWP */
+ if (ps->num_proc == 0)
+ {
+ DBG ("This is only an LWP: pid = %i; name = %s;",
+ pid, ps->name);
+ return (0);
+ }
+
+ ps->vmem_minflt = atol (fields[9]);
+ ps->vmem_majflt = atol (fields[11]);
+ ps->cpu_user = atoll (fields[13]);
+ ps->cpu_system = atoll (fields[14]);
+ ps->vmem_rss = atol (fields[23]);
+
+ /* Convert jiffies to useconds */
+ ps->cpu_user = ps->cpu_user * 1000000 / CONFIG_HZ;
+ ps->cpu_system = ps->cpu_system * 1000000 / CONFIG_HZ;
+ ps->vmem_rss = ps->vmem_rss * pagesize_g;
+
+ *state = fields[2][0];
+
+ /* success */
+ return (0);
+} /* int ps_read_process (...) */
+#endif /* KERNEL_LINUX */
+
+static void ps_read (void)
+{
+#if HAVE_THREAD_INFO
+ kern_return_t status;
+
+ int pset;
+ processor_set_t port_pset_priv;
+
+ int task;
+ task_array_t task_list;
+ mach_msg_type_number_t task_list_len;
+
+ int thread;
+ thread_act_array_t thread_list;
+ mach_msg_type_number_t thread_list_len;
+ thread_basic_info_data_t thread_data;
+ mach_msg_type_number_t thread_data_len;
+
+ int running = 0;
+ int sleeping = 0;
+ int zombies = 0;
+ int stopped = 0;
+ int blocked = 0;
+
+ /*
+ * The Mach-concept is a little different from the traditional UNIX
+ * concept: All the work is done in threads. Threads are contained in
+ * `tasks'. Therefore, `task status' doesn't make much sense, since
+ * it's actually a `thread status'.
+ * Tasks are assigned to sets of processors, so that's where you go to
+ * get a list.
+ */
+ for (pset = 0; pset < pset_list_len; pset++)
+ {
+ if ((status = host_processor_set_priv (port_host_self,
+ pset_list[pset],
+ &port_pset_priv)) != KERN_SUCCESS)
{
- syslog (LOG_ERR, "Cannot open `%s': %s", filename, strerror (errno));
+ syslog (LOG_ERR, "host_processor_set_priv failed: %s\n",
+ mach_error_string (status));
continue;
}
- if (fgets (buf, BUFSIZE, fh) == NULL)
+ if ((status = processor_set_tasks (port_pset_priv,
+ &task_list,
+ &task_list_len)) != KERN_SUCCESS)
{
- fclose (fh);
+ syslog (LOG_ERR, "processor_set_tasks failed: %s\n",
+ mach_error_string (status));
+ mach_port_deallocate (port_task_self, port_pset_priv);
continue;
}
- fclose (fh);
+ for (task = 0; task < task_list_len; task++)
+ {
+ status = task_threads (task_list[task], &thread_list,
+ &thread_list_len);
+ if (status != KERN_SUCCESS)
+ {
+ /* Apple's `top' treats this case a zombie. It
+ * makes sense to some extend: A `zombie'
+ * thread is nonsense, since the task/process
+ * is dead. */
+ zombies++;
+ DBG ("task_threads failed: %s",
+ mach_error_string (status));
+ if (task_list[task] != port_task_self)
+ mach_port_deallocate (port_task_self,
+ task_list[task]);
+ continue; /* with next task_list */
+ }
+
+ for (thread = 0; thread < thread_list_len; thread++)
+ {
+ thread_data_len = THREAD_BASIC_INFO_COUNT;
+ status = thread_info (thread_list[thread],
+ THREAD_BASIC_INFO,
+ (thread_info_t) &thread_data,
+ &thread_data_len);
+ if (status != KERN_SUCCESS)
+ {
+ syslog (LOG_ERR, "thread_info failed: %s\n",
+ mach_error_string (status));
+ if (task_list[task] != port_task_self)
+ mach_port_deallocate (port_task_self,
+ thread_list[thread]);
+ continue; /* with next thread_list */
+ }
+
+ switch (thread_data.run_state)
+ {
+ case TH_STATE_RUNNING:
+ running++;
+ break;
+ case TH_STATE_STOPPED:
+ /* What exactly is `halted'? */
+ case TH_STATE_HALTED:
+ stopped++;
+ break;
+ case TH_STATE_WAITING:
+ sleeping++;
+ break;
+ case TH_STATE_UNINTERRUPTIBLE:
+ blocked++;
+ break;
+ /* There is no `zombie' case here,
+ * since there are no zombie-threads.
+ * There's only zombie tasks, which are
+ * handled above. */
+ default:
+ syslog (LOG_WARNING,
+ "Unknown thread status: %s",
+ thread_data.run_state);
+ break;
+ } /* switch (thread_data.run_state) */
+
+ if (task_list[task] != port_task_self)
+ {
+ status = mach_port_deallocate (port_task_self,
+ thread_list[thread]);
+ if (status != KERN_SUCCESS)
+ syslog (LOG_ERR, "mach_port_deallocate failed: %s",
+ mach_error_string (status));
+ }
+ } /* for (thread_list) */
+
+ if ((status = vm_deallocate (port_task_self,
+ (vm_address_t) thread_list,
+ thread_list_len * sizeof (thread_act_t)))
+ != KERN_SUCCESS)
+ {
+ syslog (LOG_ERR, "vm_deallocate failed: %s",
+ mach_error_string (status));
+ }
+ thread_list = NULL;
+ thread_list_len = 0;
+
+ /* Only deallocate the task port, if it isn't our own.
+ * Don't know what would happen in that case, but this
+ * is what Apple's top does.. ;) */
+ if (task_list[task] != port_task_self)
+ {
+ status = mach_port_deallocate (port_task_self,
+ task_list[task]);
+ if (status != KERN_SUCCESS)
+ syslog (LOG_ERR, "mach_port_deallocate failed: %s",
+ mach_error_string (status));
+ }
+ } /* for (task_list) */
+
+ if ((status = vm_deallocate (port_task_self,
+ (vm_address_t) task_list,
+ task_list_len * sizeof (task_t))) != KERN_SUCCESS)
+ {
+ syslog (LOG_ERR, "vm_deallocate failed: %s",
+ mach_error_string (status));
+ }
+ task_list = NULL;
+ task_list_len = 0;
+
+ if ((status = mach_port_deallocate (port_task_self, port_pset_priv))
+ != KERN_SUCCESS)
+ {
+ syslog (LOG_ERR, "mach_port_deallocate failed: %s",
+ mach_error_string (status));
+ }
+ } /* for (pset_list) */
+
+ ps_submit (running, sleeping, zombies, stopped, -1, blocked);
+/* #endif HAVE_THREAD_INFO */
+
+#elif KERNEL_LINUX
+ int running = 0;
+ int sleeping = 0;
+ int zombies = 0;
+ int stopped = 0;
+ int paging = 0;
+ int blocked = 0;
+
+ struct dirent *ent;
+ DIR *proc;
+ int pid;
+
+ int status;
+ procstat_t ps;
+ char state;
+
+ procstat_t *ps_ptr;
+
+ running = sleeping = zombies = stopped = paging = blocked = 0;
+ ps_list_reset (list_head_g);
+
+ if ((proc = opendir ("/proc")) == NULL)
+ {
+ syslog (LOG_ERR, "Cannot open `/proc': %s", strerror (errno));
+ return;
+ }
+
+ while ((ent = readdir (proc)) != NULL)
+ {
+ if (!isdigit (ent->d_name[0]))
+ continue;
- if (strsplit (buf, fields, 256) < 3)
+ if ((pid = atoi (ent->d_name)) < 1)
+ continue;
+
+ status = ps_read_process (pid, &ps, &state);
+ if (status != 0)
+ {
+ DBG ("ps_read_process failed: %i", status);
continue;
+ }
- switch (fields[2][0])
+ switch (state)
{
case 'R': running++; break;
case 'S': sleeping++; break;
case 'T': stopped++; break;
case 'W': paging++; break;
}
+
+ if (list_head_g != NULL)
+ ps_list_add (list_head_g, &ps);
}
- closedir(proc);
+ closedir (proc);
ps_submit (running, sleeping, zombies, stopped, paging, blocked);
-#endif /* defined(KERNEL_LINUX) */
+
+ for (ps_ptr = list_head_g; ps_ptr != NULL; ps_ptr = ps_ptr->next)
+ ps_submit_proc (ps_ptr);
+#endif /* KERNEL_LINUX */
}
-#undef BUFSIZE
+#else
+# define ps_read NULL
+#endif /* PROCESSES_HAVE_READ */
void module_register (void)
{
plugin_register (MODULE_NAME, ps_init, ps_read, ps_write);
+ cf_register (MODULE_NAME, ps_config, config_keys, config_keys_num);
}
+#undef BUFSIZE
#undef MODULE_NAME
-#endif /* COLLECT_PROCESSES */