-// TODO
-//
-// 1. disable RESET_OR_OVERFLOW check in charts
-#ifdef HAVE_CONFIG_H
-#include <config.h>
-#endif
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-#include <time.h>
-#include <unistd.h>
-#include <sys/types.h>
-#include <sys/time.h>
-#include <sys/wait.h>
-#include <sys/stat.h>
-
-#include <sys/resource.h>
-#include <sys/stat.h>
-
-#include <errno.h>
-#include <stdarg.h>
-#include <locale.h>
-#include <ctype.h>
-#include <fcntl.h>
-
-#include <malloc.h>
-#include <dirent.h>
-#include <arpa/inet.h>
-
-#include <sys/types.h>
-#include <pwd.h>
-#include <grp.h>
-
-#include "avl.h"
+/*
+ * netdata apps.plugin
+ * (C) Copyright 2016-2017 Costa Tsaousis <costa@tsaousis.gr>
+ * Released under GPL v3+
+ */
#include "common.h"
-#include "log.h"
-#include "procfile.h"
-#include "../config.h"
-#ifdef NETDATA_INTERNAL_CHECKS
-#include <sys/prctl.h>
+#ifdef __FreeBSD__
+#include <sys/user.h>
+#endif
+
+// ----------------------------------------------------------------------------
+// per O/S configuration
+
+// the minimum PID of the system
+// this is also the pid of the init process
+#define INIT_PID 1
+
+// if the way apps.plugin will work, will read the entire process list,
+// including the resource utilization of each process, instantly
+// set this to 1
+// when set to 0, apps.plugin builds a sort list of processes, in order
+// to process children processes, before parent processes
+#ifdef __FreeBSD__
+#define ALL_PIDS_ARE_READ_INSTANTLY 1
+#else
+#define ALL_PIDS_ARE_READ_INSTANTLY 0
#endif
+// ----------------------------------------------------------------------------
+// string lengths
+
#define MAX_COMPARE_NAME 100
#define MAX_NAME 100
#define MAX_CMDLINE 1024
-long processors = 1;
-long pid_max = 32768;
-int debug = 0;
-int update_every = 1;
-unsigned long long file_counter = 0;
-int proc_pid_cmdline_is_needed = 0;
+// ----------------------------------------------------------------------------
+// the rates we are going to send to netdata will have this detail a value of:
+// - 1 will send just integer parts to netdata
+// - 100 will send 2 decimal points
+// - 1000 will send 3 decimal points
+// etc.
+#define RATES_DETAIL 10000ULL
-char *host_prefix = "";
-char *config_dir = CONFIG_DIR;
-#ifdef NETDATA_INTERNAL_CHECKS
// ----------------------------------------------------------------------------
-// memory debugger
-// do not use in production systems - it mis-aligns allocated memory
+// to avoid reallocating too frequently, we can increase the number of spare
+// file descriptors used by processes.
+// IMPORTANT:
+// having a lot of spares, increases the CPU utilization of the plugin.
+#define MAX_SPARE_FDS 1
-struct allocations {
- size_t allocations;
- size_t allocated;
- size_t allocated_max;
-} allocations = { 0, 0, 0 };
-#define MALLOC_MARK (uint32_t)(0x0BADCAFE)
-#define MALLOC_PREFIX (sizeof(uint32_t) * 2)
-#define MALLOC_SUFFIX (sizeof(uint32_t))
-#define MALLOC_OVERHEAD (MALLOC_PREFIX + MALLOC_SUFFIX)
+// ----------------------------------------------------------------------------
+// command line options
-void *mark_allocation(void *allocated_ptr, size_t size_without_overheads) {
- uint32_t *real_ptr = (uint32_t *)allocated_ptr;
- real_ptr[0] = MALLOC_MARK;
- real_ptr[1] = (uint32_t) size_without_overheads;
+static int
+ debug = 0,
+ update_every = 1,
+ enable_guest_charts = 0,
+ enable_file_charts = 1,
+ enable_users_charts = 1,
+ enable_groups_charts = 1,
+ include_exited_childs = 1;
- uint32_t *end_ptr = (uint32_t *)(allocated_ptr + MALLOC_PREFIX + size_without_overheads);
- end_ptr[0] = MALLOC_MARK;
- // fprintf(stderr, "MEMORY_POINTER: Allocated at %p, returning %p.\n", allocated_ptr, (void *)(allocated_ptr + MALLOC_PREFIX));
+// will be changed to getenv(NETDATA_CONFIG_DIR) if it exists
+static char *config_dir = CONFIG_DIR;
- return allocated_ptr + MALLOC_PREFIX;
-}
+// ----------------------------------------------------------------------------
+// internal flags
+// handled in code (automatically set)
-void *check_allocation(const char *file, int line, const char *function, void *marked_ptr, size_t *size_without_overheads_ptr) {
- uint32_t *real_ptr = (uint32_t *)(marked_ptr - MALLOC_PREFIX);
+static int
+ show_guest_time = 0, // 1 when guest values are collected
+ show_guest_time_old = 0,
+ proc_pid_cmdline_is_needed = 0; // 1 when we need to read /proc/cmdline
- // fprintf(stderr, "MEMORY_POINTER: Checking pointer at %p, real %p for %s/%u@%s.\n", marked_ptr, (void *)(marked_ptr - MALLOC_PREFIX), function, line, file);
- if(real_ptr[0] != MALLOC_MARK) fatal("MEMORY: prefix MARK is not valid for %s/%u@%s.", function, line, file);
+// ----------------------------------------------------------------------------
+// internal counters
- size_t size = real_ptr[1];
+static size_t
+ global_iterations_counter = 1,
+ calls_counter = 0,
+ file_counter = 0;
- uint32_t *end_ptr = (uint32_t *)(marked_ptr + size);
- if(end_ptr[0] != MALLOC_MARK) fatal("MEMORY: suffix MARK of allocation with size %zu is not valid for %s/%u@%s.", size, function, line, file);
- if(size_without_overheads_ptr) *size_without_overheads_ptr = size;
+// ----------------------------------------------------------------------------
+// Normalization
+//
+// With normalization we lower the collected metrics by a factor to make them
+// match the total utilization of the system.
+// The discrepancy exists because apps.plugin needs some time to collect all
+// the metrics. This results in utilization that exceeds the total utilization
+// of the system.
+//
+// With normalization we align the per-process utilization, to the total of
+// the system. We first consume the exited children utilization and it the
+// collected values is above the total, we proportionally scale each reported
+// metric.
+
+// the total system time, as reported by /proc/stat
+static kernel_uint_t
+ global_utime = 0,
+ global_stime = 0,
+ global_gtime = 0;
+
+
+// the normalization ratios, as calculated by normalize_utilization()
+double utime_fix_ratio = 1.0,
+ stime_fix_ratio = 1.0,
+ gtime_fix_ratio = 1.0,
+ minflt_fix_ratio = 1.0,
+ majflt_fix_ratio = 1.0,
+ cutime_fix_ratio = 1.0,
+ cstime_fix_ratio = 1.0,
+ cgtime_fix_ratio = 1.0,
+ cminflt_fix_ratio = 1.0,
+ cmajflt_fix_ratio = 1.0;
- return real_ptr;
-}
-void *malloc_debug(const char *file, int line, const char *function, size_t size) {
- void *ptr = malloc(size + MALLOC_OVERHEAD);
- if(!ptr) fatal("MEMORY: Cannot allocate %zu bytes for %s/%u@%s.", size, function, line, file);
-
- allocations.allocated += size;
- allocations.allocations++;
-
- debug(D_MEMORY, "MEMORY: Allocated %zu bytes for %s/%u@%s."
- " Status: allocated %zu in %zu allocs."
- , size
- , function, line, file
- , allocations.allocated
- , allocations.allocations
- );
-
- if(allocations.allocated > allocations.allocated_max) {
- debug(D_MEMORY, "MEMORY: total allocation peak increased from %zu to %zu", allocations.allocated_max, allocations.allocated);
- allocations.allocated_max = allocations.allocated;
- }
-
- size_t csize;
- check_allocation(file, line, function, mark_allocation(ptr, size), &csize);
- if(size != csize) {
- fatal("Invalid size.");
- }
-
- return mark_allocation(ptr, size);
-}
+// ----------------------------------------------------------------------------
+// target
+//
+// target is the structure that processes are aggregated to be reported
+// to netdata.
+//
+// - Each entry in /etc/apps_groups.conf creates a target.
+// - Each user and group used by a process in the system, creates a target.
-void *calloc_debug(const char *file, int line, const char *function, size_t nmemb, size_t size) {
- void *ptr = malloc_debug(file, line, function, (nmemb * size));
- bzero(ptr, nmemb * size);
- return ptr;
-}
+struct target {
+ char compare[MAX_COMPARE_NAME + 1];
+ uint32_t comparehash;
+ size_t comparelen;
+
+ char id[MAX_NAME + 1];
+ uint32_t idhash;
+
+ char name[MAX_NAME + 1];
+
+ uid_t uid;
+ gid_t gid;
+
+ kernel_uint_t minflt;
+ kernel_uint_t cminflt;
+ kernel_uint_t majflt;
+ kernel_uint_t cmajflt;
+ kernel_uint_t utime;
+ kernel_uint_t stime;
+ kernel_uint_t gtime;
+ kernel_uint_t cutime;
+ kernel_uint_t cstime;
+ kernel_uint_t cgtime;
+ kernel_uint_t num_threads;
+ // kernel_uint_t rss;
+
+ kernel_uint_t statm_size;
+ kernel_uint_t statm_resident;
+ kernel_uint_t statm_share;
+ // kernel_uint_t statm_text;
+ // kernel_uint_t statm_lib;
+ // kernel_uint_t statm_data;
+ // kernel_uint_t statm_dirty;
+
+ kernel_uint_t io_logical_bytes_read;
+ kernel_uint_t io_logical_bytes_written;
+ // kernel_uint_t io_read_calls;
+ // kernel_uint_t io_write_calls;
+ kernel_uint_t io_storage_bytes_read;
+ kernel_uint_t io_storage_bytes_written;
+ // kernel_uint_t io_cancelled_write_bytes;
+
+ int *target_fds;
+ int target_fds_size;
+
+ kernel_uint_t openfiles;
+ kernel_uint_t openpipes;
+ kernel_uint_t opensockets;
+ kernel_uint_t openinotifies;
+ kernel_uint_t openeventfds;
+ kernel_uint_t opentimerfds;
+ kernel_uint_t opensignalfds;
+ kernel_uint_t openeventpolls;
+ kernel_uint_t openother;
+
+ unsigned int processes; // how many processes have been merged to this
+ int exposed; // if set, we have sent this to netdata
+ int hidden; // if set, we set the hidden flag on the dimension
+ int debug;
+ int ends_with;
+ int starts_with; // if set, the compare string matches only the
+ // beginning of the command
+
+ struct target *target; // the one that will be reported to netdata
+ struct target *next;
+};
-void free_debug(const char *file, int line, const char *function, void *ptr) {
- size_t size;
- void *real_ptr = check_allocation(file, line, function, ptr, &size);
+struct target
+ *apps_groups_default_target = NULL, // the default target
+ *apps_groups_root_target = NULL, // apps_groups.conf defined
+ *users_root_target = NULL, // users
+ *groups_root_target = NULL; // user groups
- bzero(real_ptr, size + MALLOC_OVERHEAD);
+size_t
+ apps_groups_targets_count = 0; // # of apps_groups.conf targets
- free(real_ptr);
- allocations.allocated -= size;
- allocations.allocations--;
- debug(D_MEMORY, "MEMORY: freed %zu bytes for %s/%u@%s."
- " Status: allocated %zu in %zu allocs."
- , size
- , function, line, file
- , allocations.allocated
- , allocations.allocations
- );
-}
+// ----------------------------------------------------------------------------
+// pid_stat
+//
+// structure to store data for each process running
+// see: man proc for the description of the fields
-void *realloc_debug(const char *file, int line, const char *function, void *ptr, size_t size) {
- if(!ptr) return malloc_debug(file, line, function, size);
- if(!size) { free_debug(file, line, function, ptr); return NULL; }
+struct pid_stat {
+ int32_t pid;
+ char comm[MAX_COMPARE_NAME + 1];
+ char cmdline[MAX_CMDLINE + 1];
+
+ uint32_t log_thrown;
+
+ // char state;
+ int32_t ppid;
+ // int32_t pgrp;
+ // int32_t session;
+ // int32_t tty_nr;
+ // int32_t tpgid;
+ // uint64_t flags;
+
+ // these are raw values collected
+ kernel_uint_t minflt_raw;
+ kernel_uint_t cminflt_raw;
+ kernel_uint_t majflt_raw;
+ kernel_uint_t cmajflt_raw;
+ kernel_uint_t utime_raw;
+ kernel_uint_t stime_raw;
+ kernel_uint_t gtime_raw; // guest_time
+ kernel_uint_t cutime_raw;
+ kernel_uint_t cstime_raw;
+ kernel_uint_t cgtime_raw; // cguest_time
+
+ // these are rates
+ kernel_uint_t minflt;
+ kernel_uint_t cminflt;
+ kernel_uint_t majflt;
+ kernel_uint_t cmajflt;
+ kernel_uint_t utime;
+ kernel_uint_t stime;
+ kernel_uint_t gtime;
+ kernel_uint_t cutime;
+ kernel_uint_t cstime;
+ kernel_uint_t cgtime;
+
+ // int64_t priority;
+ // int64_t nice;
+ int32_t num_threads;
+ // int64_t itrealvalue;
+ // kernel_uint_t starttime;
+ // kernel_uint_t vsize;
+ // kernel_uint_t rss;
+ // kernel_uint_t rsslim;
+ // kernel_uint_t starcode;
+ // kernel_uint_t endcode;
+ // kernel_uint_t startstack;
+ // kernel_uint_t kstkesp;
+ // kernel_uint_t kstkeip;
+ // uint64_t signal;
+ // uint64_t blocked;
+ // uint64_t sigignore;
+ // uint64_t sigcatch;
+ // uint64_t wchan;
+ // uint64_t nswap;
+ // uint64_t cnswap;
+ // int32_t exit_signal;
+ // int32_t processor;
+ // uint32_t rt_priority;
+ // uint32_t policy;
+ // kernel_uint_t delayacct_blkio_ticks;
+
+ uid_t uid;
+ gid_t gid;
+
+ kernel_uint_t statm_size;
+ kernel_uint_t statm_resident;
+ kernel_uint_t statm_share;
+ // kernel_uint_t statm_text;
+ // kernel_uint_t statm_lib;
+ // kernel_uint_t statm_data;
+ // kernel_uint_t statm_dirty;
+
+ kernel_uint_t io_logical_bytes_read_raw;
+ kernel_uint_t io_logical_bytes_written_raw;
+ // kernel_uint_t io_read_calls_raw;
+ // kernel_uint_t io_write_calls_raw;
+ kernel_uint_t io_storage_bytes_read_raw;
+ kernel_uint_t io_storage_bytes_written_raw;
+ // kernel_uint_t io_cancelled_write_bytes_raw;
+
+ kernel_uint_t io_logical_bytes_read;
+ kernel_uint_t io_logical_bytes_written;
+ // kernel_uint_t io_read_calls;
+ // kernel_uint_t io_write_calls;
+ kernel_uint_t io_storage_bytes_read;
+ kernel_uint_t io_storage_bytes_written;
+ // kernel_uint_t io_cancelled_write_bytes;
+
+ int *fds; // array of fds it uses
+ int fds_size; // the size of the fds array
+
+ int children_count; // number of processes directly referencing this
+ char keep:1; // 1 when we need to keep this process in memory even after it exited
+ int keeploops; // increases by 1 every time keep is 1 and updated 0
+ char updated:1; // 1 when the process is currently running
+ char merged:1; // 1 when it has been merged to its parent
+ char read:1; // 1 when we have already read this process for this iteration
+
+ int sortlist; // higher numbers = top on the process tree
+ // each process gets a unique number
+
+ struct target *target; // app_groups.conf targets
+ struct target *user_target; // uid based targets
+ struct target *group_target; // gid based targets
+
+ usec_t stat_collected_usec;
+ usec_t last_stat_collected_usec;
+
+ usec_t io_collected_usec;
+ usec_t last_io_collected_usec;
+
+ char *fds_dirname; // the full directory name in /proc/PID/fd
+
+ char *stat_filename;
+ char *statm_filename;
+ char *io_filename;
+ char *cmdline_filename;
+
+ struct pid_stat *parent;
+ struct pid_stat *prev;
+ struct pid_stat *next;
+};
- size_t old_size;
- void *real_ptr = check_allocation(file, line, function, ptr, &old_size);
+// log each problem once per process
+// log flood protection flags (log_thrown)
+#define PID_LOG_IO 0x00000001
+#define PID_LOG_STATM 0x00000002
+#define PID_LOG_CMDLINE 0x00000004
+#define PID_LOG_FDS 0x00000008
+#define PID_LOG_STAT 0x00000010
+
+static struct pid_stat
+ *root_of_pids = NULL, // global list of all processes running
+ **all_pids = NULL; // to avoid allocations, we pre-allocate the
+ // the entire pid space.
+
+static size_t
+ all_pids_count = 0; // the number of processes running
+
+#if (ALL_PIDS_ARE_READ_INSTANTLY == 0)
+// Another pre-allocated list of all possible pids.
+// We need it to pids and assign them a unique sortlist id, so that we
+// read parents before children. This is needed to prevent a situation where
+// a child is found running, but until we read its parent, it has exited and
+// its parent has accumulated its resources.
+static pid_t
+ *all_pids_sortlist = NULL;
+#endif
- void *new_ptr = realloc(real_ptr, size + MALLOC_OVERHEAD);
- if(!new_ptr) fatal("MEMORY: Cannot allocate %zu bytes for %s/%u@%s.", size, function, line, file);
+// ----------------------------------------------------------------------------
+// file descriptor
+//
+// this is used to keep a global list of all open files of the system.
+// it is needed in order to calculate the unique files processes have open.
- allocations.allocated += size;
- allocations.allocated -= old_size;
+#define FILE_DESCRIPTORS_INCREASE_STEP 100
- debug(D_MEMORY, "MEMORY: Re-allocated from %zu to %zu bytes for %s/%u@%s."
- " Status: allocated %z in %zu allocs."
- , old_size, size
- , function, line, file
- , allocations.allocated
- , allocations.allocations
- );
+// types for struct file_descriptor->type
+typedef enum fd_filetype {
+ FILETYPE_OTHER,
+ FILETYPE_FILE,
+ FILETYPE_PIPE,
+ FILETYPE_SOCKET,
+ FILETYPE_INOTIFY,
+ FILETYPE_EVENTFD,
+ FILETYPE_EVENTPOLL,
+ FILETYPE_TIMERFD,
+ FILETYPE_SIGNALFD
+} FD_FILETYPE;
- if(allocations.allocated > allocations.allocated_max) {
- debug(D_MEMORY, "MEMORY: total allocation peak increased from %zu to %zu", allocations.allocated_max, allocations.allocated);
- allocations.allocated_max = allocations.allocated;
- }
+struct file_descriptor {
+ avl avl;
- return mark_allocation(new_ptr, size);
-}
+#ifdef NETDATA_INTERNAL_CHECKS
+ uint32_t magic;
+#endif /* NETDATA_INTERNAL_CHECKS */
-char *strdup_debug(const char *file, int line, const char *function, const char *ptr) {
- size_t size = 0;
- const char *s = ptr;
+ const char *name;
+ uint32_t hash;
- while(*s++) size++;
- size++;
+ FD_FILETYPE type;
+ int count;
+ int pos;
+} *all_files = NULL;
- char *p = malloc_debug(file, line, function, size);
- if(!p) fatal("Cannot allocate %zu bytes.", size);
+static int
+ all_files_len = 0,
+ all_files_size = 0;
- memcpy(p, ptr, size);
- return p;
+// ----------------------------------------------------------------------------
+// callback required by fatal()
+
+void netdata_cleanup_and_exit(int ret) {
+ exit(ret);
}
-#define malloc(size) malloc_debug(__FILE__, __LINE__, __FUNCTION__, (size))
-#define calloc(nmemb, size) calloc_debug(__FILE__, __LINE__, __FUNCTION__, (nmemb), (size))
-#define realloc(ptr, size) realloc_debug(__FILE__, __LINE__, __FUNCTION__, (ptr), (size))
-#define free(ptr) free_debug(__FILE__, __LINE__, __FUNCTION__, (ptr))
+// ----------------------------------------------------------------------------
+// apps_groups.conf
+// aggregate all processes in groups, to have a limited number of dimensions
+
+static struct target *get_users_target(uid_t uid) {
+ struct target *w;
+ for(w = users_root_target ; w ; w = w->next)
+ if(w->uid == uid) return w;
-#ifdef strdup
-#undef strdup
-#endif
-#define strdup(ptr) strdup_debug(__FILE__, __LINE__, __FUNCTION__, (ptr))
+ w = callocz(sizeof(struct target), 1);
+ snprintfz(w->compare, MAX_COMPARE_NAME, "%u", uid);
+ w->comparehash = simple_hash(w->compare);
+ w->comparelen = strlen(w->compare);
-#endif /* NETDATA_INTERNAL_CHECKS */
+ snprintfz(w->id, MAX_NAME, "%u", uid);
+ w->idhash = simple_hash(w->id);
-// ----------------------------------------------------------------------------
-// system functions
-// to retrieve settings of the system
+ struct passwd *pw = getpwuid(uid);
+ if(!pw)
+ snprintfz(w->name, MAX_NAME, "%u", uid);
+ else
+ snprintfz(w->name, MAX_NAME, "%s", pw->pw_name);
-long get_system_cpus(void) {
- procfile *ff = NULL;
+ netdata_fix_chart_name(w->name);
- int processors = 0;
+ w->uid = uid;
- char filename[FILENAME_MAX + 1];
- snprintf(filename, FILENAME_MAX, "%s/proc/stat", host_prefix);
+ w->next = users_root_target;
+ users_root_target = w;
- ff = procfile_open(filename, NULL, PROCFILE_FLAG_DEFAULT);
- if(!ff) return 1;
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: added uid %u ('%s') target\n", w->uid, w->name);
- ff = procfile_readall(ff);
- if(!ff) {
- procfile_close(ff);
- return 1;
- }
+ return w;
+}
- unsigned int i;
- for(i = 0; i < procfile_lines(ff); i++) {
- if(!procfile_linewords(ff, i)) continue;
+struct target *get_groups_target(gid_t gid)
+{
+ struct target *w;
+ for(w = groups_root_target ; w ; w = w->next)
+ if(w->gid == gid) return w;
- if(strncmp(procfile_lineword(ff, i, 0), "cpu", 3) == 0) processors++;
- }
- processors--;
- if(processors < 1) processors = 1;
+ w = callocz(sizeof(struct target), 1);
+ snprintfz(w->compare, MAX_COMPARE_NAME, "%u", gid);
+ w->comparehash = simple_hash(w->compare);
+ w->comparelen = strlen(w->compare);
- procfile_close(ff);
- return processors;
-}
+ snprintfz(w->id, MAX_NAME, "%u", gid);
+ w->idhash = simple_hash(w->id);
-long get_system_pid_max(void) {
- procfile *ff = NULL;
- long mpid = 32768;
+ struct group *gr = getgrgid(gid);
+ if(!gr)
+ snprintfz(w->name, MAX_NAME, "%u", gid);
+ else
+ snprintfz(w->name, MAX_NAME, "%s", gr->gr_name);
- char filename[FILENAME_MAX + 1];
- snprintf(filename, FILENAME_MAX, "%s/proc/sys/kernel/pid_max", host_prefix);
- ff = procfile_open(filename, NULL, PROCFILE_FLAG_DEFAULT);
- if(!ff) return mpid;
+ netdata_fix_chart_name(w->name);
- ff = procfile_readall(ff);
- if(!ff) {
- procfile_close(ff);
- return mpid;
- }
+ w->gid = gid;
- mpid = atol(procfile_lineword(ff, 0, 0));
- if(!mpid) mpid = 32768;
+ w->next = groups_root_target;
+ groups_root_target = w;
- procfile_close(ff);
- return mpid;
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: added gid %u ('%s') target\n", w->gid, w->name);
+
+ return w;
}
-// ----------------------------------------------------------------------------
-// target
-// target is the structure that process data are aggregated
+// find or create a new target
+// there are targets that are just aggregated to other target (the second argument)
+static struct target *get_apps_groups_target(const char *id, struct target *target, const char *name) {
+ int tdebug = 0, thidden = target?target->hidden:0, ends_with = 0;
+ const char *nid = id;
+
+ // extract the options
+ while(nid[0] == '-' || nid[0] == '+' || nid[0] == '*') {
+ if(nid[0] == '-') thidden = 1;
+ if(nid[0] == '+') tdebug = 1;
+ if(nid[0] == '*') ends_with = 1;
+ nid++;
+ }
+ uint32_t hash = simple_hash(id);
+
+ // find if it already exists
+ struct target *w, *last = apps_groups_root_target;
+ for(w = apps_groups_root_target ; w ; w = w->next) {
+ if(w->idhash == hash && strncmp(nid, w->id, MAX_NAME) == 0)
+ return w;
+
+ last = w;
+ }
+
+ // find an existing target
+ if(unlikely(!target)) {
+ while(*name == '-') {
+ if(*name == '-') thidden = 1;
+ name++;
+ }
+
+ for(target = apps_groups_root_target ; target != NULL ; target = target->next) {
+ if(!target->target && strcmp(name, target->name) == 0)
+ break;
+ }
+
+ if(unlikely(debug)) {
+ if(unlikely(target))
+ fprintf(stderr, "apps.plugin: REUSING TARGET NAME '%s' on ID '%s'\n", target->name, target->id);
+ else
+ fprintf(stderr, "apps.plugin: NEW TARGET NAME '%s' on ID '%s'\n", name, id);
+ }
+ }
+
+ if(target && target->target)
+ fatal("Internal Error: request to link process '%s' to target '%s' which is linked to target '%s'", id, target->id, target->target->id);
+
+ w = callocz(sizeof(struct target), 1);
+ strncpyz(w->id, nid, MAX_NAME);
+ w->idhash = simple_hash(w->id);
+
+ if(unlikely(!target))
+ // copy the name
+ strncpyz(w->name, name, MAX_NAME);
+ else
+ // copy the id
+ strncpyz(w->name, nid, MAX_NAME);
+
+ strncpyz(w->compare, nid, MAX_COMPARE_NAME);
+ size_t len = strlen(w->compare);
+ if(w->compare[len - 1] == '*') {
+ w->compare[len - 1] = '\0';
+ w->starts_with = 1;
+ }
+ w->ends_with = ends_with;
+
+ if(w->starts_with && w->ends_with)
+ proc_pid_cmdline_is_needed = 1;
+
+ w->comparehash = simple_hash(w->compare);
+ w->comparelen = strlen(w->compare);
+
+ w->hidden = thidden;
+ w->debug = tdebug;
+ w->target = target;
+
+ // append it, to maintain the order in apps_groups.conf
+ if(last) last->next = w;
+ else apps_groups_root_target = w;
+
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: ADDING TARGET ID '%s', process name '%s' (%s), aggregated on target '%s', options: %s %s\n"
+ , w->id
+ , w->compare, (w->starts_with && w->ends_with)?"substring":((w->starts_with)?"prefix":((w->ends_with)?"suffix":"exact"))
+ , w->target?w->target->name:w->name
+ , (w->hidden)?"hidden":"-"
+ , (w->debug)?"debug":"-"
+ );
+
+ return w;
+}
-struct target {
- char compare[MAX_COMPARE_NAME + 1];
- uint32_t comparehash;
- size_t comparelen;
-
- char id[MAX_NAME + 1];
- uint32_t idhash;
-
- char name[MAX_NAME + 1];
-
- uid_t uid;
- gid_t gid;
-
- unsigned long long minflt;
- unsigned long long cminflt;
- unsigned long long majflt;
- unsigned long long cmajflt;
- unsigned long long utime;
- unsigned long long stime;
- unsigned long long cutime;
- unsigned long long cstime;
- unsigned long long num_threads;
- unsigned long long rss;
-
- unsigned long long fix_minflt;
- unsigned long long fix_cminflt;
- unsigned long long fix_majflt;
- unsigned long long fix_cmajflt;
- unsigned long long fix_utime;
- unsigned long long fix_stime;
- unsigned long long fix_cutime;
- unsigned long long fix_cstime;
-
- unsigned long long statm_size;
- unsigned long long statm_resident;
- unsigned long long statm_share;
- unsigned long long statm_text;
- unsigned long long statm_lib;
- unsigned long long statm_data;
- unsigned long long statm_dirty;
-
- unsigned long long io_logical_bytes_read;
- unsigned long long io_logical_bytes_written;
- unsigned long long io_read_calls;
- unsigned long long io_write_calls;
- unsigned long long io_storage_bytes_read;
- unsigned long long io_storage_bytes_written;
- unsigned long long io_cancelled_write_bytes;
-
- unsigned long long fix_io_logical_bytes_read;
- unsigned long long fix_io_logical_bytes_written;
- unsigned long long fix_io_read_calls;
- unsigned long long fix_io_write_calls;
- unsigned long long fix_io_storage_bytes_read;
- unsigned long long fix_io_storage_bytes_written;
- unsigned long long fix_io_cancelled_write_bytes;
-
- int *fds;
- unsigned long long openfiles;
- unsigned long long openpipes;
- unsigned long long opensockets;
- unsigned long long openinotifies;
- unsigned long long openeventfds;
- unsigned long long opentimerfds;
- unsigned long long opensignalfds;
- unsigned long long openeventpolls;
- unsigned long long openother;
-
- unsigned long processes; // how many processes have been merged to this
- int exposed; // if set, we have sent this to netdata
- int hidden; // if set, we set the hidden flag on the dimension
- int debug;
- int ends_with;
- int starts_with; // if set, the compare string matches only the
- // beginning of the command
-
- struct target *target; // the one that will be reported to netdata
- struct target *next;
-};
+// read the apps_groups.conf file
+static int read_apps_groups_conf(const char *file)
+{
+ char filename[FILENAME_MAX + 1];
+ snprintfz(filename, FILENAME_MAX, "%s/apps_%s.conf", config_dir, file);
-// ----------------------------------------------------------------------------
-// apps_groups.conf
-// aggregate all processes in groups, to have a limited number of dimensions
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: process groups file: '%s'\n", filename);
-struct target *apps_groups_root_target = NULL;
-struct target *apps_groups_default_target = NULL;
-long apps_groups_targets = 0;
+ // ----------------------------------------
-struct target *users_root_target = NULL;
-struct target *groups_root_target = NULL;
+ procfile *ff = procfile_open(filename, " :\t", PROCFILE_FLAG_DEFAULT);
+ if(!ff) return 1;
-struct target *get_users_target(uid_t uid)
-{
- struct target *w;
- for(w = users_root_target ; w ; w = w->next)
- if(w->uid == uid) return w;
+ procfile_set_quotes(ff, "'\"");
- w = calloc(sizeof(struct target), 1);
- if(unlikely(!w)) {
- error("Cannot allocate %lu bytes of memory", (unsigned long)sizeof(struct target));
- return NULL;
- }
+ ff = procfile_readall(ff);
+ if(!ff)
+ return 1;
- snprintf(w->compare, MAX_COMPARE_NAME, "%d", uid);
- w->comparehash = simple_hash(w->compare);
- w->comparelen = strlen(w->compare);
+ size_t line, lines = procfile_lines(ff);
- snprintf(w->id, MAX_NAME, "%d", uid);
- w->idhash = simple_hash(w->id);
+ for(line = 0; line < lines ;line++) {
+ size_t word, words = procfile_linewords(ff, line);
+ if(!words) continue;
- struct passwd *pw = getpwuid(uid);
- if(!pw)
- snprintf(w->name, MAX_NAME, "%d", uid);
- else
- snprintf(w->name, MAX_NAME, "%s", pw->pw_name);
+ char *name = procfile_lineword(ff, line, 0);
+ if(!name || !*name) continue;
- netdata_fix_chart_name(w->name);
+ // find a possibly existing target
+ struct target *w = NULL;
- w->uid = uid;
+ // loop through all words, skipping the first one (the name)
+ for(word = 0; word < words ;word++) {
+ char *s = procfile_lineword(ff, line, word);
+ if(!s || !*s) continue;
+ if(*s == '#') break;
- w->next = users_root_target;
- users_root_target = w;
+ // is this the first word? skip it
+ if(s == name) continue;
- if(unlikely(debug))
- fprintf(stderr, "apps.plugin: added uid %d ('%s') target\n", w->uid, w->name);
+ // add this target
+ struct target *n = get_apps_groups_target(s, w, name);
+ if(!n) {
+ error("Cannot create target '%s' (line %zu, word %zu)", s, line, word);
+ continue;
+ }
- return w;
-}
+ // just some optimization
+ // to avoid searching for a target for each process
+ if(!w) w = n->target?n->target:n;
+ }
+ }
-struct target *get_groups_target(gid_t gid)
-{
- struct target *w;
- for(w = groups_root_target ; w ; w = w->next)
- if(w->gid == gid) return w;
+ procfile_close(ff);
- w = calloc(sizeof(struct target), 1);
- if(unlikely(!w)) {
- error("Cannot allocate %lu bytes of memory", (unsigned long)sizeof(struct target));
- return NULL;
- }
+ apps_groups_default_target = get_apps_groups_target("p+!o@w#e$i^r&7*5(-i)l-o_", NULL, "other"); // match nothing
+ if(!apps_groups_default_target)
+ fatal("Cannot create default target");
- snprintf(w->compare, MAX_COMPARE_NAME, "%d", gid);
- w->comparehash = simple_hash(w->compare);
- w->comparelen = strlen(w->compare);
+ // allow the user to override group 'other'
+ if(apps_groups_default_target->target)
+ apps_groups_default_target = apps_groups_default_target->target;
- snprintf(w->id, MAX_NAME, "%d", gid);
- w->idhash = simple_hash(w->id);
+ return 0;
+}
- struct group *gr = getgrgid(gid);
- if(!gr)
- snprintf(w->name, MAX_NAME, "%d", gid);
- else
- snprintf(w->name, MAX_NAME, "%s", gr->gr_name);
- netdata_fix_chart_name(w->name);
+// ----------------------------------------------------------------------------
+// struct pid_stat management
+
+static inline struct pid_stat *get_pid_entry(pid_t pid) {
+ if(unlikely(all_pids[pid]))
+ return all_pids[pid];
+
+ struct pid_stat *p = callocz(sizeof(struct pid_stat), 1);
+ p->fds = callocz(sizeof(int), MAX_SPARE_FDS);
+ p->fds_size = MAX_SPARE_FDS;
- w->gid = gid;
+ if(likely(root_of_pids))
+ root_of_pids->prev = p;
- w->next = groups_root_target;
- groups_root_target = w;
+ p->next = root_of_pids;
+ root_of_pids = p;
- if(unlikely(debug))
- fprintf(stderr, "apps.plugin: added gid %d ('%s') target\n", w->gid, w->name);
+ p->pid = pid;
- return w;
+ all_pids[pid] = p;
+ all_pids_count++;
+
+ return p;
}
-// find or create a new target
-// there are targets that are just aggregated to other target (the second argument)
-struct target *get_apps_groups_target(const char *id, struct target *target)
-{
- int tdebug = 0, thidden = 0, ends_with = 0;
- const char *nid = id;
-
- while(nid[0] == '-' || nid[0] == '+' || nid[0] == '*') {
- if(nid[0] == '-') thidden = 1;
- if(nid[0] == '+') tdebug = 1;
- if(nid[0] == '*') ends_with = 1;
- nid++;
- }
- uint32_t hash = simple_hash(id);
-
- struct target *w;
- for(w = apps_groups_root_target ; w ; w = w->next) {
- if(w->idhash == hash && strncmp(nid, w->id, MAX_NAME) == 0)
- return w;
- }
-
- w = calloc(sizeof(struct target), 1);
- if(unlikely(!w)) {
- error("Cannot allocate %lu bytes of memory", (unsigned long)sizeof(struct target));
- return NULL;
- }
-
- strncpy(w->id, nid, MAX_NAME);
- w->idhash = simple_hash(w->id);
-
- strncpy(w->name, nid, MAX_NAME);
-
- strncpy(w->compare, nid, MAX_COMPARE_NAME);
- int len = strlen(w->compare);
- if(w->compare[len - 1] == '*') {
- w->compare[len - 1] = '\0';
- w->starts_with = 1;
- }
- w->ends_with = ends_with;
-
- if(w->starts_with && w->ends_with)
- proc_pid_cmdline_is_needed = 1;
-
- w->comparehash = simple_hash(w->compare);
- w->comparelen = strlen(w->compare);
-
- w->hidden = thidden;
- w->debug = tdebug;
- w->target = target;
-
- w->next = apps_groups_root_target;
- apps_groups_root_target = w;
-
- if(unlikely(debug))
- fprintf(stderr, "apps.plugin: ADDING TARGET ID '%s', process name '%s' (%s), aggregated on target '%s', options: %s %s\n"
- , w->id
- , w->compare, (w->starts_with && w->ends_with)?"substring":((w->starts_with)?"prefix":((w->ends_with)?"suffix":"exact"))
- , w->target?w->target->id:w->id
- , (w->hidden)?"hidden":"-"
- , (w->debug)?"debug":"-"
- );
-
- return w;
+static inline void del_pid_entry(pid_t pid) {
+ struct pid_stat *p = all_pids[pid];
+
+ if(unlikely(!p)) {
+ error("attempted to free pid %d that is not allocated.", pid);
+ return;
+ }
+
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: process %d %s exited, deleting it.\n", pid, p->comm);
+
+ if(root_of_pids == p)
+ root_of_pids = p->next;
+
+ if(p->next) p->next->prev = p->prev;
+ if(p->prev) p->prev->next = p->next;
+
+ freez(p->fds);
+ freez(p->fds_dirname);
+ freez(p->stat_filename);
+ freez(p->statm_filename);
+ freez(p->io_filename);
+ freez(p->cmdline_filename);
+ freez(p);
+
+ all_pids[pid] = NULL;
+ all_pids_count--;
}
-// read the apps_groups.conf file
-int read_apps_groups_conf(const char *name)
-{
- char filename[FILENAME_MAX + 1];
+// ----------------------------------------------------------------------------
- snprintf(filename, FILENAME_MAX, "%s/apps_%s.conf", config_dir, name);
+static inline int managed_log(struct pid_stat *p, uint32_t log, int status) {
+ if(unlikely(!status)) {
+ // error("command failed log %u, errno %d", log, errno);
+
+ if(unlikely(debug || errno != ENOENT)) {
+ if(unlikely(debug || !(p->log_thrown & log))) {
+ p->log_thrown |= log;
+ switch(log) {
+ case PID_LOG_IO:
+ error("Cannot process %s/proc/%d/io (command '%s')", netdata_configured_host_prefix, p->pid, p->comm);
+ break;
+
+ case PID_LOG_STATM:
+ error("Cannot process %s/proc/%d/statm (command '%s')", netdata_configured_host_prefix, p->pid, p->comm);
+ break;
+
+ case PID_LOG_CMDLINE:
+ error("Cannot process %s/proc/%d/cmdline (command '%s')", netdata_configured_host_prefix, p->pid, p->comm);
+ break;
+
+ case PID_LOG_FDS:
+ error("Cannot process entries in %s/proc/%d/fd (command '%s')", netdata_configured_host_prefix, p->pid, p->comm);
+ break;
+
+ case PID_LOG_STAT:
+ break;
+
+ default:
+ error("unhandled error for pid %d, command '%s'", p->pid, p->comm);
+ break;
+ }
+ }
+ }
+ errno = 0;
+ }
+ else if(unlikely(p->log_thrown & log)) {
+ // error("unsetting log %u on pid %d", log, p->pid);
+ p->log_thrown &= ~log;
+ }
+
+ return status;
+}
- if(unlikely(debug))
- fprintf(stderr, "apps.plugin: process groups file: '%s'\n", filename);
+static inline void assign_target_to_pid(struct pid_stat *p) {
+ uint32_t hash = simple_hash(p->comm);
+ size_t pclen = strlen(p->comm);
- // ----------------------------------------
+ struct target *w;
+ for(w = apps_groups_root_target; w ; w = w->next) {
+ // if(debug || (p->target && p->target->debug)) fprintf(stderr, "apps.plugin: \t\tcomparing '%s' with '%s'\n", w->compare, p->comm);
- procfile *ff = procfile_open(filename, " :\t", PROCFILE_FLAG_DEFAULT);
- if(!ff) return 1;
+ // find it - 4 cases:
+ // 1. the target is not a pattern
+ // 2. the target has the prefix
+ // 3. the target has the suffix
+ // 4. the target is something inside cmdline
- procfile_set_quotes(ff, "'\"");
+ if(unlikely(( (!w->starts_with && !w->ends_with && w->comparehash == hash && !strcmp(w->compare, p->comm))
+ || (w->starts_with && !w->ends_with && !strncmp(w->compare, p->comm, w->comparelen))
+ || (!w->starts_with && w->ends_with && pclen >= w->comparelen && !strcmp(w->compare, &p->comm[pclen - w->comparelen]))
+ || (proc_pid_cmdline_is_needed && w->starts_with && w->ends_with && strstr(p->cmdline, w->compare))
+ ))) {
- ff = procfile_readall(ff);
- if(!ff) {
- procfile_close(ff);
- return 1;
- }
+ if(w->target) p->target = w->target;
+ else p->target = w;
- unsigned long line, lines = procfile_lines(ff);
+ if(debug || (p->target && p->target->debug))
+ fprintf(stderr, "apps.plugin: \t\t%s linked to target %s\n", p->comm, p->target->name);
- for(line = 0; line < lines ;line++) {
- unsigned long word, words = procfile_linewords(ff, line);
- struct target *w = NULL;
+ break;
+ }
+ }
+}
- char *t = procfile_lineword(ff, line, 0);
- if(!t || !*t) continue;
- for(word = 0; word < words ;word++) {
- char *s = procfile_lineword(ff, line, word);
- if(!s || !*s) continue;
- if(*s == '#') break;
+// ----------------------------------------------------------------------------
+// update pids from proc
- if(t == s) continue;
+static inline int read_proc_pid_cmdline(struct pid_stat *p) {
- struct target *n = get_apps_groups_target(s, w);
- if(!n) {
- error("Cannot create target '%s' (line %d, word %d)", s, line, word);
- continue;
- }
+#ifdef __FreeBSD__
+ size_t i, bytes = MAX_CMDLINE;
+ int mib[4];
- if(!w) w = n;
- }
+ mib[0] = CTL_KERN;
+ mib[1] = KERN_PROC;
+ mib[2] = KERN_PROC_ARGS;
+ mib[3] = p->pid;
+ if (unlikely(sysctl(mib, 4, p->cmdline, &bytes, NULL, 0)))
+ goto cleanup;
+#else
+ if(unlikely(!p->cmdline_filename)) {
+ char filename[FILENAME_MAX + 1];
+ snprintfz(filename, FILENAME_MAX, "%s/proc/%d/cmdline", netdata_configured_host_prefix, p->pid);
+ p->cmdline_filename = strdupz(filename);
+ }
- if(w) {
- int tdebug = 0, thidden = 0;
+ int fd = open(p->cmdline_filename, O_RDONLY, 0666);
+ if(unlikely(fd == -1)) goto cleanup;
- while(t[0] == '-' || t[0] == '+') {
- if(t[0] == '-') thidden = 1;
- if(t[0] == '+') tdebug = 1;
- t++;
- }
+ ssize_t i, bytes = read(fd, p->cmdline, MAX_CMDLINE);
+ close(fd);
- strncpy(w->name, t, MAX_NAME);
- w->name[MAX_NAME] = '\0';
- w->hidden = thidden;
- w->debug = tdebug;
+ if(unlikely(bytes < 0)) goto cleanup;
+#endif
- if(unlikely(debug))
- fprintf(stderr, "apps.plugin: AGGREGATION TARGET NAME '%s' on ID '%s', process name '%s' (%s), aggregated on target '%s', options: %s %s\n"
- , w->name
- , w->id
- , w->compare, (w->starts_with && w->ends_with)?"substring":((w->starts_with)?"prefix":((w->ends_with)?"suffix":"exact"))
- , w->target?w->target->id:w->id
- , (w->hidden)?"hidden":"-"
- , (w->debug)?"debug":"-"
- );
- }
- }
+ p->cmdline[bytes] = '\0';
+ for(i = 0; i < bytes ; i++)
+ if(unlikely(!p->cmdline[i])) p->cmdline[i] = ' ';
- procfile_close(ff);
+ if(unlikely(debug))
+ fprintf(stderr, "Read file '%s' contents: %s\n", p->cmdline_filename, p->cmdline);
- apps_groups_default_target = get_apps_groups_target("p+!o@w#e$i^r&7*5(-i)l-o_", NULL); // match nothing
- if(!apps_groups_default_target)
- error("Cannot create default target");
- else
- strncpy(apps_groups_default_target->name, "other", MAX_NAME);
+ return 1;
- return 0;
+cleanup:
+ // copy the command to the command line
+ strncpyz(p->cmdline, p->comm, MAX_CMDLINE);
+ return 0;
}
+static inline int read_proc_pid_ownership(struct pid_stat *p, void *ptr) {
+ (void)ptr;
+#ifdef __FreeBSD__
+ struct kinfo_proc *proc_info = (struct kinfo_proc *)ptr;
-// ----------------------------------------------------------------------------
-// data to store for each pid
-// see: man proc
+ p->uid = proc_info->ki_uid;
+ p->gid = proc_info->ki_groups[0];
-struct pid_stat {
- int32_t pid;
- char comm[MAX_COMPARE_NAME + 1];
- char cmdline[MAX_CMDLINE + 1];
-
- // char state;
- int32_t ppid;
- // int32_t pgrp;
- // int32_t session;
- // int32_t tty_nr;
- // int32_t tpgid;
- // uint64_t flags;
- unsigned long long minflt;
- unsigned long long cminflt;
- unsigned long long majflt;
- unsigned long long cmajflt;
- unsigned long long utime;
- unsigned long long stime;
- unsigned long long cutime;
- unsigned long long cstime;
- // int64_t priority;
- // int64_t nice;
- int32_t num_threads;
- // int64_t itrealvalue;
- // unsigned long long starttime;
- // unsigned long long vsize;
- unsigned long long rss;
- // unsigned long long rsslim;
- // unsigned long long starcode;
- // unsigned long long endcode;
- // unsigned long long startstack;
- // unsigned long long kstkesp;
- // unsigned long long kstkeip;
- // uint64_t signal;
- // uint64_t blocked;
- // uint64_t sigignore;
- // uint64_t sigcatch;
- // uint64_t wchan;
- // uint64_t nswap;
- // uint64_t cnswap;
- // int32_t exit_signal;
- // int32_t processor;
- // uint32_t rt_priority;
- // uint32_t policy;
- // unsigned long long delayacct_blkio_ticks;
- // uint64_t guest_time;
- // int64_t cguest_time;
-
- uid_t uid;
- gid_t gid;
-
- unsigned long long statm_size;
- unsigned long long statm_resident;
- unsigned long long statm_share;
- unsigned long long statm_text;
- unsigned long long statm_lib;
- unsigned long long statm_data;
- unsigned long long statm_dirty;
-
- unsigned long long io_logical_bytes_read;
- unsigned long long io_logical_bytes_written;
- unsigned long long io_read_calls;
- unsigned long long io_write_calls;
- unsigned long long io_storage_bytes_read;
- unsigned long long io_storage_bytes_written;
- unsigned long long io_cancelled_write_bytes;
-
- // we need the last values
- // for all incremental counters
- // so that when a process switches users/groups
- // we will subtract these values from the old
- // target
- unsigned long long last_minflt;
- unsigned long long last_cminflt;
- unsigned long long last_majflt;
- unsigned long long last_cmajflt;
- unsigned long long last_utime;
- unsigned long long last_stime;
- unsigned long long last_cutime;
- unsigned long long last_cstime;
-
- unsigned long long last_io_logical_bytes_read;
- unsigned long long last_io_logical_bytes_written;
- unsigned long long last_io_read_calls;
- unsigned long long last_io_write_calls;
- unsigned long long last_io_storage_bytes_read;
- unsigned long long last_io_storage_bytes_written;
- unsigned long long last_io_cancelled_write_bytes;
-
-#ifdef AGGREGATE_CHILDREN_TO_PARENTS
- unsigned long long old_utime;
- unsigned long long old_stime;
- unsigned long long old_minflt;
- unsigned long long old_majflt;
-
- unsigned long long old_cutime;
- unsigned long long old_cstime;
- unsigned long long old_cminflt;
- unsigned long long old_cmajflt;
-
- unsigned long long fix_cutime;
- unsigned long long fix_cstime;
- unsigned long long fix_cminflt;
- unsigned long long fix_cmajflt;
-
- unsigned long long diff_cutime;
- unsigned long long diff_cstime;
- unsigned long long diff_cminflt;
- unsigned long long diff_cmajflt;
-#endif /* AGGREGATE_CHILDREN_TO_PARENTS */
-
- int *fds; // array of fds it uses
- int fds_size; // the size of the fds array
-
- int children_count; // number of processes directly referencing this
- int updated; // 1 when update
- int merged; // 1 when it has been merged to its parent
- int new_entry;
-
- struct target *target; // app_groups.conf targets
- struct target *user_target; // uid based targets
- struct target *group_target; // gid based targets
-
- struct pid_stat *parent;
- struct pid_stat *prev;
- struct pid_stat *next;
-
-} *root_of_pids = NULL, **all_pids;
-
-long all_pids_count = 0;
-
-struct pid_stat *get_pid_entry(pid_t pid)
-{
- if(all_pids[pid]) {
- all_pids[pid]->new_entry = 0;
- return all_pids[pid];
- }
-
- all_pids[pid] = calloc(sizeof(struct pid_stat), 1);
- if(!all_pids[pid]) {
- error("Cannot allocate %lu bytes of memory", (unsigned long)sizeof(struct pid_stat));
- return NULL;
- }
-
- all_pids[pid]->fds = calloc(sizeof(int), 100);
- if(!all_pids[pid]->fds)
- error("Cannot allocate %ld bytes of memory", (unsigned long)(sizeof(int) * 100));
- else all_pids[pid]->fds_size = 100;
-
- if(root_of_pids) root_of_pids->prev = all_pids[pid];
- all_pids[pid]->next = root_of_pids;
- root_of_pids = all_pids[pid];
-
- all_pids[pid]->pid = pid;
- all_pids[pid]->new_entry = 1;
-
- return all_pids[pid];
-}
+ return 1;
+#else
+ if(unlikely(!p->stat_filename)) {
+ error("pid %d does not have a stat_filename", p->pid);
+ return 0;
+ }
-void del_pid_entry(pid_t pid)
-{
- if(!all_pids[pid]) return;
+ // ----------------------------------------
+ // read uid and gid
- if(debug) fprintf(stderr, "apps.plugin: process %d %s exited, deleting it.\n", pid, all_pids[pid]->comm);
+ struct stat st;
+ if(stat(p->stat_filename, &st) != 0) {
+ error("Cannot stat file '%s'", p->stat_filename);
+ return 1;
+ }
- if(root_of_pids == all_pids[pid]) root_of_pids = all_pids[pid]->next;
- if(all_pids[pid]->next) all_pids[pid]->next->prev = all_pids[pid]->prev;
- if(all_pids[pid]->prev) all_pids[pid]->prev->next = all_pids[pid]->next;
+ p->uid = st.st_uid;
+ p->gid = st.st_gid;
- if(all_pids[pid]->fds) free(all_pids[pid]->fds);
- free(all_pids[pid]);
- all_pids[pid] = NULL;
+ return 1;
+#endif
}
+// ----------------------------------------------------------------------------
+// macro to calculate the incremental rate of a value
+// each parameter is accessed only ONCE - so it is safe to pass function calls
+// or other macros as parameters
+
+#define incremental_rate(rate_variable, last_kernel_variable, new_kernel_value, collected_usec, last_collected_usec) { \
+ kernel_uint_t _new_tmp = new_kernel_value; \
+ rate_variable = (_new_tmp - last_kernel_variable) * (USEC_PER_SEC * RATES_DETAIL) / (collected_usec - last_collected_usec); \
+ last_kernel_variable = _new_tmp; \
+ }
+
+// the same macro for struct pid members
+#define pid_incremental_rate(type, var, value) \
+ incremental_rate(var, var##_raw, value, p->type##_collected_usec, p->last_##type##_collected_usec)
+
// ----------------------------------------------------------------------------
-// update pids from proc
-int read_proc_pid_cmdline(struct pid_stat *p) {
- char filename[FILENAME_MAX + 1];
- snprintf(filename, FILENAME_MAX, "%s/proc/%d/cmdline", host_prefix, p->pid);
+static inline int read_proc_pid_stat(struct pid_stat *p, void *ptr) {
+ (void)ptr;
+
+#ifdef __FreeBSD__
+ struct kinfo_proc *proc_info = (struct kinfo_proc *)ptr;
+
+ if (unlikely(proc_info->ki_tdflags & TDF_IDLETD))
+ goto cleanup;
+#else
+ static procfile *ff = NULL;
+
+ if(unlikely(!p->stat_filename)) {
+ char filename[FILENAME_MAX + 1];
+ snprintfz(filename, FILENAME_MAX, "%s/proc/%d/stat", netdata_configured_host_prefix, p->pid);
+ p->stat_filename = strdupz(filename);
+ }
- int fd = open(filename, O_RDONLY, 0666);
- if(unlikely(fd == -1)) return 1;
+ int set_quotes = (!ff)?1:0;
- int i, bytes = read(fd, p->cmdline, MAX_CMDLINE);
- close(fd);
+ ff = procfile_reopen(ff, p->stat_filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
+ if(unlikely(!ff)) goto cleanup;
- if(bytes <= 0) {
- // copy the command to the command line
- strncpy(p->cmdline, p->comm, MAX_CMDLINE);
- p->cmdline[MAX_CMDLINE] = '\0';
- return 0;
- }
+ // if(set_quotes) procfile_set_quotes(ff, "()");
+ if(unlikely(set_quotes))
+ procfile_set_open_close(ff, "(", ")");
- p->cmdline[bytes] = '\0';
- for(i = 0; i < bytes ; i++)
- if(!p->cmdline[i]) p->cmdline[i] = ' ';
+ ff = procfile_readall(ff);
+ if(unlikely(!ff)) goto cleanup;
+#endif
- if(unlikely(debug))
- fprintf(stderr, "Read file '%s' contents: %s\n", filename, p->cmdline);
+ p->last_stat_collected_usec = p->stat_collected_usec;
+ p->stat_collected_usec = now_monotonic_usec();
+ calls_counter++;
+
+#ifdef __FreeBSD__
+ char *comm = proc_info->ki_comm;
+ p->ppid = proc_info->ki_ppid;
+#else
+ // p->pid = str2pid_t(procfile_lineword(ff, 0, 0));
+ char *comm = procfile_lineword(ff, 0, 1);
+ // p->state = *(procfile_lineword(ff, 0, 2));
+ p->ppid = (int32_t)str2pid_t(procfile_lineword(ff, 0, 3));
+ // p->pgrp = (int32_t)str2pid_t(procfile_lineword(ff, 0, 4));
+ // p->session = (int32_t)str2pid_t(procfile_lineword(ff, 0, 5));
+ // p->tty_nr = (int32_t)str2pid_t(procfile_lineword(ff, 0, 6));
+ // p->tpgid = (int32_t)str2pid_t(procfile_lineword(ff, 0, 7));
+ // p->flags = str2uint64_t(procfile_lineword(ff, 0, 8));
+#endif
+
+ if(strcmp(p->comm, comm)) {
+ if(unlikely(debug)) {
+ if(p->comm[0])
+ fprintf(stderr, "apps.plugin: \tpid %d (%s) changed name to '%s'\n", p->pid, p->comm, comm);
+ else
+ fprintf(stderr, "apps.plugin: \tJust added %d (%s)\n", p->pid, comm);
+ }
+
+ strncpyz(p->comm, comm, MAX_COMPARE_NAME);
+
+ // /proc/<pid>/cmdline
+ if(likely(proc_pid_cmdline_is_needed))
+ managed_log(p, PID_LOG_CMDLINE, read_proc_pid_cmdline(p));
+
+ assign_target_to_pid(p);
+ }
+
+#ifdef __FreeBSD__
+ pid_incremental_rate(stat, p->minflt, (kernel_uint_t)proc_info->ki_rusage.ru_minflt);
+ pid_incremental_rate(stat, p->cminflt, (kernel_uint_t)proc_info->ki_rusage_ch.ru_minflt);
+ pid_incremental_rate(stat, p->majflt, (kernel_uint_t)proc_info->ki_rusage.ru_majflt);
+ pid_incremental_rate(stat, p->cmajflt, (kernel_uint_t)proc_info->ki_rusage_ch.ru_majflt);
+ pid_incremental_rate(stat, p->utime, (kernel_uint_t)proc_info->ki_rusage.ru_utime.tv_sec * 100 + proc_info->ki_rusage.ru_utime.tv_usec / 10000);
+ pid_incremental_rate(stat, p->stime, (kernel_uint_t)proc_info->ki_rusage.ru_stime.tv_sec * 100 + proc_info->ki_rusage.ru_stime.tv_usec / 10000);
+ pid_incremental_rate(stat, p->cutime, (kernel_uint_t)proc_info->ki_rusage_ch.ru_utime.tv_sec * 100 + proc_info->ki_rusage_ch.ru_utime.tv_usec / 10000);
+ pid_incremental_rate(stat, p->cstime, (kernel_uint_t)proc_info->ki_rusage_ch.ru_stime.tv_sec * 100 + proc_info->ki_rusage_ch.ru_utime.tv_usec / 10000);
+
+ p->num_threads = proc_info->ki_numthreads;
+
+ if(enable_guest_charts) {
+ enable_guest_charts = 0;
+ info("Guest charts aren't supported by FreeBSD");
+ }
+#else
+ pid_incremental_rate(stat, p->minflt, str2kernel_uint_t(procfile_lineword(ff, 0, 9)));
+ pid_incremental_rate(stat, p->cminflt, str2kernel_uint_t(procfile_lineword(ff, 0, 10)));
+ pid_incremental_rate(stat, p->majflt, str2kernel_uint_t(procfile_lineword(ff, 0, 11)));
+ pid_incremental_rate(stat, p->cmajflt, str2kernel_uint_t(procfile_lineword(ff, 0, 12)));
+ pid_incremental_rate(stat, p->utime, str2kernel_uint_t(procfile_lineword(ff, 0, 13)));
+ pid_incremental_rate(stat, p->stime, str2kernel_uint_t(procfile_lineword(ff, 0, 14)));
+ pid_incremental_rate(stat, p->cutime, str2kernel_uint_t(procfile_lineword(ff, 0, 15)));
+ pid_incremental_rate(stat, p->cstime, str2kernel_uint_t(procfile_lineword(ff, 0, 16)));
+ // p->priority = str2kernel_uint_t(procfile_lineword(ff, 0, 17));
+ // p->nice = str2kernel_uint_t(procfile_lineword(ff, 0, 18));
+ p->num_threads = (int32_t)str2uint32_t(procfile_lineword(ff, 0, 19));
+ // p->itrealvalue = str2kernel_uint_t(procfile_lineword(ff, 0, 20));
+ // p->starttime = str2kernel_uint_t(procfile_lineword(ff, 0, 21));
+ // p->vsize = str2kernel_uint_t(procfile_lineword(ff, 0, 22));
+ // p->rss = str2kernel_uint_t(procfile_lineword(ff, 0, 23));
+ // p->rsslim = str2kernel_uint_t(procfile_lineword(ff, 0, 24));
+ // p->starcode = str2kernel_uint_t(procfile_lineword(ff, 0, 25));
+ // p->endcode = str2kernel_uint_t(procfile_lineword(ff, 0, 26));
+ // p->startstack = str2kernel_uint_t(procfile_lineword(ff, 0, 27));
+ // p->kstkesp = str2kernel_uint_t(procfile_lineword(ff, 0, 28));
+ // p->kstkeip = str2kernel_uint_t(procfile_lineword(ff, 0, 29));
+ // p->signal = str2kernel_uint_t(procfile_lineword(ff, 0, 30));
+ // p->blocked = str2kernel_uint_t(procfile_lineword(ff, 0, 31));
+ // p->sigignore = str2kernel_uint_t(procfile_lineword(ff, 0, 32));
+ // p->sigcatch = str2kernel_uint_t(procfile_lineword(ff, 0, 33));
+ // p->wchan = str2kernel_uint_t(procfile_lineword(ff, 0, 34));
+ // p->nswap = str2kernel_uint_t(procfile_lineword(ff, 0, 35));
+ // p->cnswap = str2kernel_uint_t(procfile_lineword(ff, 0, 36));
+ // p->exit_signal = str2kernel_uint_t(procfile_lineword(ff, 0, 37));
+ // p->processor = str2kernel_uint_t(procfile_lineword(ff, 0, 38));
+ // p->rt_priority = str2kernel_uint_t(procfile_lineword(ff, 0, 39));
+ // p->policy = str2kernel_uint_t(procfile_lineword(ff, 0, 40));
+ // p->delayacct_blkio_ticks = str2kernel_uint_t(procfile_lineword(ff, 0, 41));
+
+ if(enable_guest_charts) {
+
+ pid_incremental_rate(stat, p->gtime, str2kernel_uint_t(procfile_lineword(ff, 0, 42)));
+ pid_incremental_rate(stat, p->cgtime, str2kernel_uint_t(procfile_lineword(ff, 0, 43)));
+
+ if (show_guest_time || p->gtime || p->cgtime) {
+ p->utime -= (p->utime >= p->gtime) ? p->gtime : p->utime;
+ p->cutime -= (p->cutime >= p->cgtime) ? p->cgtime : p->cutime;
+ show_guest_time = 1;
+ }
+ }
+#endif
- return 0;
+ if(unlikely(debug || (p->target && p->target->debug)))
+ fprintf(stderr, "apps.plugin: READ PROC/PID/STAT: %s/proc/%d/stat, process: '%s' on target '%s' (dt=%llu) VALUES: utime=" KERNEL_UINT_FORMAT ", stime=" KERNEL_UINT_FORMAT ", cutime=" KERNEL_UINT_FORMAT ", cstime=" KERNEL_UINT_FORMAT ", minflt=" KERNEL_UINT_FORMAT ", majflt=" KERNEL_UINT_FORMAT ", cminflt=" KERNEL_UINT_FORMAT ", cmajflt=" KERNEL_UINT_FORMAT ", threads=%d\n", netdata_configured_host_prefix, p->pid, p->comm, (p->target)?p->target->name:"UNSET", p->stat_collected_usec - p->last_stat_collected_usec, p->utime, p->stime, p->cutime, p->cstime, p->minflt, p->majflt, p->cminflt, p->cmajflt, p->num_threads);
+
+ if(unlikely(global_iterations_counter == 1)) {
+ p->minflt = 0;
+ p->cminflt = 0;
+ p->majflt = 0;
+ p->cmajflt = 0;
+ p->utime = 0;
+ p->stime = 0;
+ p->gtime = 0;
+ p->cutime = 0;
+ p->cstime = 0;
+ p->cgtime = 0;
+ }
+
+ return 1;
+
+cleanup:
+ p->minflt = 0;
+ p->cminflt = 0;
+ p->majflt = 0;
+ p->cmajflt = 0;
+ p->utime = 0;
+ p->stime = 0;
+ p->gtime = 0;
+ p->cutime = 0;
+ p->cstime = 0;
+ p->cgtime = 0;
+ p->num_threads = 0;
+ // p->rss = 0;
+ return 0;
}
-int read_proc_pid_ownership(struct pid_stat *p) {
- char filename[FILENAME_MAX + 1];
+static inline int read_proc_pid_statm(struct pid_stat *p, void *ptr) {
+ (void)ptr;
+#ifdef __FreeBSD__
+ struct kinfo_proc *proc_info = (struct kinfo_proc *)ptr;
+#else
+ static procfile *ff = NULL;
- snprintf(filename, FILENAME_MAX, "%s/proc/%d", host_prefix, p->pid);
+ if(unlikely(!p->statm_filename)) {
+ char filename[FILENAME_MAX + 1];
+ snprintfz(filename, FILENAME_MAX, "%s/proc/%d/statm", netdata_configured_host_prefix, p->pid);
+ p->statm_filename = strdupz(filename);
+ }
- // ----------------------------------------
- // read uid and gid
+ ff = procfile_reopen(ff, p->statm_filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
+ if(unlikely(!ff)) goto cleanup;
- struct stat st;
- if(stat(filename, &st) != 0)
- return 1;
+ ff = procfile_readall(ff);
+ if(unlikely(!ff)) goto cleanup;
+#endif
- p->uid = st.st_uid;
- p->gid = st.st_gid;
+ calls_counter++;
- return 0;
-}
+#ifdef __FreeBSD__
+ p->statm_size = proc_info->ki_size / sysconf(_SC_PAGESIZE);
+ p->statm_resident = proc_info->ki_rssize;
+ p->statm_share = 0; // do we have to use ru_ixrss here?
+#else
+ p->statm_size = str2kernel_uint_t(procfile_lineword(ff, 0, 0));
+ p->statm_resident = str2kernel_uint_t(procfile_lineword(ff, 0, 1));
+ p->statm_share = str2kernel_uint_t(procfile_lineword(ff, 0, 2));
+ // p->statm_text = str2kernel_uint_t(procfile_lineword(ff, 0, 3));
+ // p->statm_lib = str2kernel_uint_t(procfile_lineword(ff, 0, 4));
+ // p->statm_data = str2kernel_uint_t(procfile_lineword(ff, 0, 5));
+ // p->statm_dirty = str2kernel_uint_t(procfile_lineword(ff, 0, 6));
+#endif
-int read_proc_pid_stat(struct pid_stat *p) {
- static procfile *ff = NULL;
-
- char filename[FILENAME_MAX + 1];
-
- snprintf(filename, FILENAME_MAX, "%s/proc/%d/stat", host_prefix, p->pid);
-
- // ----------------------------------------
-
- int set_quotes = (!ff)?1:0;
-
- ff = procfile_reopen(ff, filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
- if(!ff) return 1;
-
- // if(set_quotes) procfile_set_quotes(ff, "()");
- if(set_quotes) procfile_set_open_close(ff, "(", ")");
-
- ff = procfile_readall(ff);
- if(!ff) {
- // procfile_close(ff);
- return 1;
- }
-
- file_counter++;
-
- // parse the process name
- unsigned int i = 0;
- strncpy(p->comm, procfile_lineword(ff, 0, 1), MAX_COMPARE_NAME);
- p->comm[MAX_COMPARE_NAME] = '\0';
-
- // p->pid = atol(procfile_lineword(ff, 0, 0+i));
- // comm is at 1
- // p->state = *(procfile_lineword(ff, 0, 2+i));
- p->ppid = (int32_t) atol(procfile_lineword(ff, 0, 3 + i));
- // p->pgrp = atol(procfile_lineword(ff, 0, 4+i));
- // p->session = atol(procfile_lineword(ff, 0, 5+i));
- // p->tty_nr = atol(procfile_lineword(ff, 0, 6+i));
- // p->tpgid = atol(procfile_lineword(ff, 0, 7+i));
- // p->flags = strtoull(procfile_lineword(ff, 0, 8+i), NULL, 10);
- p->minflt = strtoull(procfile_lineword(ff, 0, 9+i), NULL, 10);
- p->cminflt = strtoull(procfile_lineword(ff, 0, 10+i), NULL, 10);
- p->majflt = strtoull(procfile_lineword(ff, 0, 11+i), NULL, 10);
- p->cmajflt = strtoull(procfile_lineword(ff, 0, 12+i), NULL, 10);
- p->utime = strtoull(procfile_lineword(ff, 0, 13+i), NULL, 10);
- p->stime = strtoull(procfile_lineword(ff, 0, 14+i), NULL, 10);
- p->cutime = strtoull(procfile_lineword(ff, 0, 15+i), NULL, 10);
- p->cstime = strtoull(procfile_lineword(ff, 0, 16+i), NULL, 10);
- // p->priority = strtoull(procfile_lineword(ff, 0, 17+i), NULL, 10);
- // p->nice = strtoull(procfile_lineword(ff, 0, 18+i), NULL, 10);
- p->num_threads = (int32_t) atol(procfile_lineword(ff, 0, 19 + i));
- // p->itrealvalue = strtoull(procfile_lineword(ff, 0, 20+i), NULL, 10);
- // p->starttime = strtoull(procfile_lineword(ff, 0, 21+i), NULL, 10);
- // p->vsize = strtoull(procfile_lineword(ff, 0, 22+i), NULL, 10);
- p->rss = strtoull(procfile_lineword(ff, 0, 23+i), NULL, 10);
- // p->rsslim = strtoull(procfile_lineword(ff, 0, 24+i), NULL, 10);
- // p->starcode = strtoull(procfile_lineword(ff, 0, 25+i), NULL, 10);
- // p->endcode = strtoull(procfile_lineword(ff, 0, 26+i), NULL, 10);
- // p->startstack = strtoull(procfile_lineword(ff, 0, 27+i), NULL, 10);
- // p->kstkesp = strtoull(procfile_lineword(ff, 0, 28+i), NULL, 10);
- // p->kstkeip = strtoull(procfile_lineword(ff, 0, 29+i), NULL, 10);
- // p->signal = strtoull(procfile_lineword(ff, 0, 30+i), NULL, 10);
- // p->blocked = strtoull(procfile_lineword(ff, 0, 31+i), NULL, 10);
- // p->sigignore = strtoull(procfile_lineword(ff, 0, 32+i), NULL, 10);
- // p->sigcatch = strtoull(procfile_lineword(ff, 0, 33+i), NULL, 10);
- // p->wchan = strtoull(procfile_lineword(ff, 0, 34+i), NULL, 10);
- // p->nswap = strtoull(procfile_lineword(ff, 0, 35+i), NULL, 10);
- // p->cnswap = strtoull(procfile_lineword(ff, 0, 36+i), NULL, 10);
- // p->exit_signal = atol(procfile_lineword(ff, 0, 37+i));
- // p->processor = atol(procfile_lineword(ff, 0, 38+i));
- // p->rt_priority = strtoul(procfile_lineword(ff, 0, 39+i), NULL, 10);
- // p->policy = strtoul(procfile_lineword(ff, 0, 40+i), NULL, 10);
- // p->delayacct_blkio_ticks = strtoull(procfile_lineword(ff, 0, 41+i), NULL, 10);
- // p->guest_time = strtoull(procfile_lineword(ff, 0, 42+i), NULL, 10);
- // p->cguest_time = strtoull(procfile_lineword(ff, 0, 43), NULL, 10);
-
- if(debug || (p->target && p->target->debug))
- fprintf(stderr, "apps.plugin: READ PROC/PID/STAT: %s/proc/%d/stat, process: '%s' VALUES: utime=%llu, stime=%llu, cutime=%llu, cstime=%llu, minflt=%llu, majflt=%llu, cminflt=%llu, cmajflt=%llu, threads=%d\n", host_prefix, p->pid, p->comm, p->utime, p->stime, p->cutime, p->cstime, p->minflt, p->majflt, p->cminflt, p->cmajflt, p->num_threads);
-
- // procfile_close(ff);
- return 0;
+ return 1;
+
+cleanup:
+ p->statm_size = 0;
+ p->statm_resident = 0;
+ p->statm_share = 0;
+ // p->statm_text = 0;
+ // p->statm_lib = 0;
+ // p->statm_data = 0;
+ // p->statm_dirty = 0;
+ return 0;
}
-int read_proc_pid_statm(struct pid_stat *p) {
- static procfile *ff = NULL;
+static inline int read_proc_pid_io(struct pid_stat *p, void *ptr) {
+ (void)ptr;
+#ifdef __FreeBSD__
+ struct kinfo_proc *proc_info = (struct kinfo_proc *)ptr;
+#else
+ static procfile *ff = NULL;
- char filename[FILENAME_MAX + 1];
+ if(unlikely(!p->io_filename)) {
+ char filename[FILENAME_MAX + 1];
+ snprintfz(filename, FILENAME_MAX, "%s/proc/%d/io", netdata_configured_host_prefix, p->pid);
+ p->io_filename = strdupz(filename);
+ }
- snprintf(filename, FILENAME_MAX, "%s/proc/%d/statm", host_prefix, p->pid);
+ // open the file
+ ff = procfile_reopen(ff, p->io_filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
+ if(unlikely(!ff)) goto cleanup;
- ff = procfile_reopen(ff, filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
- if(!ff) return 1;
+ ff = procfile_readall(ff);
+ if(unlikely(!ff)) goto cleanup;
+#endif
- ff = procfile_readall(ff);
- if(!ff) {
- // procfile_close(ff);
- return 1;
- }
+ calls_counter++;
- file_counter++;
+ p->last_io_collected_usec = p->io_collected_usec;
+ p->io_collected_usec = now_monotonic_usec();
- p->statm_size = strtoull(procfile_lineword(ff, 0, 0), NULL, 10);
- p->statm_resident = strtoull(procfile_lineword(ff, 0, 1), NULL, 10);
- p->statm_share = strtoull(procfile_lineword(ff, 0, 2), NULL, 10);
- p->statm_text = strtoull(procfile_lineword(ff, 0, 3), NULL, 10);
- p->statm_lib = strtoull(procfile_lineword(ff, 0, 4), NULL, 10);
- p->statm_data = strtoull(procfile_lineword(ff, 0, 5), NULL, 10);
- p->statm_dirty = strtoull(procfile_lineword(ff, 0, 6), NULL, 10);
+#ifdef __FreeBSD__
+ pid_incremental_rate(io, p->io_storage_bytes_read, proc_info->ki_rusage.ru_inblock);
+ pid_incremental_rate(io, p->io_storage_bytes_written, proc_info->ki_rusage.ru_oublock);
+#else
+ pid_incremental_rate(io, p->io_logical_bytes_read, str2kernel_uint_t(procfile_lineword(ff, 0, 1)));
+ pid_incremental_rate(io, p->io_logical_bytes_written, str2kernel_uint_t(procfile_lineword(ff, 1, 1)));
+ // pid_incremental_rate(io, p->io_read_calls, str2kernel_uint_t(procfile_lineword(ff, 2, 1)));
+ // pid_incremental_rate(io, p->io_write_calls, str2kernel_uint_t(procfile_lineword(ff, 3, 1)));
+ pid_incremental_rate(io, p->io_storage_bytes_read, str2kernel_uint_t(procfile_lineword(ff, 4, 1)));
+ pid_incremental_rate(io, p->io_storage_bytes_written, str2kernel_uint_t(procfile_lineword(ff, 5, 1)));
+ // pid_incremental_rate(io, p->io_cancelled_write_bytes, str2kernel_uint_t(procfile_lineword(ff, 6, 1)));
+#endif
- // procfile_close(ff);
- return 0;
+ if(unlikely(global_iterations_counter == 1)) {
+ p->io_logical_bytes_read = 0;
+ p->io_logical_bytes_written = 0;
+ // p->io_read_calls = 0;
+ // p->io_write_calls = 0;
+ p->io_storage_bytes_read = 0;
+ p->io_storage_bytes_written = 0;
+ // p->io_cancelled_write_bytes = 0;
+ }
+
+ return 1;
+
+cleanup:
+ p->io_logical_bytes_read = 0;
+ p->io_logical_bytes_written = 0;
+ // p->io_read_calls = 0;
+ // p->io_write_calls = 0;
+ p->io_storage_bytes_read = 0;
+ p->io_storage_bytes_written = 0;
+ // p->io_cancelled_write_bytes = 0;
+ return 0;
}
-int read_proc_pid_io(struct pid_stat *p) {
- static procfile *ff = NULL;
+static inline int read_proc_stat() {
+#ifdef __FreeBSD__
+ long cp_time[CPUSTATES];
+ int i;
- char filename[FILENAME_MAX + 1];
+ if (unlikely(CPUSTATES != 5)) {
+ error("FREEBSD: There are %d CPU states (5 was expected)", CPUSTATES);
+ goto cleanup;
+ }
+ if (unlikely(GETSYSCTL("kern.cp_time", cp_time))) goto cleanup;
+#else
+ static char filename[FILENAME_MAX + 1] = "";
+ static procfile *ff = NULL;
+#endif
+ static kernel_uint_t utime_raw = 0, stime_raw = 0, gtime_raw = 0, gntime_raw = 0, ntime_raw = 0;
+ static usec_t collected_usec = 0, last_collected_usec = 0;
+
+#ifndef __FreeBSD__
+ if(unlikely(!ff)) {
+ snprintfz(filename, FILENAME_MAX, "%s/proc/stat", netdata_configured_host_prefix);
+ ff = procfile_open(filename, " \t:", PROCFILE_FLAG_DEFAULT);
+ if(unlikely(!ff)) goto cleanup;
+ }
+
+ ff = procfile_readall(ff);
+ if(unlikely(!ff)) goto cleanup;
+#endif
- snprintf(filename, FILENAME_MAX, "%s/proc/%d/io", host_prefix, p->pid);
+ last_collected_usec = collected_usec;
+ collected_usec = now_monotonic_usec();
- ff = procfile_reopen(ff, filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
- if(!ff) return 1;
+ calls_counter++;
- ff = procfile_readall(ff);
- if(!ff) {
- // procfile_close(ff);
- return 1;
- }
+ // temporary - it is added global_ntime;
+ kernel_uint_t global_ntime = 0;
- file_counter++;
+#ifdef __FreeBSD__
+ incremental_rate(global_utime, utime_raw, cp_time[0], collected_usec, last_collected_usec);
+ incremental_rate(global_ntime, ntime_raw, cp_time[1], collected_usec, last_collected_usec);
+ incremental_rate(global_stime, stime_raw, cp_time[2], collected_usec, last_collected_usec);
+#else
+ incremental_rate(global_utime, utime_raw, str2kernel_uint_t(procfile_lineword(ff, 0, 1)), collected_usec, last_collected_usec);
+ incremental_rate(global_ntime, ntime_raw, str2kernel_uint_t(procfile_lineword(ff, 0, 2)), collected_usec, last_collected_usec);
+ incremental_rate(global_stime, stime_raw, str2kernel_uint_t(procfile_lineword(ff, 0, 3)), collected_usec, last_collected_usec);
+ incremental_rate(global_gtime, gtime_raw, str2kernel_uint_t(procfile_lineword(ff, 0, 10)), collected_usec, last_collected_usec);
+#endif
- p->io_logical_bytes_read = strtoull(procfile_lineword(ff, 0, 1), NULL, 10);
- p->io_logical_bytes_written = strtoull(procfile_lineword(ff, 1, 1), NULL, 10);
- p->io_read_calls = strtoull(procfile_lineword(ff, 2, 1), NULL, 10);
- p->io_write_calls = strtoull(procfile_lineword(ff, 3, 1), NULL, 10);
- p->io_storage_bytes_read = strtoull(procfile_lineword(ff, 4, 1), NULL, 10);
- p->io_storage_bytes_written = strtoull(procfile_lineword(ff, 5, 1), NULL, 10);
- p->io_cancelled_write_bytes = strtoull(procfile_lineword(ff, 6, 1), NULL, 10);
+ global_utime += global_ntime;
- // procfile_close(ff);
- return 0;
-}
+#ifdef __FreeBSD__
+ if(enable_guest_charts) {
+ enable_guest_charts = 0;
+ info("Guest charts aren't supported by FreeBSD");
+ }
+#else
+ if(enable_guest_charts) {
+ // temporary - it is added global_ntime;
+ kernel_uint_t global_gntime = 0;
+ // guest nice time, on guest time
+ incremental_rate(global_gntime, gntime_raw, str2kernel_uint_t(procfile_lineword(ff, 0, 11)), collected_usec, last_collected_usec);
-// ----------------------------------------------------------------------------
-// file descriptor
-// this is used to keep a global list of all open files of the system
-// it is needed in order to calculate the unique files processes have open
+ global_gtime += global_gntime;
-#define FILE_DESCRIPTORS_INCREASE_STEP 100
+ // remove guest time from user time
+ global_utime -= (global_utime > global_gtime) ? global_gtime : global_utime;
+ }
+#endif
-struct file_descriptor {
- avl avl;
-#ifdef NETDATA_INTERNAL_CHECKS
- uint32_t magic;
-#endif /* NETDATA_INTERNAL_CHECKS */
- uint32_t hash;
- const char *name;
- int type;
- int count;
- int pos;
-} *all_files = NULL;
+ if(unlikely(global_iterations_counter == 1)) {
+ global_utime = 0;
+ global_stime = 0;
+ global_gtime = 0;
+ }
+
+ return 1;
+
+cleanup:
+ global_utime = 0;
+ global_stime = 0;
+ global_gtime = 0;
+ return 0;
+}
-int all_files_len = 0;
-int all_files_size = 0;
+
+// ----------------------------------------------------------------------------
int file_descriptor_compare(void* a, void* b) {
#ifdef NETDATA_INTERNAL_CHECKS
- if(((struct file_descriptor *)a)->magic != 0x0BADCAFE || ((struct file_descriptor *)b)->magic != 0x0BADCAFE)
- error("Corrupted index data detected. Please report this.");
+ if(((struct file_descriptor *)a)->magic != 0x0BADCAFE || ((struct file_descriptor *)b)->magic != 0x0BADCAFE)
+ error("Corrupted index data detected. Please report this.");
#endif /* NETDATA_INTERNAL_CHECKS */
- if(((struct file_descriptor *)a)->hash < ((struct file_descriptor *)b)->hash)
- return -1;
+ if(((struct file_descriptor *)a)->hash < ((struct file_descriptor *)b)->hash)
+ return -1;
- else if(((struct file_descriptor *)a)->hash > ((struct file_descriptor *)b)->hash)
- return 1;
+ else if(((struct file_descriptor *)a)->hash > ((struct file_descriptor *)b)->hash)
+ return 1;
- else
- return strcmp(((struct file_descriptor *)a)->name, ((struct file_descriptor *)b)->name);
+ else
+ return strcmp(((struct file_descriptor *)a)->name, ((struct file_descriptor *)b)->name);
}
int file_descriptor_iterator(avl *a) { if(a) {}; return 0; }
avl_tree all_files_index = {
- NULL,
- file_descriptor_compare
+ NULL,
+ file_descriptor_compare
};
static struct file_descriptor *file_descriptor_find(const char *name, uint32_t hash) {
- struct file_descriptor *result = NULL, tmp;
- tmp.hash = (hash)?hash:simple_hash(name);
- tmp.name = name;
- tmp.count = 0;
- tmp.pos = 0;
+ struct file_descriptor tmp;
+ tmp.hash = (hash)?hash:simple_hash(name);
+ tmp.name = name;
+ tmp.count = 0;
+ tmp.pos = 0;
#ifdef NETDATA_INTERNAL_CHECKS
- tmp.magic = 0x0BADCAFE;
+ tmp.magic = 0x0BADCAFE;
#endif /* NETDATA_INTERNAL_CHECKS */
- avl_search(&all_files_index, (avl *) &tmp, file_descriptor_iterator, (avl **) &result);
- return result;
+ return (struct file_descriptor *)avl_search(&all_files_index, (avl *) &tmp);
}
#define file_descriptor_add(fd) avl_insert(&all_files_index, (avl *)(fd))
#define file_descriptor_remove(fd) avl_remove(&all_files_index, (avl *)(fd))
-#define FILETYPE_OTHER 0
-#define FILETYPE_FILE 1
-#define FILETYPE_PIPE 2
-#define FILETYPE_SOCKET 3
-#define FILETYPE_INOTIFY 4
-#define FILETYPE_EVENTFD 5
-#define FILETYPE_EVENTPOLL 6
-#define FILETYPE_TIMERFD 7
-#define FILETYPE_SIGNALFD 8
-
-void file_descriptor_not_used(int id)
+// ----------------------------------------------------------------------------
+
+static inline void file_descriptor_not_used(int id)
{
- if(id > 0 && id < all_files_size) {
+ if(id > 0 && id < all_files_size) {
#ifdef NETDATA_INTERNAL_CHECKS
- if(all_files[id].magic != 0x0BADCAFE) {
- error("Ignoring request to remove empty file id %d.", id);
- return;
- }
+ if(all_files[id].magic != 0x0BADCAFE) {
+ error("Ignoring request to remove empty file id %d.", id);
+ return;
+ }
#endif /* NETDATA_INTERNAL_CHECKS */
- if(debug) fprintf(stderr, "apps.plugin: decreasing slot %d (count = %d).\n", id, all_files[id].count);
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: decreasing slot %d (count = %d).\n", id, all_files[id].count);
+
+ if(all_files[id].count > 0) {
+ all_files[id].count--;
- if(all_files[id].count > 0) {
- all_files[id].count--;
+ if(!all_files[id].count) {
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: >> slot %d is empty.\n", id);
+
+ if(unlikely(file_descriptor_remove(&all_files[id]) != (void *)&all_files[id]))
+ error("INTERNAL ERROR: removal of unused fd from index, removed a different fd");
- if(!all_files[id].count) {
- if(debug) fprintf(stderr, "apps.plugin: >> slot %d is empty.\n", id);
- file_descriptor_remove(&all_files[id]);
#ifdef NETDATA_INTERNAL_CHECKS
- all_files[id].magic = 0x00000000;
+ all_files[id].magic = 0x00000000;
#endif /* NETDATA_INTERNAL_CHECKS */
- all_files_len--;
- }
- }
- else
- error("Request to decrease counter of fd %d (%s), while the use counter is 0", id, all_files[id].name);
- }
- else error("Request to decrease counter of fd %d, which is outside the array size (1 to %d)", id, all_files_size);
+ all_files_len--;
+ }
+ }
+ else
+ error("Request to decrease counter of fd %d (%s), while the use counter is 0", id, all_files[id].name);
+ }
+ else error("Request to decrease counter of fd %d, which is outside the array size (1 to %d)", id, all_files_size);
}
-int file_descriptor_find_or_add(const char *name)
-{
- static int last_pos = 0;
- uint32_t hash = simple_hash(name);
-
- if(debug) fprintf(stderr, "apps.plugin: adding or finding name '%s' with hash %u\n", name, hash);
-
- struct file_descriptor *fd = file_descriptor_find(name, hash);
- if(fd) {
- // found
- if(debug) fprintf(stderr, "apps.plugin: >> found on slot %d\n", fd->pos);
- fd->count++;
- return fd->pos;
- }
- // not found
-
- // check we have enough memory to add it
- if(!all_files || all_files_len == all_files_size) {
- void *old = all_files;
- int i;
-
- // there is no empty slot
- if(debug) fprintf(stderr, "apps.plugin: extending fd array to %d entries\n", all_files_size + FILE_DESCRIPTORS_INCREASE_STEP);
- all_files = realloc(all_files, (all_files_size + FILE_DESCRIPTORS_INCREASE_STEP) * sizeof(struct file_descriptor));
-
- // if the address changed, we have to rebuild the index
- // since all pointers are now invalid
- if(old && old != (void *)all_files) {
- if(debug) fprintf(stderr, "apps.plugin: >> re-indexing.\n");
- all_files_index.root = NULL;
- for(i = 0; i < all_files_size; i++) {
- if(!all_files[i].count) continue;
- file_descriptor_add(&all_files[i]);
- }
- if(debug) fprintf(stderr, "apps.plugin: >> re-indexing done.\n");
- }
-
- for(i = all_files_size; i < (all_files_size + FILE_DESCRIPTORS_INCREASE_STEP); i++) {
- all_files[i].count = 0;
- all_files[i].name = NULL;
+static inline void all_files_grow() {
+ void *old = all_files;
+ int i;
+
+ // there is no empty slot
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: extending fd array to %d entries\n", all_files_size + FILE_DESCRIPTORS_INCREASE_STEP);
+
+ all_files = reallocz(all_files, (all_files_size + FILE_DESCRIPTORS_INCREASE_STEP) * sizeof(struct file_descriptor));
+
+ // if the address changed, we have to rebuild the index
+ // since all pointers are now invalid
+
+ if(unlikely(old && old != (void *)all_files)) {
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: >> re-indexing.\n");
+
+ all_files_index.root = NULL;
+ for(i = 0; i < all_files_size; i++) {
+ if(!all_files[i].count) continue;
+ if(unlikely(file_descriptor_add(&all_files[i]) != (void *)&all_files[i]))
+ error("INTERNAL ERROR: duplicate indexing of fd during realloc.");
+ }
+
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: >> re-indexing done.\n");
+ }
+
+ // initialize the newly added entries
+
+ for(i = all_files_size; i < (all_files_size + FILE_DESCRIPTORS_INCREASE_STEP); i++) {
+ all_files[i].count = 0;
+ all_files[i].name = NULL;
#ifdef NETDATA_INTERNAL_CHECKS
- all_files[i].magic = 0x00000000;
+ all_files[i].magic = 0x00000000;
#endif /* NETDATA_INTERNAL_CHECKS */
- all_files[i].pos = i;
- }
+ all_files[i].pos = i;
+ }
- if(!all_files_size) all_files_len = 1;
- all_files_size += FILE_DESCRIPTORS_INCREASE_STEP;
- }
+ if(unlikely(!all_files_size)) all_files_len = 1;
+ all_files_size += FILE_DESCRIPTORS_INCREASE_STEP;
+}
+
+static inline int file_descriptor_set_on_empty_slot(const char *name, uint32_t hash, FD_FILETYPE type) {
+ // check we have enough memory to add it
+ if(!all_files || all_files_len == all_files_size)
+ all_files_grow();
- if(debug) fprintf(stderr, "apps.plugin: >> searching for empty slot.\n");
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: >> searching for empty slot.\n");
- // search for an empty slot
- int i, c;
- for(i = 0, c = last_pos ; i < all_files_size ; i++, c++) {
- if(c >= all_files_size) c = 0;
- if(c == 0) continue;
+ // search for an empty slot
- if(!all_files[c].count) {
- if(debug) fprintf(stderr, "apps.plugin: >> Examining slot %d.\n", c);
+ static int last_pos = 0;
+ int i, c;
+ for(i = 0, c = last_pos ; i < all_files_size ; i++, c++) {
+ if(c >= all_files_size) c = 0;
+ if(c == 0) continue;
+
+ if(!all_files[c].count) {
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: >> Examining slot %d.\n", c);
#ifdef NETDATA_INTERNAL_CHECKS
- if(all_files[c].magic == 0x0BADCAFE && all_files[c].name && file_descriptor_find(all_files[c].name, all_files[c].hash))
- error("fd on position %d is not cleared properly. It still has %s in it.\n", c, all_files[c].name);
+ if(all_files[c].magic == 0x0BADCAFE && all_files[c].name && file_descriptor_find(all_files[c].name, all_files[c].hash))
+ error("fd on position %d is not cleared properly. It still has %s in it.\n", c, all_files[c].name);
#endif /* NETDATA_INTERNAL_CHECKS */
- if(debug) fprintf(stderr, "apps.plugin: >> %s fd position %d for %s (last name: %s)\n", all_files[c].name?"re-using":"using", c, name, all_files[c].name);
- if(all_files[c].name) free((void *)all_files[c].name);
- all_files[c].name = NULL;
- last_pos = c;
- break;
- }
- }
- if(i == all_files_size) {
- fatal("We should find an empty slot, but there isn't any");
- exit(1);
- }
- if(debug) fprintf(stderr, "apps.plugin: >> updating slot %d.\n", c);
-
- all_files_len++;
-
- // else we have an empty slot in 'c'
-
- int type;
- if(name[0] == '/') type = FILETYPE_FILE;
- else if(strncmp(name, "pipe:", 5) == 0) type = FILETYPE_PIPE;
- else if(strncmp(name, "socket:", 7) == 0) type = FILETYPE_SOCKET;
- else if(strcmp(name, "anon_inode:inotify") == 0 || strcmp(name, "inotify") == 0) type = FILETYPE_INOTIFY;
- else if(strcmp(name, "anon_inode:[eventfd]") == 0) type = FILETYPE_EVENTFD;
- else if(strcmp(name, "anon_inode:[eventpoll]") == 0) type = FILETYPE_EVENTPOLL;
- else if(strcmp(name, "anon_inode:[timerfd]") == 0) type = FILETYPE_TIMERFD;
- else if(strcmp(name, "anon_inode:[signalfd]") == 0) type = FILETYPE_SIGNALFD;
- else if(strncmp(name, "anon_inode:", 11) == 0) {
- if(debug) fprintf(stderr, "apps.plugin: FIXME: unknown anonymous inode: %s\n", name);
- type = FILETYPE_OTHER;
- }
- else {
- if(debug) fprintf(stderr, "apps.plugin: FIXME: cannot understand linkname: %s\n", name);
- type = FILETYPE_OTHER;
- }
-
- all_files[c].name = strdup(name);
- all_files[c].hash = hash;
- all_files[c].type = type;
- all_files[c].pos = c;
- all_files[c].count = 1;
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: >> %s fd position %d for %s (last name: %s)\n", all_files[c].name?"re-using":"using", c, name, all_files[c].name);
+
+ freez((void *)all_files[c].name);
+ all_files[c].name = NULL;
+ last_pos = c;
+ break;
+ }
+ }
+
+ all_files_len++;
+
+ if(i == all_files_size) {
+ fatal("We should find an empty slot, but there isn't any");
+ exit(1);
+ }
+ // else we have an empty slot in 'c'
+
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: >> updating slot %d.\n", c);
+
+ all_files[c].name = strdupz(name);
+ all_files[c].hash = hash;
+ all_files[c].type = type;
+ all_files[c].pos = c;
+ all_files[c].count = 1;
#ifdef NETDATA_INTERNAL_CHECKS
- all_files[c].magic = 0x0BADCAFE;
+ all_files[c].magic = 0x0BADCAFE;
#endif /* NETDATA_INTERNAL_CHECKS */
- file_descriptor_add(&all_files[c]);
+ if(unlikely(file_descriptor_add(&all_files[c]) != (void *)&all_files[c]))
+ error("INTERNAL ERROR: duplicate indexing of fd.");
+
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: using fd position %d (name: %s)\n", c, all_files[c].name);
+
+ return c;
+}
+
+static inline int file_descriptor_find_or_add(const char *name)
+{
+ uint32_t hash = simple_hash(name);
+
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: adding or finding name '%s' with hash %u\n", name, hash);
+
+ struct file_descriptor *fd = file_descriptor_find(name, hash);
+ if(fd) {
+ // found
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: >> found on slot %d\n", fd->pos);
+
+ fd->count++;
+ return fd->pos;
+ }
+ // not found
+
+ FD_FILETYPE type;
+ if(likely(name[0] == '/')) type = FILETYPE_FILE;
+ else if(likely(strncmp(name, "pipe:", 5) == 0)) type = FILETYPE_PIPE;
+ else if(likely(strncmp(name, "socket:", 7) == 0)) type = FILETYPE_SOCKET;
+ else if(likely(strncmp(name, "anon_inode:", 11) == 0)) {
+ const char *t = &name[11];
+
+ if(strcmp(t, "inotify") == 0) type = FILETYPE_INOTIFY;
+ else if(strcmp(t, "[eventfd]") == 0) type = FILETYPE_EVENTFD;
+ else if(strcmp(t, "[eventpoll]") == 0) type = FILETYPE_EVENTPOLL;
+ else if(strcmp(t, "[timerfd]") == 0) type = FILETYPE_TIMERFD;
+ else if(strcmp(t, "[signalfd]") == 0) type = FILETYPE_SIGNALFD;
+ else {
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: FIXME: unknown anonymous inode: %s\n", name);
+
+ type = FILETYPE_OTHER;
+ }
+ }
+ else if(likely(strcmp(name, "inotify") == 0)) type = FILETYPE_INOTIFY;
+ else {
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: FIXME: cannot understand linkname: %s\n", name);
+
+ type = FILETYPE_OTHER;
+ }
+
+ return file_descriptor_set_on_empty_slot(name, hash, type);
+}
- if(debug) fprintf(stderr, "apps.plugin: using fd position %d (name: %s)\n", c, all_files[c].name);
+static inline void make_all_pid_fds_negative(struct pid_stat *p) {
+ int *fd = p->fds, *end = &p->fds[p->fds_size];
+ while(fd < end) {
+ *fd = -(*fd);
+ fd++;
+ }
+}
- return c;
+static inline void cleanup_negative_pid_fds(struct pid_stat *p) {
+ int *fd = p->fds, *fdend = &p->fds[p->fds_size];
+
+ while(fd < fdend) {
+ if(unlikely(*fd < 0)) {
+ file_descriptor_not_used(-(*fd));
+ *fd++ = 0;
+ }
+ else
+ fd++;
+ }
}
-int read_pid_file_descriptors(struct pid_stat *p) {
- char dirname[FILENAME_MAX+1];
-
- snprintf(dirname, FILENAME_MAX, "%s/proc/%d/fd", host_prefix, p->pid);
- DIR *fds = opendir(dirname);
- if(fds) {
- int c;
- struct dirent *de;
- char fdname[FILENAME_MAX + 1];
- char linkname[FILENAME_MAX + 1];
-
- // make the array negative
- for(c = 0 ; c < p->fds_size ; c++)
- p->fds[c] = -p->fds[c];
-
- while((de = readdir(fds))) {
- if(strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0)
- continue;
-
- // check if the fds array is small
- int fdid = atoi(de->d_name);
- if(fdid < 0) continue;
- if(fdid >= p->fds_size) {
- // it is small, extend it
- if(debug) fprintf(stderr, "apps.plugin: extending fd memory slots for %s from %d to %d\n", p->comm, p->fds_size, fdid + 100);
- p->fds = realloc(p->fds, (fdid + 100) * sizeof(int));
- if(!p->fds) {
- error("Cannot re-allocate fds for %s", p->comm);
- break;
- }
-
- // and initialize it
- for(c = p->fds_size ; c < (fdid + 100) ; c++) p->fds[c] = 0;
- p->fds_size = fdid + 100;
- }
-
- if(p->fds[fdid] == 0) {
- // we don't know this fd, get it
-
- sprintf(fdname, "%s/proc/%d/fd/%s", host_prefix, p->pid, de->d_name);
- ssize_t l = readlink(fdname, linkname, FILENAME_MAX);
- if(l == -1) {
- if(debug || (p->target && p->target->debug)) {
- if(debug || (p->target && p->target->debug))
- error("Cannot read link %s", fdname);
- }
- continue;
- }
- linkname[l] = '\0';
- file_counter++;
-
- // if another process already has this, we will get
- // the same id
- p->fds[fdid] = file_descriptor_find_or_add(linkname);
- }
-
- // else make it positive again, we need it
- // of course, the actual file may have changed, but we don't care so much
- // FIXME: we could compare the inode as returned by readdir direct structure
- else p->fds[fdid] = -p->fds[fdid];
- }
- closedir(fds);
-
- // remove all the negative file descriptors
- for(c = 0 ; c < p->fds_size ; c++) if(p->fds[c] < 0) {
- file_descriptor_not_used(-p->fds[c]);
- p->fds[c] = 0;
- }
- }
- else return 1;
-
- return 0;
+static inline void zero_pid_fds(struct pid_stat *p, int first, int size) {
+ int *fd = &p->fds[first], *end = &p->fds[first + size];
+ while(fd < end) *fd++ = 0;
+}
+
+static inline int read_pid_file_descriptors(struct pid_stat *p, void *ptr) {
+ (void)ptr;
+#ifdef __FreeBSD__
+ int mib[4];
+ size_t size;
+ struct kinfo_file *fds;
+ static char *fdsbuf;
+ char *bfdsbuf, *efdsbuf;
+ char fdsname[FILENAME_MAX + 1];
+
+ // we make all pid fds negative, so that
+ // we can detect unused file descriptors
+ // at the end, to free them
+ make_all_pid_fds_negative(p);
+
+ mib[0] = CTL_KERN;
+ mib[1] = KERN_PROC;
+ mib[2] = KERN_PROC_FILEDESC;
+ mib[3] = p->pid;
+
+ if (unlikely(sysctl(mib, 4, NULL, &size, NULL, 0))) {
+ error("sysctl error: Can't get file descriptors data size for pid %d", p->pid);
+ return 0;
+ }
+ if (likely(size > 0))
+ fdsbuf = reallocz(fdsbuf, size);
+ if (unlikely(sysctl(mib, 4, fdsbuf, &size, NULL, 0))) {
+ error("sysctl error: Can't get file descriptors data for pid %d", p->pid);
+ return 0;
+ }
+
+ bfdsbuf = fdsbuf;
+ efdsbuf = fdsbuf + size;
+ while (bfdsbuf < efdsbuf) {
+ fds = (struct kinfo_file *)(uintptr_t)bfdsbuf;
+ if (unlikely(fds->kf_structsize == 0))
+ break;
+
+ // do not process file descriptors for current working directory, root directory,
+ // jail directory, ktrace vnode, text vnode and controlling terminal
+ if (unlikely(fds->kf_fd < 0)) {
+ bfdsbuf += fds->kf_structsize;
+ continue;
+ }
+
+ // get file descriptors array index
+ int fdid = fds->kf_fd;
+
+ // check if the fds array is small
+ if (unlikely(fdid >= p->fds_size)) {
+ // it is small, extend it
+
+ if (unlikely(debug))
+ fprintf(stderr, "apps.plugin: extending fd memory slots for %s from %d to %d\n", p->comm, p->fds_size, fdid + MAX_SPARE_FDS);
+
+ p->fds = reallocz(p->fds, (fdid + MAX_SPARE_FDS) * sizeof(int));
+
+ // and initialize it
+ zero_pid_fds(p, p->fds_size, (fdid + MAX_SPARE_FDS) - p->fds_size);
+ p->fds_size = fdid + MAX_SPARE_FDS;
+ }
+
+ if (unlikely(p->fds[fdid] == 0)) {
+ // we don't know this fd, get it
+
+ switch (fds->kf_type) {
+ case KF_TYPE_FIFO:
+ case KF_TYPE_VNODE:
+ if (unlikely(!fds->kf_path[0])) {
+ sprintf(fdsname, "other: inode: %lu", fds->kf_un.kf_file.kf_file_fileid);
+ break;
+ }
+ sprintf(fdsname, "%s", fds->kf_path);
+ break;
+ case KF_TYPE_SOCKET:
+ switch (fds->kf_sock_domain) {
+ case AF_INET:
+ case AF_INET6:
+ if (fds->kf_sock_protocol == IPPROTO_TCP)
+ sprintf(fdsname, "socket: %d %lx", fds->kf_sock_protocol, fds->kf_un.kf_sock.kf_sock_inpcb);
+ else
+ sprintf(fdsname, "socket: %d %lx", fds->kf_sock_protocol, fds->kf_un.kf_sock.kf_sock_pcb);
+ break;
+ case AF_UNIX:
+ /* print address of pcb and connected pcb */
+ sprintf(fdsname, "socket: %lx %lx", fds->kf_un.kf_sock.kf_sock_pcb, fds->kf_un.kf_sock.kf_sock_unpconn);
+ break;
+ default:
+ /* print protocol number and socket address */
+ sprintf(fdsname, "socket: other: %d %s %s", fds->kf_sock_protocol, fds->kf_sa_local.__ss_pad1, fds->kf_sa_local.__ss_pad2);
+ }
+ break;
+ case KF_TYPE_PIPE:
+ sprintf(fdsname, "pipe: %lu %lu", fds->kf_un.kf_pipe.kf_pipe_addr, fds->kf_un.kf_pipe.kf_pipe_peer);
+ break;
+ case KF_TYPE_PTS:
+ sprintf(fdsname, "other: pts: %u", fds->kf_un.kf_pts.kf_pts_dev);
+ break;
+ case KF_TYPE_SHM:
+ sprintf(fdsname, "other: shm: %s size: %lu", fds->kf_path, fds->kf_un.kf_file.kf_file_size);
+ break;
+ case KF_TYPE_SEM:
+ sprintf(fdsname, "other: sem: %u", fds->kf_un.kf_sem.kf_sem_value);
+ break;
+ default:
+ sprintf(fdsname, "other: pid: %d fd: %d", fds->kf_un.kf_proc.kf_pid, fds->kf_fd);
+ }
+
+ // if another process already has this, we will get
+ // the same id
+ p->fds[fdid] = file_descriptor_find_or_add(fdsname);
+ }
+
+ // else make it positive again, we need it
+ // of course, the actual file may have changed, but we don't care so much
+ // FIXME: we could compare the inode as returned by readdir dirent structure
+
+ else
+ p->fds[fdid] = -p->fds[fdid];
+
+ bfdsbuf += fds->kf_structsize;
+ }
+#else
+ if(unlikely(!p->fds_dirname)) {
+ char dirname[FILENAME_MAX+1];
+ snprintfz(dirname, FILENAME_MAX, "%s/proc/%d/fd", netdata_configured_host_prefix, p->pid);
+ p->fds_dirname = strdupz(dirname);
+ }
+
+ DIR *fds = opendir(p->fds_dirname);
+ if(unlikely(!fds)) return 0;
+
+ struct dirent *de;
+ char fdname[FILENAME_MAX + 1];
+ char linkname[FILENAME_MAX + 1];
+
+ // we make all pid fds negative, so that
+ // we can detect unused file descriptors
+ // at the end, to free them
+ make_all_pid_fds_negative(p);
+
+ while((de = readdir(fds))) {
+ // we need only files with numeric names
+
+ if(unlikely(de->d_name[0] < '0' || de->d_name[0] > '9'))
+ continue;
+
+ // get its number
+ int fdid = (int) str2l(de->d_name);
+ if(unlikely(fdid < 0)) continue;
+
+ // check if the fds array is small
+ if(unlikely(fdid >= p->fds_size)) {
+ // it is small, extend it
+
+ if(unlikely(debug))
+ fprintf(stderr
+ , "apps.plugin: extending fd memory slots for %s from %d to %d\n"
+ , p->comm
+ , p->fds_size
+ , fdid + MAX_SPARE_FDS
+ );
+
+ p->fds = reallocz(p->fds, (fdid + MAX_SPARE_FDS) * sizeof(int));
+
+ // and initialize it
+ zero_pid_fds(p, p->fds_size, (fdid + MAX_SPARE_FDS) - p->fds_size);
+ p->fds_size = fdid + MAX_SPARE_FDS;
+ }
+
+ if(unlikely(p->fds[fdid] == 0)) {
+ // we don't know this fd, get it
+
+ sprintf(fdname, "%s/proc/%d/fd/%s", netdata_configured_host_prefix, p->pid, de->d_name);
+ ssize_t l = readlink(fdname, linkname, FILENAME_MAX);
+ if(unlikely(l == -1)) {
+ if(debug || (p->target && p->target->debug)) {
+ if(debug || (p->target && p->target->debug))
+ error("Cannot read link %s", fdname);
+ }
+ continue;
+ }
+ else
+ linkname[l] = '\0';
+
+ file_counter++;
+
+ // if another process already has this, we will get
+ // the same id
+ p->fds[fdid] = file_descriptor_find_or_add(linkname);
+ }
+
+ // else make it positive again, we need it
+ // of course, the actual file may have changed, but we don't care so much
+ // FIXME: we could compare the inode as returned by readdir dirent structure
+ // UPDATE: no we cannot use inodes - under /proc inodes don't change when the link is changed
+
+ else
+ p->fds[fdid] = -p->fds[fdid];
+ }
+
+ closedir(fds);
+#endif
+ cleanup_negative_pid_fds(p);
+
+ return 1;
+}
+
+// ----------------------------------------------------------------------------
+
+static inline int print_process_and_parents(struct pid_stat *p, usec_t time) {
+ char *prefix = "\\_ ";
+ int indent = 0;
+
+ if(p->parent)
+ indent = print_process_and_parents(p->parent, p->stat_collected_usec);
+ else
+ prefix = " > ";
+
+ char buffer[indent + 1];
+ int i;
+
+ for(i = 0; i < indent ;i++) buffer[i] = ' ';
+ buffer[i] = '\0';
+
+ fprintf(stderr, " %s %s%s (%d %s %llu"
+ , buffer
+ , prefix
+ , p->comm
+ , p->pid
+ , p->updated?"running":"exited"
+ , p->stat_collected_usec - time
+ );
+
+ if(p->utime) fprintf(stderr, " utime=" KERNEL_UINT_FORMAT, p->utime);
+ if(p->stime) fprintf(stderr, " stime=" KERNEL_UINT_FORMAT, p->stime);
+ if(p->gtime) fprintf(stderr, " gtime=" KERNEL_UINT_FORMAT, p->gtime);
+ if(p->cutime) fprintf(stderr, " cutime=" KERNEL_UINT_FORMAT, p->cutime);
+ if(p->cstime) fprintf(stderr, " cstime=" KERNEL_UINT_FORMAT, p->cstime);
+ if(p->cgtime) fprintf(stderr, " cgtime=" KERNEL_UINT_FORMAT, p->cgtime);
+ if(p->minflt) fprintf(stderr, " minflt=" KERNEL_UINT_FORMAT, p->minflt);
+ if(p->cminflt) fprintf(stderr, " cminflt=" KERNEL_UINT_FORMAT, p->cminflt);
+ if(p->majflt) fprintf(stderr, " majflt=" KERNEL_UINT_FORMAT, p->majflt);
+ if(p->cmajflt) fprintf(stderr, " cmajflt=" KERNEL_UINT_FORMAT, p->cmajflt);
+ fprintf(stderr, ")\n");
+
+ return indent + 1;
+}
+
+static inline void print_process_tree(struct pid_stat *p, char *msg) {
+ log_date(stderr);
+ fprintf(stderr, "%s: process %s (%d, %s) with parents:\n", msg, p->comm, p->pid, p->updated?"running":"exited");
+ print_process_and_parents(p, p->stat_collected_usec);
+}
+
+static inline void find_lost_child_debug(struct pid_stat *pe, kernel_uint_t lost, int type) {
+ int found = 0;
+ struct pid_stat *p = NULL;
+
+ for(p = root_of_pids; p ; p = p->next) {
+ if(p == pe) continue;
+
+ switch(type) {
+ case 1:
+ if(p->cminflt > lost) {
+ fprintf(stderr, " > process %d (%s) could use the lost exited child minflt " KERNEL_UINT_FORMAT " of process %d (%s)\n", p->pid, p->comm, lost, pe->pid, pe->comm);
+ found++;
+ }
+ break;
+
+ case 2:
+ if(p->cmajflt > lost) {
+ fprintf(stderr, " > process %d (%s) could use the lost exited child majflt " KERNEL_UINT_FORMAT " of process %d (%s)\n", p->pid, p->comm, lost, pe->pid, pe->comm);
+ found++;
+ }
+ break;
+
+ case 3:
+ if(p->cutime > lost) {
+ fprintf(stderr, " > process %d (%s) could use the lost exited child utime " KERNEL_UINT_FORMAT " of process %d (%s)\n", p->pid, p->comm, lost, pe->pid, pe->comm);
+ found++;
+ }
+ break;
+
+ case 4:
+ if(p->cstime > lost) {
+ fprintf(stderr, " > process %d (%s) could use the lost exited child stime " KERNEL_UINT_FORMAT " of process %d (%s)\n", p->pid, p->comm, lost, pe->pid, pe->comm);
+ found++;
+ }
+ break;
+
+ case 5:
+ if(p->cgtime > lost) {
+ fprintf(stderr, " > process %d (%s) could use the lost exited child gtime " KERNEL_UINT_FORMAT " of process %d (%s)\n", p->pid, p->comm, lost, pe->pid, pe->comm);
+ found++;
+ }
+ break;
+ }
+ }
+
+ if(!found) {
+ switch(type) {
+ case 1:
+ fprintf(stderr, " > cannot find any process to use the lost exited child minflt " KERNEL_UINT_FORMAT " of process %d (%s)\n", lost, pe->pid, pe->comm);
+ break;
+
+ case 2:
+ fprintf(stderr, " > cannot find any process to use the lost exited child majflt " KERNEL_UINT_FORMAT " of process %d (%s)\n", lost, pe->pid, pe->comm);
+ break;
+
+ case 3:
+ fprintf(stderr, " > cannot find any process to use the lost exited child utime " KERNEL_UINT_FORMAT " of process %d (%s)\n", lost, pe->pid, pe->comm);
+ break;
+
+ case 4:
+ fprintf(stderr, " > cannot find any process to use the lost exited child stime " KERNEL_UINT_FORMAT " of process %d (%s)\n", lost, pe->pid, pe->comm);
+ break;
+
+ case 5:
+ fprintf(stderr, " > cannot find any process to use the lost exited child gtime " KERNEL_UINT_FORMAT " of process %d (%s)\n", lost, pe->pid, pe->comm);
+ break;
+ }
+ }
+}
+
+static inline kernel_uint_t remove_exited_child_from_parent(kernel_uint_t *field, kernel_uint_t *pfield) {
+ kernel_uint_t absorbed = 0;
+
+ if(*field > *pfield) {
+ absorbed += *pfield;
+ *field -= *pfield;
+ *pfield = 0;
+ }
+ else {
+ absorbed += *field;
+ *pfield -= *field;
+ *field = 0;
+ }
+
+ return absorbed;
+}
+
+static inline void process_exited_processes() {
+ struct pid_stat *p;
+
+ for(p = root_of_pids; p ; p = p->next) {
+ if(p->updated || !p->stat_collected_usec)
+ continue;
+
+ kernel_uint_t utime = (p->utime_raw + p->cutime_raw) * (USEC_PER_SEC * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
+ kernel_uint_t stime = (p->stime_raw + p->cstime_raw) * (USEC_PER_SEC * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
+ kernel_uint_t gtime = (p->gtime_raw + p->cgtime_raw) * (USEC_PER_SEC * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
+ kernel_uint_t minflt = (p->minflt_raw + p->cminflt_raw) * (USEC_PER_SEC * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
+ kernel_uint_t majflt = (p->majflt_raw + p->cmajflt_raw) * (USEC_PER_SEC * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
+
+ if(utime + stime + gtime + minflt + majflt == 0)
+ continue;
+
+ if(unlikely(debug)) {
+ log_date(stderr);
+ fprintf(stderr, "Absorb %s (%d %s total resources: utime=" KERNEL_UINT_FORMAT " stime=" KERNEL_UINT_FORMAT " gtime=" KERNEL_UINT_FORMAT " minflt=" KERNEL_UINT_FORMAT " majflt=" KERNEL_UINT_FORMAT ")\n"
+ , p->comm
+ , p->pid
+ , p->updated?"running":"exited"
+ , utime
+ , stime
+ , gtime
+ , minflt
+ , majflt
+ );
+ print_process_tree(p, "Searching parents");
+ }
+
+ struct pid_stat *pp;
+ for(pp = p->parent; pp ; pp = pp->parent) {
+ if(!pp->updated) continue;
+
+ kernel_uint_t absorbed;
+ absorbed = remove_exited_child_from_parent(&utime, &pp->cutime);
+ if(unlikely(debug && absorbed))
+ fprintf(stderr, " > process %s (%d %s) absorbed " KERNEL_UINT_FORMAT " utime (remaining: " KERNEL_UINT_FORMAT ")\n", pp->comm, pp->pid, pp->updated?"running":"exited", absorbed, utime);
+
+ absorbed = remove_exited_child_from_parent(&stime, &pp->cstime);
+ if(unlikely(debug && absorbed))
+ fprintf(stderr, " > process %s (%d %s) absorbed " KERNEL_UINT_FORMAT " stime (remaining: " KERNEL_UINT_FORMAT ")\n", pp->comm, pp->pid, pp->updated?"running":"exited", absorbed, stime);
+
+ absorbed = remove_exited_child_from_parent(>ime, &pp->cgtime);
+ if(unlikely(debug && absorbed))
+ fprintf(stderr, " > process %s (%d %s) absorbed " KERNEL_UINT_FORMAT " gtime (remaining: " KERNEL_UINT_FORMAT ")\n", pp->comm, pp->pid, pp->updated?"running":"exited", absorbed, gtime);
+
+ absorbed = remove_exited_child_from_parent(&minflt, &pp->cminflt);
+ if(unlikely(debug && absorbed))
+ fprintf(stderr, " > process %s (%d %s) absorbed " KERNEL_UINT_FORMAT " minflt (remaining: " KERNEL_UINT_FORMAT ")\n", pp->comm, pp->pid, pp->updated?"running":"exited", absorbed, minflt);
+
+ absorbed = remove_exited_child_from_parent(&majflt, &pp->cmajflt);
+ if(unlikely(debug && absorbed))
+ fprintf(stderr, " > process %s (%d %s) absorbed " KERNEL_UINT_FORMAT " majflt (remaining: " KERNEL_UINT_FORMAT ")\n", pp->comm, pp->pid, pp->updated?"running":"exited", absorbed, majflt);
+ }
+
+ if(unlikely(utime + stime + gtime + minflt + majflt > 0)) {
+ if(unlikely(debug)) {
+ if(utime) find_lost_child_debug(p, utime, 3);
+ if(stime) find_lost_child_debug(p, stime, 4);
+ if(gtime) find_lost_child_debug(p, gtime, 5);
+ if(minflt) find_lost_child_debug(p, minflt, 1);
+ if(majflt) find_lost_child_debug(p, majflt, 2);
+ }
+
+ p->keep = 1;
+
+ if(unlikely(debug))
+ fprintf(stderr, " > remaining resources - KEEP - for another loop: %s (%d %s total resources: utime=" KERNEL_UINT_FORMAT " stime=" KERNEL_UINT_FORMAT " gtime=" KERNEL_UINT_FORMAT " minflt=" KERNEL_UINT_FORMAT " majflt=" KERNEL_UINT_FORMAT ")\n"
+ , p->comm
+ , p->pid
+ , p->updated?"running":"exited"
+ , utime
+ , stime
+ , gtime
+ , minflt
+ , majflt
+ );
+
+ for(pp = p->parent; pp ; pp = pp->parent) {
+ if(pp->updated) break;
+ pp->keep = 1;
+
+ if(unlikely(debug))
+ fprintf(stderr, " > - KEEP - parent for another loop: %s (%d %s)\n"
+ , pp->comm
+ , pp->pid
+ , pp->updated?"running":"exited"
+ );
+ }
+
+ p->utime_raw = utime * (p->stat_collected_usec - p->last_stat_collected_usec) / (USEC_PER_SEC * RATES_DETAIL);
+ p->stime_raw = stime * (p->stat_collected_usec - p->last_stat_collected_usec) / (USEC_PER_SEC * RATES_DETAIL);
+ p->gtime_raw = gtime * (p->stat_collected_usec - p->last_stat_collected_usec) / (USEC_PER_SEC * RATES_DETAIL);
+ p->minflt_raw = minflt * (p->stat_collected_usec - p->last_stat_collected_usec) / (USEC_PER_SEC * RATES_DETAIL);
+ p->majflt_raw = majflt * (p->stat_collected_usec - p->last_stat_collected_usec) / (USEC_PER_SEC * RATES_DETAIL);
+ p->cutime_raw = p->cstime_raw = p->cgtime_raw = p->cminflt_raw = p->cmajflt_raw = 0;
+
+ if(unlikely(debug))
+ fprintf(stderr, "\n");
+ }
+ else if(unlikely(debug)) {
+ fprintf(stderr, " > totally absorbed - DONE - %s (%d %s)\n"
+ , p->comm
+ , p->pid
+ , p->updated?"running":"exited"
+ );
+ }
+ }
+}
+
+static inline void link_all_processes_to_their_parents(void) {
+ struct pid_stat *p, *pp;
+
+ // link all children to their parents
+ // and update children count on parents
+ for(p = root_of_pids; p ; p = p->next) {
+ // for each process found
+
+ p->sortlist = 0;
+ p->parent = NULL;
+
+ if(unlikely(!p->ppid)) {
+ p->parent = NULL;
+ continue;
+ }
+
+ pp = all_pids[p->ppid];
+ if(likely(pp)) {
+ p->parent = pp;
+ pp->children_count++;
+
+ if(unlikely(debug || (p->target && p->target->debug)))
+ fprintf(stderr, "apps.plugin: \tchild %d (%s, %s) on target '%s' has parent %d (%s, %s). Parent: utime=" KERNEL_UINT_FORMAT ", stime=" KERNEL_UINT_FORMAT ", gtime=" KERNEL_UINT_FORMAT ", minflt=" KERNEL_UINT_FORMAT ", majflt=" KERNEL_UINT_FORMAT ", cutime=" KERNEL_UINT_FORMAT ", cstime=" KERNEL_UINT_FORMAT ", cgtime=" KERNEL_UINT_FORMAT ", cminflt=" KERNEL_UINT_FORMAT ", cmajflt=" KERNEL_UINT_FORMAT "\n", p->pid, p->comm, p->updated?"running":"exited", (p->target)?p->target->name:"UNSET", pp->pid, pp->comm, pp->updated?"running":"exited", pp->utime, pp->stime, pp->gtime, pp->minflt, pp->majflt, pp->cutime, pp->cstime, pp->cgtime, pp->cminflt, pp->cmajflt);
+ }
+ else {
+ p->parent = NULL;
+ error("pid %d %s states parent %d, but the later does not exist.", p->pid, p->comm, p->ppid);
+ }
+ }
}
// ----------------------------------------------------------------------------
// to avoid filling up all disk space
// if debug is enabled, all errors are printed
-int collect_data_for_all_processes_from_proc(void)
-{
- char dirname[FILENAME_MAX + 1];
+static int compar_pid(const void *pid1, const void *pid2) {
- snprintf(dirname, FILENAME_MAX, "%s/proc", host_prefix);
- DIR *dir = opendir(dirname);
- if(!dir) return 0;
+ struct pid_stat *p1 = all_pids[*((pid_t *)pid1)];
+ struct pid_stat *p2 = all_pids[*((pid_t *)pid2)];
- struct dirent *file = NULL;
- struct pid_stat *p = NULL;
+ if(p1->sortlist > p2->sortlist)
+ return -1;
+ else
+ return 1;
+}
- // mark them all as un-updated
- all_pids_count = 0;
- for(p = root_of_pids; p ; p = p->next) {
- all_pids_count++;
- p->parent = NULL;
- p->updated = 0;
- p->children_count = 0;
- p->merged = 0;
- p->new_entry = 0;
+static inline int collect_data_for_pid(pid_t pid, void *ptr) {
+ if(unlikely(pid < INIT_PID || pid > pid_max)) {
+ error("Invalid pid %d read (expected %d to %d). Ignoring process.", pid, INIT_PID, pid_max);
+ return 0;
+ }
- p->last_minflt = p->minflt;
- p->last_cminflt = p->cminflt;
- p->last_majflt = p->majflt;
- p->last_cmajflt = p->cmajflt;
- p->last_utime = p->utime;
- p->last_stime = p->stime;
- p->last_cutime = p->cutime;
- p->last_cstime = p->cstime;
-
- p->last_io_logical_bytes_read = p->io_logical_bytes_read;
- p->last_io_logical_bytes_written = p->io_logical_bytes_written;
- p->last_io_read_calls = p->io_read_calls;
- p->last_io_write_calls = p->io_write_calls;
- p->last_io_storage_bytes_read = p->io_storage_bytes_read;
- p->last_io_storage_bytes_written = p->io_storage_bytes_written;
- p->last_io_cancelled_write_bytes = p->io_cancelled_write_bytes;
- }
-
- while((file = readdir(dir))) {
- char *endptr = file->d_name;
- pid_t pid = (pid_t) strtoul(file->d_name, &endptr, 10);
-
- // make sure we read a valid number
- if(unlikely(pid <= 0 || pid > pid_max || endptr == file->d_name || *endptr != '\0'))
- continue;
-
- p = get_pid_entry(pid);
- if(unlikely(!p)) continue;
-
-
- // --------------------------------------------------------------------
- // /proc/<pid>/stat
-
- if(unlikely(read_proc_pid_stat(p))) {
- error("Cannot process %s/proc/%d/stat", host_prefix, pid);
-
- // there is no reason to proceed if we cannot get its status
- continue;
- }
-
- // check its parent pid
- if(unlikely(p->ppid < 0 || p->ppid > pid_max)) {
- error("Pid %d states invalid parent pid %d. Using 0.", pid, p->ppid);
-
- p->ppid = 0;
- }
-
- // --------------------------------------------------------------------
- // /proc/<pid>/cmdline
-
- if(proc_pid_cmdline_is_needed) {
- if(unlikely(read_proc_pid_cmdline(p))) {
- error("Cannot process %s/proc/%d/cmdline", host_prefix, pid);
- }
- }
-
- // --------------------------------------------------------------------
- // /proc/<pid>/statm
-
- if(unlikely(read_proc_pid_statm(p))) {
- error("Cannot process %s/proc/%d/statm", host_prefix, pid);
-
- // there is no reason to proceed if we cannot get its memory status
- continue;
- }
-
-
- // --------------------------------------------------------------------
- // /proc/<pid>/io
-
- if(unlikely(read_proc_pid_io(p))) {
- error("Cannot process %s/proc/%d/io", host_prefix, pid);
-
- // on systems without /proc/X/io
- // allow proceeding without I/O information
- // continue;
- }
-
- // --------------------------------------------------------------------
- // <pid> ownership
-
- if(unlikely(read_proc_pid_ownership(p))) {
- error("Cannot stat %s/proc/%d", host_prefix, pid);
- }
-
- // --------------------------------------------------------------------
- // link it
-
- // check if it is target
- // we do this only once, the first time this pid is loaded
- if(unlikely(p->new_entry)) {
- if(debug) fprintf(stderr, "apps.plugin: \tJust added %s\n", p->comm);
- uint32_t hash = simple_hash(p->comm);
- size_t pclen = strlen(p->comm);
-
- struct target *w;
- for(w = apps_groups_root_target; w ; w = w->next) {
- // if(debug || (p->target && p->target->debug)) fprintf(stderr, "apps.plugin: \t\tcomparing '%s' with '%s'\n", w->compare, p->comm);
-
- // find it - 4 cases:
- // 1. the target is not a pattern
- // 2. the target has the prefix
- // 3. the target has the suffix
- // 4. the target is something inside cmdline
- if( (!w->starts_with && !w->ends_with && w->comparehash == hash && !strcmp(w->compare, p->comm))
- || (w->starts_with && !w->ends_with && !strncmp(w->compare, p->comm, w->comparelen))
- || (!w->starts_with && w->ends_with && pclen >= w->comparelen && !strcmp(w->compare, &p->comm[pclen - w->comparelen]))
- || (proc_pid_cmdline_is_needed && w->starts_with && w->ends_with && strstr(p->cmdline, w->compare))
- ) {
- if(w->target) p->target = w->target;
- else p->target = w;
+ struct pid_stat *p = get_pid_entry(pid);
+ if(unlikely(!p || p->read)) return 0;
+ p->read = 1;
- if(debug || (p->target && p->target->debug))
- fprintf(stderr, "apps.plugin: \t\t%s linked to target %s\n", p->comm, p->target->name);
- }
- }
- }
+ // fprintf(stderr, "Reading process %d (%s), sortlist %d\n", p->pid, p->comm, p->sortlist);
- // --------------------------------------------------------------------
- // /proc/<pid>/fd
+ // --------------------------------------------------------------------
+ // /proc/<pid>/stat
- if(unlikely(read_pid_file_descriptors(p))) {
- error("Cannot process entries in %s/proc/%d/fd", host_prefix, pid);
- }
+ if(unlikely(!managed_log(p, PID_LOG_STAT, read_proc_pid_stat(p, ptr))))
+ // there is no reason to proceed if we cannot get its status
+ return 0;
- // --------------------------------------------------------------------
- // done!
+ read_proc_pid_ownership(p, ptr);
- // mark it as updated
- p->updated = 1;
- }
+ // check its parent pid
+ if(unlikely(p->ppid < 0 || p->ppid > pid_max)) {
+ error("Pid %d (command '%s') states invalid parent pid %d. Using 0.", pid, p->comm, p->ppid);
+ p->ppid = 0;
+ }
- closedir(dir);
+ // --------------------------------------------------------------------
+ // /proc/<pid>/io
- return 1;
-}
+ managed_log(p, PID_LOG_IO, read_proc_pid_io(p, ptr));
+ // --------------------------------------------------------------------
+ // /proc/<pid>/statm
-// ----------------------------------------------------------------------------
+ if(unlikely(!managed_log(p, PID_LOG_STATM, read_proc_pid_statm(p, ptr))))
+ // there is no reason to proceed if we cannot get its memory status
+ return 0;
+
+ // --------------------------------------------------------------------
+ // /proc/<pid>/fd
+
+ if(enable_file_charts)
+ managed_log(p, PID_LOG_FDS, read_pid_file_descriptors(p, ptr));
-#ifdef AGGREGATE_CHILDREN_TO_PARENTS
-// print a tree view of all processes
-int debug_childrens_aggregations(pid_t pid, int level) {
- struct pid_stat *p = NULL;
- char b[level+3];
- int i, ret = 0;
-
- for(i = 0; i < level; i++) b[i] = '\t';
- b[level] = '|';
- b[level+1] = '-';
- b[level+2] = '\0';
-
- for(p = root_of_pids; p ; p = p->next) {
- if(p->ppid == pid) {
- ret += debug_childrens_aggregations(p->pid, level+1);
- }
- }
-
- p = all_pids[pid];
- if(p) {
- if(!p->updated) ret += 1;
- if(ret) fprintf(stderr, "%s %s %d [%s, %s] c=%d u=%llu+%llu, s=%llu+%llu, cu=%llu+%llu, cs=%llu+%llu, n=%llu+%llu, j=%llu+%llu, cn=%llu+%llu, cj=%llu+%llu\n"
- , b, p->comm, p->pid, p->updated?"OK":"KILLED", p->target->name, p->children_count
- , p->utime, p->utime - p->old_utime
- , p->stime, p->stime - p->old_stime
- , p->cutime, p->cutime - p->old_cutime
- , p->cstime, p->cstime - p->old_cstime
- , p->minflt, p->minflt - p->old_minflt
- , p->majflt, p->majflt - p->old_majflt
- , p->cminflt, p->cminflt - p->old_cminflt
- , p->cmajflt, p->cmajflt - p->old_cmajflt
- );
- }
-
- return ret;
+ // --------------------------------------------------------------------
+ // done!
+
+ if(unlikely(debug && include_exited_childs && all_pids_count && p->ppid && all_pids[p->ppid] && !all_pids[p->ppid]->read))
+ fprintf(stderr, "Read process %d (%s) sortlisted %d, but its parent %d (%s) sortlisted %d, is not read\n", p->pid, p->comm, p->sortlist, all_pids[p->ppid]->pid, all_pids[p->ppid]->comm, all_pids[p->ppid]->sortlist);
+
+ // mark it as updated
+ p->updated = 1;
+ p->keep = 0;
+ p->keeploops = 0;
+
+ return 1;
}
-#endif /* AGGREGATE_CHILDREN_TO_PARENTS */
+static int collect_data_for_all_processes(void) {
+ struct pid_stat *p = NULL;
+
+#ifdef __FreeBSD__
+ int i, procnum;
+ size_t procbase_size;
+ static struct kinfo_proc *procbase;
+
+ int mib[3];
+
+ mib[0] = CTL_KERN;
+ mib[1] = KERN_PROC;
+ mib[2] = KERN_PROC_PROC;
+ if (unlikely(sysctl(mib, 3, NULL, &procbase_size, NULL, 0))) {
+ error("sysctl error: Can't get processes data size");
+ return 0;
+ }
+ procbase = reallocz(procbase, procbase_size);
+ if (unlikely(sysctl(mib, 3, procbase, &procbase_size, NULL, 0))) {
+ error("sysctl error: Can't get processes data");
+ return 0;
+ }
+ procnum = procbase_size / sizeof(struct kinfo_proc);
+#endif
+
+ if(all_pids_count) {
+ size_t slc = 0;
+ for(p = root_of_pids; p ; p = p->next) {
+ p->read = 0; // mark it as not read, so that collect_data_for_pid() will read it
+ p->updated = 0;
+ p->merged = 0;
+ p->children_count = 0;
+ p->parent = NULL;
+
+#if (ALL_PIDS_ARE_READ_INSTANTLY == 0)
+ all_pids_sortlist[slc++] = p->pid;
+#endif
+ }
+
+#if (ALL_PIDS_ARE_READ_INSTANTLY == 0)
+ if(unlikely(slc != all_pids_count)) {
+ error("Internal error: I was thinking I had %zu processes in my arrays, but it seems there are more.", all_pids_count);
+ all_pids_count = slc;
+ }
+
+ if(include_exited_childs) {
+ // Read parents before childs
+ // This is needed to prevent a situation where
+ // a child is found running, but until we read
+ // its parent, it has exited and its parent
+ // has accumulated its resources.
+
+ qsort((void *)all_pids_sortlist, (size_t)all_pids_count, sizeof(pid_t), compar_pid);
+
+ // we forward read all running processes
+ // collect_data_for_pid() is smart enough,
+ // not to read the same pid twice per iteration
+ for(slc = 0; slc < all_pids_count; slc++)
+ collect_data_for_pid(all_pids_sortlist[slc], NULL);
+ }
+#endif
+ }
+
+#ifdef __FreeBSD__
+ for (i = INIT_PID; i < procnum - INIT_PID; ++i) {
+ pid_t pid = procbase[i].ki_pid;
+ collect_data_for_pid(pid, &procbase[i]);
+ }
+#else
+ char dirname[FILENAME_MAX + 1];
+
+ snprintfz(dirname, FILENAME_MAX, "%s/proc", netdata_configured_host_prefix);
+ DIR *dir = opendir(dirname);
+ if(!dir) return 0;
+
+ struct dirent *de = NULL;
+
+ while((de = readdir(dir))) {
+ char *endptr = de->d_name;
+
+ if(unlikely(de->d_type != DT_DIR || de->d_name[0] < '0' || de->d_name[0] > '9'))
+ continue;
+
+ pid_t pid = (pid_t) strtoul(de->d_name, &endptr, 10);
+
+ // make sure we read a valid number
+ if(unlikely(endptr == de->d_name || *endptr != '\0'))
+ continue;
+
+ collect_data_for_pid(pid, NULL);
+ }
+ closedir(dir);
+#endif
+
+ if(!all_pids_count)
+ return 0;
+ // we need /proc/stat to normalize the cpu consumption of the exited childs
+ read_proc_stat();
+
+ // build the process tree
+ link_all_processes_to_their_parents();
+
+ // normally this is done
+ // however we may have processes exited while we collected values
+ // so let's find the exited ones
+ // we do this by collecting the ownership of process
+ // if we manage to get the ownership, the process still runs
+ process_exited_processes();
+
+ return 1;
+}
// ----------------------------------------------------------------------------
// update statistics on the targets
// 9. find the unique file count for each target
// check: update_apps_groups_statistics()
-void link_all_processes_to_their_parents(void) {
- struct pid_stat *p = NULL;
-
- // link all children to their parents
- // and update children count on parents
- for(p = root_of_pids; p ; p = p->next) {
- // for each process found running
-
- if(p->ppid > 0
- && p->ppid <= pid_max
- && all_pids[p->ppid]
- ) {
- // for valid processes
-
- if(debug || (p->target && p->target->debug))
- fprintf(stderr, "apps.plugin: \tparent of %d (%s) is %d (%s)\n", p->pid, p->comm, p->ppid, all_pids[p->ppid]->comm);
-
- p->parent = all_pids[p->ppid];
- p->parent->children_count++;
- }
- else if(p->ppid != 0)
- error("pid %d %s states parent %d, but the later does not exist.", p->pid, p->comm, p->ppid);
- }
-}
-
-#ifdef AGGREGATE_CHILDREN_TO_PARENTS
-void aggregate_children_to_parents(void) {
- struct pid_stat *p = NULL;
-
- // for each killed process, remove its values from the parents
- // sums (we had already added them in a previous loop)
- for(p = root_of_pids; p ; p = p->next) {
- if(p->updated) continue;
-
- if(debug) fprintf(stderr, "apps.plugin: UNMERGING %d %s\n", p->pid, p->comm);
-
- unsigned long long diff_utime = p->utime + p->cutime + p->fix_cutime;
- unsigned long long diff_stime = p->stime + p->cstime + p->fix_cstime;
- unsigned long long diff_minflt = p->minflt + p->cminflt + p->fix_cminflt;
- unsigned long long diff_majflt = p->majflt + p->cmajflt + p->fix_cmajflt;
-
- struct pid_stat *t = p;
- while((t = t->parent)) {
- if(!t->updated) continue;
-
- unsigned long long x;
- if(diff_utime && t->diff_cutime) {
- x = (t->diff_cutime < diff_utime)?t->diff_cutime:diff_utime;
- diff_utime -= x;
- t->diff_cutime -= x;
- t->fix_cutime += x;
- if(debug) fprintf(stderr, "apps.plugin: \t cutime %llu from %d %s %s\n", x, t->pid, t->comm, t->target->name);
- }
- if(diff_stime && t->diff_cstime) {
- x = (t->diff_cstime < diff_stime)?t->diff_cstime:diff_stime;
- diff_stime -= x;
- t->diff_cstime -= x;
- t->fix_cstime += x;
- if(debug) fprintf(stderr, "apps.plugin: \t cstime %llu from %d %s %s\n", x, t->pid, t->comm, t->target->name);
- }
- if(diff_minflt && t->diff_cminflt) {
- x = (t->diff_cminflt < diff_minflt)?t->diff_cminflt:diff_minflt;
- diff_minflt -= x;
- t->diff_cminflt -= x;
- t->fix_cminflt += x;
- if(debug) fprintf(stderr, "apps.plugin: \t cminflt %llu from %d %s %s\n", x, t->pid, t->comm, t->target->name);
- }
- if(diff_majflt && t->diff_cmajflt) {
- x = (t->diff_cmajflt < diff_majflt)?t->diff_cmajflt:diff_majflt;
- diff_majflt -= x;
- t->diff_cmajflt -= x;
- t->fix_cmajflt += x;
- if(debug) fprintf(stderr, "apps.plugin: \t cmajflt %llu from %d %s %s\n", x, t->pid, t->comm, t->target->name);
- }
- }
-
- if(diff_utime) error("Cannot fix up utime %llu", diff_utime);
- if(diff_stime) error("Cannot fix up stime %llu", diff_stime);
- if(diff_minflt) error("Cannot fix up minflt %llu", diff_minflt);
- if(diff_majflt) error("Cannot fix up majflt %llu", diff_majflt);
- }
-}
-#endif /* AGGREGATE_CHILDREN_TO_PARENTS */
-
-void cleanup_non_existing_pids(void) {
- int c;
- struct pid_stat *p = NULL;
-
- for(p = root_of_pids; p ;) {
- if(!p->updated) {
-// fprintf(stderr, "\tEXITED %d %s [parent %d %s, target %s] utime=%llu, stime=%llu, cutime=%llu, cstime=%llu, minflt=%llu, majflt=%llu, cminflt=%llu, cmajflt=%llu\n", p->pid, p->comm, p->parent->pid, p->parent->comm, p->target->name, p->utime, p->stime, p->cutime, p->cstime, p->minflt, p->majflt, p->cminflt, p->cmajflt);
-
- for(c = 0 ; c < p->fds_size ; c++) if(p->fds[c] > 0) {
- file_descriptor_not_used(p->fds[c]);
- p->fds[c] = 0;
- }
-
- pid_t r = p->pid;
- p = p->next;
- del_pid_entry(r);
- }
- else p = p->next;
- }
-}
-
-void apply_apps_groups_targets_inheritance(void) {
- struct pid_stat *p = NULL;
-
- // children that do not have a target
- // inherit their target from their parent
- int found = 1;
- while(found) {
- found = 0;
- for(p = root_of_pids; p ; p = p->next) {
- // if this process does not have a target
- // and it has a parent
- // and its parent has a target
- // then, set the parent's target to this process
- if(unlikely(!p->target && p->parent && p->parent->target)) {
- p->target = p->parent->target;
- found++;
-
- if(debug || (p->target && p->target->debug))
- fprintf(stderr, "apps.plugin: \t\tTARGET INHERITANCE: %s is inherited by %d (%s) from its parent %d (%s).\n", p->target->name, p->pid, p->comm, p->parent->pid, p->parent->comm);
- }
- }
- }
-
-
- // find all the procs with 0 childs and merge them to their parents
- // repeat, until nothing more can be done.
- found = 1;
- while(found) {
- found = 0;
- for(p = root_of_pids; p ; p = p->next) {
- // if this process does not have any children
- // and is not already merged
- // and has a parent
- // and its parent has children
- // and the target of this process and its parent is the same, or the parent does not have a target
- // and its parent is not init
- // then, mark them as merged.
- if(unlikely(
- !p->children_count
- && !p->merged
- && p->parent
- && p->parent->children_count
- && (p->target == p->parent->target || !p->parent->target)
- && p->ppid != 1
- )) {
- p->parent->children_count--;
- p->merged = 1;
-
- // the parent inherits the child's target, if it does not have a target itself
- if(unlikely(p->target && !p->parent->target)) {
- p->parent->target = p->target;
-
- if(debug || (p->target && p->target->debug))
- fprintf(stderr, "apps.plugin: \t\tTARGET INHERITANCE: %s is inherited by %d (%s) from its child %d (%s).\n", p->target->name, p->parent->pid, p->parent->comm, p->pid, p->comm);
- }
-
- found++;
- }
- }
-
- if(debug)
- fprintf(stderr, "apps.plugin: merged %d processes\n", found);
- }
-
- // init goes always to default target
- if(all_pids[1])
- all_pids[1]->target = apps_groups_default_target;
-
- // give a default target on all top level processes
- for(p = root_of_pids; p ; p = p->next) {
- // if the process is not merged itself
- // then is is a top level process
- if(!p->merged && !p->target)
- p->target = apps_groups_default_target;
-
-#ifdef AGGREGATE_CHILDREN_TO_PARENTS
- // by the way, update the diffs
- // will be used later for subtracting killed process times
- p->diff_cutime = p->utime - p->cutime;
- p->diff_cstime = p->stime - p->cstime;
- p->diff_cminflt = p->minflt - p->cminflt;
- p->diff_cmajflt = p->majflt - p->cmajflt;
-#endif /* AGGREGATE_CHILDREN_TO_PARENTS */
- }
-
- // give a target to all merged child processes
- found = 1;
- while(found) {
- found = 0;
- for(p = root_of_pids; p ; p = p->next) {
- if(unlikely(!p->target && p->merged && p->parent && p->parent->target)) {
- p->target = p->parent->target;
- found++;
-
- if(debug || (p->target && p->target->debug))
- fprintf(stderr, "apps.plugin: \t\tTARGET INHERITANCE: %s is inherited by %d (%s) from its parent %d (%s) at phase 2.\n", p->target->name, p->pid, p->comm, p->parent->pid, p->parent->comm);
- }
- }
- }
+static void cleanup_exited_pids(void) {
+ int c;
+ struct pid_stat *p = NULL;
+
+ for(p = root_of_pids; p ;) {
+ if(!p->updated && (!p->keep || p->keeploops > 0)) {
+ if(unlikely(debug && (p->keep || p->keeploops)))
+ fprintf(stderr, " > CLEANUP cannot keep exited process %d (%s) anymore - removing it.\n", p->pid, p->comm);
+
+ for(c = 0; c < p->fds_size; c++)
+ if(p->fds[c] > 0) {
+ file_descriptor_not_used(p->fds[c]);
+ p->fds[c] = 0;
+ }
+
+ pid_t r = p->pid;
+ p = p->next;
+ del_pid_entry(r);
+ }
+ else {
+ if(unlikely(p->keep)) p->keeploops++;
+ p->keep = 0;
+ p = p->next;
+ }
+ }
}
-long zero_all_targets(struct target *root) {
- struct target *w;
- long count = 0;
-
- for (w = root; w ; w = w->next) {
- count++;
-
- if(w->fds) free(w->fds);
- w->fds = NULL;
-
- w->minflt = 0;
- w->majflt = 0;
- w->utime = 0;
- w->stime = 0;
- w->cminflt = 0;
- w->cmajflt = 0;
- w->cutime = 0;
- w->cstime = 0;
- w->num_threads = 0;
- w->rss = 0;
- w->processes = 0;
-
- w->statm_size = 0;
- w->statm_resident = 0;
- w->statm_share = 0;
- w->statm_text = 0;
- w->statm_lib = 0;
- w->statm_data = 0;
- w->statm_dirty = 0;
-
- w->io_logical_bytes_read = 0;
- w->io_logical_bytes_written = 0;
- w->io_read_calls = 0;
- w->io_write_calls = 0;
- w->io_storage_bytes_read = 0;
- w->io_storage_bytes_written = 0;
- w->io_cancelled_write_bytes = 0;
- }
-
- return count;
+static void apply_apps_groups_targets_inheritance(void) {
+ struct pid_stat *p = NULL;
+
+ // children that do not have a target
+ // inherit their target from their parent
+ int found = 1, loops = 0;
+ while(found) {
+ if(unlikely(debug)) loops++;
+ found = 0;
+ for(p = root_of_pids; p ; p = p->next) {
+ // if this process does not have a target
+ // and it has a parent
+ // and its parent has a target
+ // then, set the parent's target to this process
+ if(unlikely(!p->target && p->parent && p->parent->target)) {
+ p->target = p->parent->target;
+ found++;
+
+ if(debug || (p->target && p->target->debug))
+ fprintf(stderr, "apps.plugin: \t\tTARGET INHERITANCE: %s is inherited by %d (%s) from its parent %d (%s).\n", p->target->name, p->pid, p->comm, p->parent->pid, p->parent->comm);
+ }
+ }
+ }
+
+ // find all the procs with 0 childs and merge them to their parents
+ // repeat, until nothing more can be done.
+ int sortlist = 1;
+ found = 1;
+ while(found) {
+ if(unlikely(debug)) loops++;
+ found = 0;
+
+ for(p = root_of_pids; p ; p = p->next) {
+ if(unlikely(!p->sortlist && !p->children_count))
+ p->sortlist = sortlist++;
+
+ // if this process does not have any children
+ // and is not already merged
+ // and has a parent
+ // and its parent has children
+ // and the target of this process and its parent is the same, or the parent does not have a target
+ // and its parent is not init
+ // then, mark them as merged.
+ if(unlikely(
+ !p->children_count
+ && !p->merged
+ && p->parent
+ && p->parent->children_count
+ && (p->target == p->parent->target || !p->parent->target)
+ && p->ppid != INIT_PID
+ )) {
+ p->parent->children_count--;
+ p->merged = 1;
+
+ // the parent inherits the child's target, if it does not have a target itself
+ if(unlikely(p->target && !p->parent->target)) {
+ p->parent->target = p->target;
+
+ if(debug || (p->target && p->target->debug))
+ fprintf(stderr, "apps.plugin: \t\tTARGET INHERITANCE: %s is inherited by %d (%s) from its child %d (%s).\n", p->target->name, p->parent->pid, p->parent->comm, p->pid, p->comm);
+ }
+
+ found++;
+ }
+ }
+
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: TARGET INHERITANCE: merged %d processes\n", found);
+ }
+
+ // init goes always to default target
+ if(all_pids[INIT_PID])
+ all_pids[INIT_PID]->target = apps_groups_default_target;
+
+ // give a default target on all top level processes
+ if(unlikely(debug)) loops++;
+ for(p = root_of_pids; p ; p = p->next) {
+ // if the process is not merged itself
+ // then is is a top level process
+ if(unlikely(!p->merged && !p->target))
+ p->target = apps_groups_default_target;
+
+ // make sure all processes have a sortlist
+ if(unlikely(!p->sortlist))
+ p->sortlist = sortlist++;
+ }
+
+ if(all_pids[1])
+ all_pids[1]->sortlist = sortlist++;
+
+ // give a target to all merged child processes
+ found = 1;
+ while(found) {
+ if(unlikely(debug)) loops++;
+ found = 0;
+ for(p = root_of_pids; p ; p = p->next) {
+ if(unlikely(!p->target && p->merged && p->parent && p->parent->target)) {
+ p->target = p->parent->target;
+ found++;
+
+ if(debug || (p->target && p->target->debug))
+ fprintf(stderr, "apps.plugin: \t\tTARGET INHERITANCE: %s is inherited by %d (%s) from its parent %d (%s) at phase 2.\n", p->target->name, p->pid, p->comm, p->parent->pid, p->parent->comm);
+ }
+ }
+ }
+
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: apply_apps_groups_targets_inheritance() made %d loops on the process tree\n", loops);
}
-void aggregate_pid_on_target(struct target *w, struct pid_stat *p, struct target *o) {
- if(unlikely(!w->fds)) {
- w->fds = calloc(sizeof(int), (size_t) all_files_size);
- if(unlikely(!w->fds))
- error("Cannot allocate memory for fds in %s", w->name);
- }
-
- if(likely(p->updated)) {
- w->cutime += p->cutime; // - p->fix_cutime;
- w->cstime += p->cstime; // - p->fix_cstime;
- w->cminflt += p->cminflt; // - p->fix_cminflt;
- w->cmajflt += p->cmajflt; // - p->fix_cmajflt;
-
- w->utime += p->utime; //+ (p->pid != 1)?(p->cutime - p->fix_cutime):0;
- w->stime += p->stime; //+ (p->pid != 1)?(p->cstime - p->fix_cstime):0;
- w->minflt += p->minflt; //+ (p->pid != 1)?(p->cminflt - p->fix_cminflt):0;
- w->majflt += p->majflt; //+ (p->pid != 1)?(p->cmajflt - p->fix_cmajflt):0;
-
- //if(p->num_threads < 0)
- // error("Negative threads number for pid '%s' (%d): %d", p->comm, p->pid, p->num_threads);
-
- //if(p->num_threads > 10000)
- // error("Excessive threads number for pid '%s' (%d): %d", p->comm, p->pid, p->num_threads);
-
- w->num_threads += p->num_threads;
- w->rss += p->rss;
-
- w->statm_size += p->statm_size;
- w->statm_resident += p->statm_resident;
- w->statm_share += p->statm_share;
- w->statm_text += p->statm_text;
- w->statm_lib += p->statm_lib;
- w->statm_data += p->statm_data;
- w->statm_dirty += p->statm_dirty;
-
- w->io_logical_bytes_read += p->io_logical_bytes_read;
- w->io_logical_bytes_written += p->io_logical_bytes_written;
- w->io_read_calls += p->io_read_calls;
- w->io_write_calls += p->io_write_calls;
- w->io_storage_bytes_read += p->io_storage_bytes_read;
- w->io_storage_bytes_written += p->io_storage_bytes_written;
- w->io_cancelled_write_bytes += p->io_cancelled_write_bytes;
-
- w->processes++;
-
- if(likely(w->fds)) {
- int c;
- for(c = 0; c < p->fds_size ;c++) {
- if(p->fds[c] == 0) continue;
-
- if(likely(p->fds[c] < all_files_size)) {
- if(w->fds) w->fds[p->fds[c]]++;
- }
- else
- error("Invalid fd number %d", p->fds[c]);
- }
- }
-
- if(unlikely(debug || w->debug))
- fprintf(stderr, "apps.plugin: \tAgregating %s pid %d on %s utime=%llu, stime=%llu, cutime=%llu, cstime=%llu, minflt=%llu, majflt=%llu, cminflt=%llu, cmajflt=%llu\n", p->comm, p->pid, w->name, p->utime, p->stime, p->cutime, p->cstime, p->minflt, p->majflt, p->cminflt, p->cmajflt);
-
-/* if(p->utime - p->old_utime > 100) fprintf(stderr, "BIG CHANGE: %d %s utime increased by %llu from %llu to %llu\n", p->pid, p->comm, p->utime - p->old_utime, p->old_utime, p->utime);
- if(p->cutime - p->old_cutime > 100) fprintf(stderr, "BIG CHANGE: %d %s cutime increased by %llu from %llu to %llu\n", p->pid, p->comm, p->cutime - p->old_cutime, p->old_cutime, p->cutime);
- if(p->stime - p->old_stime > 100) fprintf(stderr, "BIG CHANGE: %d %s stime increased by %llu from %llu to %llu\n", p->pid, p->comm, p->stime - p->old_stime, p->old_stime, p->stime);
- if(p->cstime - p->old_cstime > 100) fprintf(stderr, "BIG CHANGE: %d %s cstime increased by %llu from %llu to %llu\n", p->pid, p->comm, p->cstime - p->old_cstime, p->old_cstime, p->cstime);
- if(p->minflt - p->old_minflt > 5000) fprintf(stderr, "BIG CHANGE: %d %s minflt increased by %llu from %llu to %llu\n", p->pid, p->comm, p->minflt - p->old_minflt, p->old_minflt, p->minflt);
- if(p->majflt - p->old_majflt > 5000) fprintf(stderr, "BIG CHANGE: %d %s majflt increased by %llu from %llu to %llu\n", p->pid, p->comm, p->majflt - p->old_majflt, p->old_majflt, p->majflt);
- if(p->cminflt - p->old_cminflt > 15000) fprintf(stderr, "BIG CHANGE: %d %s cminflt increased by %llu from %llu to %llu\n", p->pid, p->comm, p->cminflt - p->old_cminflt, p->old_cminflt, p->cminflt);
- if(p->cmajflt - p->old_cmajflt > 15000) fprintf(stderr, "BIG CHANGE: %d %s cmajflt increased by %llu from %llu to %llu\n", p->pid, p->comm, p->cmajflt - p->old_cmajflt, p->old_cmajflt, p->cmajflt);
-*/
-#ifdef AGGREGATE_CHILDREN_TO_PARENTS
- p->old_utime = p->utime;
- p->old_cutime = p->cutime;
- p->old_stime = p->stime;
- p->old_cstime = p->cstime;
- p->old_minflt = p->minflt;
- p->old_majflt = p->majflt;
- p->old_cminflt = p->cminflt;
- p->old_cmajflt = p->cmajflt;
-#endif /* AGGREGATE_CHILDREN_TO_PARENTS */
-
- if(o) {
- // since the process switched target
- // for all incremental values
- // we have to subtract its OLD values from the new target
- // and add its OLD values to the old target
-
- // IMPORTANT
- // We add/subtract the last/OLD values we added to the target
-
- w->fix_cutime -= p->last_cutime;
- w->fix_cstime -= p->last_cstime;
- w->fix_cminflt -= p->last_cminflt;
- w->fix_cmajflt -= p->last_cmajflt;
-
- w->fix_utime -= p->last_utime;
- w->fix_stime -= p->last_stime;
- w->fix_minflt -= p->last_minflt;
- w->fix_majflt -= p->last_majflt;
-
-
- w->fix_io_logical_bytes_read -= p->last_io_logical_bytes_read;
- w->fix_io_logical_bytes_written -= p->last_io_logical_bytes_written;
- w->fix_io_read_calls -= p->last_io_read_calls;
- w->fix_io_write_calls -= p->last_io_write_calls;
- w->fix_io_storage_bytes_read -= p->last_io_storage_bytes_read;
- w->fix_io_storage_bytes_written -= p->last_io_storage_bytes_written;
- w->fix_io_cancelled_write_bytes -= p->last_io_cancelled_write_bytes;
-
- // ---
-
- o->fix_cutime += p->last_cutime;
- o->fix_cstime += p->last_cstime;
- o->fix_cminflt += p->last_cminflt;
- o->fix_cmajflt += p->last_cmajflt;
-
- o->fix_utime += p->last_utime;
- o->fix_stime += p->last_stime;
- o->fix_minflt += p->last_minflt;
- o->fix_majflt += p->last_majflt;
-
- o->fix_io_logical_bytes_read += p->last_io_logical_bytes_read;
- o->fix_io_logical_bytes_written += p->last_io_logical_bytes_written;
- o->fix_io_read_calls += p->last_io_read_calls;
- o->fix_io_write_calls += p->last_io_write_calls;
- o->fix_io_storage_bytes_read += p->last_io_storage_bytes_read;
- o->fix_io_storage_bytes_written += p->last_io_storage_bytes_written;
- o->fix_io_cancelled_write_bytes += p->last_io_cancelled_write_bytes;
- }
- }
- else {
- // if(o) fprintf(stderr, "apps.plugin: \t\tpid %d (%s) is not updated by OLD target %s (%s) is present.\n", p->pid, p->comm, o->id, o->name);
-
- // since the process has exited, the user
- // will see a drop in our charts, because the incremental
- // values of this process will not be there
-
- // add them to the fix_* values and they will be added to
- // the reported values, so that the report goes steady
- w->fix_minflt += p->minflt;
- w->fix_majflt += p->majflt;
- w->fix_utime += p->utime;
- w->fix_stime += p->stime;
- w->fix_cminflt += p->cminflt;
- w->fix_cmajflt += p->cmajflt;
- w->fix_cutime += p->cutime;
- w->fix_cstime += p->cstime;
-
- w->fix_io_logical_bytes_read += p->io_logical_bytes_read;
- w->fix_io_logical_bytes_written += p->io_logical_bytes_written;
- w->fix_io_read_calls += p->io_read_calls;
- w->fix_io_write_calls += p->io_write_calls;
- w->fix_io_storage_bytes_read += p->io_storage_bytes_read;
- w->fix_io_storage_bytes_written += p->io_storage_bytes_written;
- w->fix_io_cancelled_write_bytes += p->io_cancelled_write_bytes;
- }
-
+static size_t zero_all_targets(struct target *root) {
+ struct target *w;
+ size_t count = 0;
+
+ for (w = root; w ; w = w->next) {
+ count++;
+
+ w->minflt = 0;
+ w->majflt = 0;
+ w->utime = 0;
+ w->stime = 0;
+ w->gtime = 0;
+ w->cminflt = 0;
+ w->cmajflt = 0;
+ w->cutime = 0;
+ w->cstime = 0;
+ w->cgtime = 0;
+ w->num_threads = 0;
+ // w->rss = 0;
+ w->processes = 0;
+
+ w->statm_size = 0;
+ w->statm_resident = 0;
+ w->statm_share = 0;
+ // w->statm_text = 0;
+ // w->statm_lib = 0;
+ // w->statm_data = 0;
+ // w->statm_dirty = 0;
+
+ w->io_logical_bytes_read = 0;
+ w->io_logical_bytes_written = 0;
+ // w->io_read_calls = 0;
+ // w->io_write_calls = 0;
+ w->io_storage_bytes_read = 0;
+ w->io_storage_bytes_written = 0;
+ // w->io_cancelled_write_bytes = 0;
+
+ // zero file counters
+ if(w->target_fds) {
+ memset(w->target_fds, 0, sizeof(int) * w->target_fds_size);
+ w->openfiles = 0;
+ w->openpipes = 0;
+ w->opensockets = 0;
+ w->openinotifies = 0;
+ w->openeventfds = 0;
+ w->opentimerfds = 0;
+ w->opensignalfds = 0;
+ w->openeventpolls = 0;
+ w->openother = 0;
+ }
+ }
+
+ return count;
}
-void count_targets_fds(struct target *root) {
- int c;
- struct target *w;
-
- for (w = root; w ; w = w->next) {
- if(!w->fds) continue;
-
- w->openfiles = 0;
- w->openpipes = 0;
- w->opensockets = 0;
- w->openinotifies = 0;
- w->openeventfds = 0;
- w->opentimerfds = 0;
- w->opensignalfds = 0;
- w->openeventpolls = 0;
- w->openother = 0;
-
- for(c = 1; c < all_files_size ;c++) {
- if(w->fds[c] > 0)
- switch(all_files[c].type) {
- case FILETYPE_FILE:
- w->openfiles++;
- break;
-
- case FILETYPE_PIPE:
- w->openpipes++;
- break;
-
- case FILETYPE_SOCKET:
- w->opensockets++;
- break;
-
- case FILETYPE_INOTIFY:
- w->openinotifies++;
- break;
-
- case FILETYPE_EVENTFD:
- w->openeventfds++;
- break;
-
- case FILETYPE_TIMERFD:
- w->opentimerfds++;
- break;
-
- case FILETYPE_SIGNALFD:
- w->opensignalfds++;
- break;
-
- case FILETYPE_EVENTPOLL:
- w->openeventpolls++;
- break;
-
- default:
- w->openother++;
- }
- }
-
- free(w->fds);
- w->fds = NULL;
- }
-}
+static inline void reallocate_target_fds(struct target *w) {
+ if(unlikely(!w))
+ return;
-void calculate_netdata_statistics(void)
-{
- link_all_processes_to_their_parents();
- apply_apps_groups_targets_inheritance();
-
-#ifdef AGGREGATE_CHILDREN_TO_PARENTS
- aggregate_children_to_parents();
-#endif /* AGGREGATE_CHILDREN_TO_PARENTS */
-
- zero_all_targets(users_root_target);
- zero_all_targets(groups_root_target);
- apps_groups_targets = zero_all_targets(apps_groups_root_target);
-
-#ifdef AGGREGATE_CHILDREN_TO_PARENTS
- if(debug)
- debug_childrens_aggregations(0, 1);
-#endif /* AGGREGATE_CHILDREN_TO_PARENTS */
-
- // this has to be done, before the cleanup
- struct pid_stat *p = NULL;
- struct target *w = NULL, *o = NULL;
-
- // concentrate everything on the apps_groups_targets
- for(p = root_of_pids; p ; p = p->next) {
-
- // --------------------------------------------------------------------
- // apps_groups targets
- if(likely(p->target))
- aggregate_pid_on_target(p->target, p, NULL);
- else
- error("pid %d %s was left without a target!", p->pid, p->comm);
-
-
- // --------------------------------------------------------------------
- // user targets
- o = p->user_target;
- if(likely(p->user_target && p->user_target->uid == p->uid))
- w = p->user_target;
- else {
- if(unlikely(debug && p->user_target))
- fprintf(stderr, "apps.plugin: \t\tpid %d (%s) switched user from %d (%s) to %d.\n", p->pid, p->comm, p->user_target->uid, p->user_target->name, p->uid);
-
- w = p->user_target = get_users_target(p->uid);
- }
-
- if(likely(w))
- aggregate_pid_on_target(w, p, o);
- else
- error("pid %d %s was left without a user target!", p->pid, p->comm);
-
-
- // --------------------------------------------------------------------
- // group targets
- o = p->group_target;
- if(likely(p->group_target && p->group_target->gid == p->gid))
- w = p->group_target;
- else {
- if(unlikely(debug && p->group_target))
- fprintf(stderr, "apps.plugin: \t\tpid %d (%s) switched group from %d (%s) to %d.\n", p->pid, p->comm, p->group_target->gid, p->group_target->name, p->gid);
-
- w = p->group_target = get_groups_target(p->gid);
- }
-
- if(likely(w))
- aggregate_pid_on_target(w, p, o);
- else
- error("pid %d %s was left without a group target!", p->pid, p->comm);
-
- }
-
- count_targets_fds(apps_groups_root_target);
- count_targets_fds(users_root_target);
- count_targets_fds(groups_root_target);
-
- cleanup_non_existing_pids();
+ if(unlikely(!w->target_fds || w->target_fds_size < all_files_size)) {
+ w->target_fds = reallocz(w->target_fds, sizeof(int) * all_files_size);
+ memset(&w->target_fds[w->target_fds_size], 0, sizeof(int) * (all_files_size - w->target_fds_size));
+ w->target_fds_size = all_files_size;
+ }
}
-// ----------------------------------------------------------------------------
-// update chart dimensions
-
-unsigned long long send_resource_usage_to_netdata() {
- static struct timeval last = { 0, 0 };
- static struct rusage me_last;
-
- struct timeval now;
- struct rusage me;
-
- unsigned long long usec;
- unsigned long long cpuuser;
- unsigned long long cpusyst;
-
- if(!last.tv_sec) {
- gettimeofday(&last, NULL);
- getrusage(RUSAGE_SELF, &me_last);
-
- // the first time, give a zero to allow
- // netdata calibrate to the current time
- // usec = update_every * 1000000ULL;
- usec = 0ULL;
- cpuuser = 0;
- cpusyst = 0;
- }
- else {
- gettimeofday(&now, NULL);
- getrusage(RUSAGE_SELF, &me);
-
- usec = usecdiff(&now, &last);
- cpuuser = me.ru_utime.tv_sec * 1000000ULL + me.ru_utime.tv_usec;
- cpusyst = me.ru_stime.tv_sec * 1000000ULL + me.ru_stime.tv_usec;
-
- bcopy(&now, &last, sizeof(struct timeval));
- bcopy(&me, &me_last, sizeof(struct rusage));
- }
-
- fprintf(stdout, "BEGIN netdata.apps_cpu %llu\n", usec);
- fprintf(stdout, "SET user = %llu\n", cpuuser);
- fprintf(stdout, "SET system = %llu\n", cpusyst);
- fprintf(stdout, "END\n");
-
- fprintf(stdout, "BEGIN netdata.apps_files %llu\n", usec);
- fprintf(stdout, "SET files = %llu\n", file_counter);
- fprintf(stdout, "SET pids = %ld\n", all_pids_count);
- fprintf(stdout, "SET fds = %d\n", all_files_len);
- fprintf(stdout, "SET targets = %ld\n", apps_groups_targets);
- fprintf(stdout, "END\n");
-
- return usec;
+static inline void aggregate_fd_on_target(int fd, struct target *w) {
+ if(unlikely(!w))
+ return;
+
+ if(unlikely(w->target_fds[fd])) {
+ // it is already aggregated
+ // just increase its usage counter
+ w->target_fds[fd]++;
+ return;
+ }
+
+ // increase its usage counter
+ // so that we will not add it again
+ w->target_fds[fd]++;
+
+ switch(all_files[fd].type) {
+ case FILETYPE_FILE:
+ w->openfiles++;
+ break;
+
+ case FILETYPE_PIPE:
+ w->openpipes++;
+ break;
+
+ case FILETYPE_SOCKET:
+ w->opensockets++;
+ break;
+
+ case FILETYPE_INOTIFY:
+ w->openinotifies++;
+ break;
+
+ case FILETYPE_EVENTFD:
+ w->openeventfds++;
+ break;
+
+ case FILETYPE_TIMERFD:
+ w->opentimerfds++;
+ break;
+
+ case FILETYPE_SIGNALFD:
+ w->opensignalfds++;
+ break;
+
+ case FILETYPE_EVENTPOLL:
+ w->openeventpolls++;
+ break;
+
+ case FILETYPE_OTHER:
+ w->openother++;
+ break;
+ }
}
-void send_collected_data_to_netdata(struct target *root, const char *type, unsigned long long usec)
-{
- struct target *w;
-
- fprintf(stdout, "BEGIN %s.cpu %llu\n", type, usec);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
-
- fprintf(stdout, "SET %s = %llu\n", w->name, w->utime + w->stime + w->fix_utime + w->fix_stime);
- }
- fprintf(stdout, "END\n");
+static inline void aggregate_pid_fds_on_targets(struct pid_stat *p) {
- fprintf(stdout, "BEGIN %s.cpu_user %llu\n", type, usec);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
+ if(unlikely(!p->updated)) {
+ // the process is not running
+ return;
+ }
- fprintf(stdout, "SET %s = %llu\n", w->name, w->utime + w->fix_utime);
- }
- fprintf(stdout, "END\n");
+ struct target *w = p->target, *u = p->user_target, *g = p->group_target;
- fprintf(stdout, "BEGIN %s.cpu_system %llu\n", type, usec);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
+ reallocate_target_fds(w);
+ reallocate_target_fds(u);
+ reallocate_target_fds(g);
- fprintf(stdout, "SET %s = %llu\n", w->name, w->stime + w->fix_stime);
- }
- fprintf(stdout, "END\n");
+ int c, size = p->fds_size, *fds = p->fds;
+ for(c = 0; c < size ;c++) {
+ int fd = fds[c];
- fprintf(stdout, "BEGIN %s.threads %llu\n", type, usec);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
+ if(likely(fd <= 0 || fd >= all_files_size))
+ continue;
- fprintf(stdout, "SET %s = %llu\n", w->name, w->num_threads);
- }
- fprintf(stdout, "END\n");
-
- fprintf(stdout, "BEGIN %s.processes %llu\n", type, usec);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
+ aggregate_fd_on_target(fd, w);
+ aggregate_fd_on_target(fd, u);
+ aggregate_fd_on_target(fd, g);
+ }
+}
- fprintf(stdout, "SET %s = %lu\n", w->name, w->processes);
- }
- fprintf(stdout, "END\n");
+static inline void aggregate_pid_on_target(struct target *w, struct pid_stat *p, struct target *o) {
+ (void)o;
+
+ if(unlikely(!p->updated)) {
+ // the process is not running
+ return;
+ }
+
+ if(unlikely(!w)) {
+ error("pid %d %s was left without a target!", p->pid, p->comm);
+ return;
+ }
+
+ w->cutime += p->cutime;
+ w->cstime += p->cstime;
+ w->cgtime += p->cgtime;
+ w->cminflt += p->cminflt;
+ w->cmajflt += p->cmajflt;
+
+ w->utime += p->utime;
+ w->stime += p->stime;
+ w->gtime += p->gtime;
+ w->minflt += p->minflt;
+ w->majflt += p->majflt;
+
+ // w->rss += p->rss;
+
+ w->statm_size += p->statm_size;
+ w->statm_resident += p->statm_resident;
+ w->statm_share += p->statm_share;
+ // w->statm_text += p->statm_text;
+ // w->statm_lib += p->statm_lib;
+ // w->statm_data += p->statm_data;
+ // w->statm_dirty += p->statm_dirty;
+
+ w->io_logical_bytes_read += p->io_logical_bytes_read;
+ w->io_logical_bytes_written += p->io_logical_bytes_written;
+ // w->io_read_calls += p->io_read_calls;
+ // w->io_write_calls += p->io_write_calls;
+ w->io_storage_bytes_read += p->io_storage_bytes_read;
+ w->io_storage_bytes_written += p->io_storage_bytes_written;
+ // w->io_cancelled_write_bytes += p->io_cancelled_write_bytes;
+
+ w->processes++;
+ w->num_threads += p->num_threads;
+
+ if(unlikely(debug || w->debug))
+ fprintf(stderr, "apps.plugin: \taggregating '%s' pid %d on target '%s' utime=" KERNEL_UINT_FORMAT ", stime=" KERNEL_UINT_FORMAT ", gtime=" KERNEL_UINT_FORMAT ", cutime=" KERNEL_UINT_FORMAT ", cstime=" KERNEL_UINT_FORMAT ", cgtime=" KERNEL_UINT_FORMAT ", minflt=" KERNEL_UINT_FORMAT ", majflt=" KERNEL_UINT_FORMAT ", cminflt=" KERNEL_UINT_FORMAT ", cmajflt=" KERNEL_UINT_FORMAT "\n", p->comm, p->pid, w->name, p->utime, p->stime, p->gtime, p->cutime, p->cstime, p->cgtime, p->minflt, p->majflt, p->cminflt, p->cmajflt);
+}
- fprintf(stdout, "BEGIN %s.mem %llu\n", type, usec);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
+static void calculate_netdata_statistics(void) {
- fprintf(stdout, "SET %s = %lld\n", w->name, (long long)w->statm_resident - (long long)w->statm_share);
- }
- fprintf(stdout, "END\n");
+ apply_apps_groups_targets_inheritance();
- fprintf(stdout, "BEGIN %s.minor_faults %llu\n", type, usec);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
+ zero_all_targets(users_root_target);
+ zero_all_targets(groups_root_target);
+ apps_groups_targets_count = zero_all_targets(apps_groups_root_target);
- fprintf(stdout, "SET %s = %llu\n", w->name, w->minflt + w->fix_minflt);
- }
- fprintf(stdout, "END\n");
+ // this has to be done, before the cleanup
+ struct pid_stat *p = NULL;
+ struct target *w = NULL, *o = NULL;
- fprintf(stdout, "BEGIN %s.major_faults %llu\n", type, usec);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
+ // concentrate everything on the targets
+ for(p = root_of_pids; p ; p = p->next) {
- fprintf(stdout, "SET %s = %llu\n", w->name, w->majflt + w->fix_majflt);
- }
- fprintf(stdout, "END\n");
+ // --------------------------------------------------------------------
+ // apps_groups target
- fprintf(stdout, "BEGIN %s.lreads %llu\n", type, usec);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
+ aggregate_pid_on_target(p->target, p, NULL);
- fprintf(stdout, "SET %s = %llu\n", w->name, w->io_logical_bytes_read + w->fix_io_logical_bytes_read);
- }
- fprintf(stdout, "END\n");
- fprintf(stdout, "BEGIN %s.lwrites %llu\n", type, usec);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
+ // --------------------------------------------------------------------
+ // user target
- fprintf(stdout, "SET %s = %llu\n", w->name, w->io_logical_bytes_written + w->fix_io_logical_bytes_written);
- }
- fprintf(stdout, "END\n");
+ o = p->user_target;
+ if(likely(p->user_target && p->user_target->uid == p->uid))
+ w = p->user_target;
+ else {
+ if(unlikely(debug && p->user_target))
+ fprintf(stderr, "apps.plugin: \t\tpid %d (%s) switched user from %u (%s) to %u.\n", p->pid, p->comm, p->user_target->uid, p->user_target->name, p->uid);
- fprintf(stdout, "BEGIN %s.preads %llu\n", type, usec);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
+ w = p->user_target = get_users_target(p->uid);
+ }
- fprintf(stdout, "SET %s = %llu\n", w->name, w->io_storage_bytes_read + w->fix_io_storage_bytes_read);
- }
- fprintf(stdout, "END\n");
+ aggregate_pid_on_target(w, p, o);
- fprintf(stdout, "BEGIN %s.pwrites %llu\n", type, usec);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
- fprintf(stdout, "SET %s = %llu\n", w->name, w->io_storage_bytes_written + w->fix_io_storage_bytes_written);
- }
- fprintf(stdout, "END\n");
+ // --------------------------------------------------------------------
+ // user group target
- fprintf(stdout, "BEGIN %s.files %llu\n", type, usec);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
+ o = p->group_target;
+ if(likely(p->group_target && p->group_target->gid == p->gid))
+ w = p->group_target;
+ else {
+ if(unlikely(debug && p->group_target))
+ fprintf(stderr, "apps.plugin: \t\tpid %d (%s) switched group from %u (%s) to %u.\n", p->pid, p->comm, p->group_target->gid, p->group_target->name, p->gid);
- fprintf(stdout, "SET %s = %llu\n", w->name, w->openfiles);
- }
- fprintf(stdout, "END\n");
+ w = p->group_target = get_groups_target(p->gid);
+ }
- fprintf(stdout, "BEGIN %s.sockets %llu\n", type, usec);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
+ aggregate_pid_on_target(w, p, o);
- fprintf(stdout, "SET %s = %llu\n", w->name, w->opensockets);
- }
- fprintf(stdout, "END\n");
- fprintf(stdout, "BEGIN %s.pipes %llu\n", type, usec);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
+ // --------------------------------------------------------------------
+ // aggregate all file descriptors
- fprintf(stdout, "SET %s = %llu\n", w->name, w->openpipes);
- }
- fprintf(stdout, "END\n");
+ if(enable_file_charts)
+ aggregate_pid_fds_on_targets(p);
+ }
- fflush(stdout);
+ cleanup_exited_pids();
}
-
// ----------------------------------------------------------------------------
-// generate the charts
-
-void send_charts_updates_to_netdata(struct target *root, const char *type, const char *title)
-{
- struct target *w;
- int newly_added = 0;
-
- for(w = root ; w ; w = w->next)
- if(!w->exposed && w->processes) {
- newly_added++;
- w->exposed = 1;
- if(debug || w->debug) fprintf(stderr, "apps.plugin: %s just added - regenerating charts.\n", w->name);
- }
-
- // nothing more to show
- if(!newly_added) return;
-
- // we have something new to show
- // update the charts
- fprintf(stdout, "CHART %s.cpu '' '%s CPU Time (%ld%% = %ld core%s)' 'cpu time %%' cpu %s.cpu stacked 20001 %d\n", type, title, (processors * 100), processors, (processors>1)?"s":"", type, update_every);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
-
- fprintf(stdout, "DIMENSION %s '' incremental 100 %u %s\n", w->name, hz, w->hidden ? "hidden,noreset" : "noreset");
- }
-
- fprintf(stdout, "CHART %s.mem '' '%s Dedicated Memory (w/o shared)' 'MB' mem %s.mem stacked 20003 %d\n", type, title, type, update_every);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
-
- fprintf(stdout, "DIMENSION %s '' absolute %ld %ld noreset\n", w->name, sysconf(_SC_PAGESIZE), 1024L*1024L);
- }
-
- fprintf(stdout, "CHART %s.threads '' '%s Threads' 'threads' processes %s.threads stacked 20005 %d\n", type, title, type, update_every);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
-
- fprintf(stdout, "DIMENSION %s '' absolute 1 1 noreset\n", w->name);
- }
+// update chart dimensions
- fprintf(stdout, "CHART %s.processes '' '%s Processes' 'processes' processes %s.processes stacked 20004 %d\n", type, title, type, update_every);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
+int print_calculated_number(char *str, calculated_number value) { (void)str; (void)value; return 0; }
- fprintf(stdout, "DIMENSION %s '' absolute 1 1 noreset\n", w->name);
- }
+static inline void send_BEGIN(const char *type, const char *id, usec_t usec) {
+ fprintf(stdout, "BEGIN %s.%s %llu\n", type, id, usec);
+}
- fprintf(stdout, "CHART %s.cpu_user '' '%s CPU User Time (%ld%% = %ld core%s)' 'cpu time %%' cpu %s.cpu_user stacked 20020 %d\n", type, title, (processors * 100), processors, (processors>1)?"s":"", type, update_every);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
+static inline void send_SET(const char *name, kernel_uint_t value) {
+ fprintf(stdout, "SET %s = " KERNEL_UINT_FORMAT "\n", name, value);
+}
- fprintf(stdout, "DIMENSION %s '' incremental 100 %ld noreset\n", w->name, hz * processors);
- }
+static inline void send_END(void) {
+ fprintf(stdout, "END\n");
+}
- fprintf(stdout, "CHART %s.cpu_system '' '%s CPU System Time (%ld%% = %ld core%s)' 'cpu time %%' cpu %s.cpu_system stacked 20021 %d\n", type, title, (processors * 100), processors, (processors>1)?"s":"", type, update_every);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
+static usec_t send_resource_usage_to_netdata() {
+ static struct timeval last = { 0, 0 };
+ static struct rusage me_last;
+
+ struct timeval now;
+ struct rusage me;
+
+ usec_t usec;
+ usec_t cpuuser;
+ usec_t cpusyst;
+
+ if(!last.tv_sec) {
+ now_monotonic_timeval(&last);
+ getrusage(RUSAGE_SELF, &me_last);
+
+ // the first time, give a zero to allow
+ // netdata calibrate to the current time
+ // usec = update_every * USEC_PER_SEC;
+ usec = 0ULL;
+ cpuuser = 0;
+ cpusyst = 0;
+ }
+ else {
+ now_monotonic_timeval(&now);
+ getrusage(RUSAGE_SELF, &me);
+
+ usec = dt_usec(&now, &last);
+ cpuuser = me.ru_utime.tv_sec * USEC_PER_SEC + me.ru_utime.tv_usec;
+ cpusyst = me.ru_stime.tv_sec * USEC_PER_SEC + me.ru_stime.tv_usec;
+
+ memmove(&last, &now, sizeof(struct timeval));
+ memmove(&me_last, &me, sizeof(struct rusage));
+ }
+
+ static char created_charts = 0;
+ if(unlikely(!created_charts)) {
+ created_charts = 1;
+
+ fprintf(stdout
+ , "CHART netdata.apps_cpu '' 'Apps Plugin CPU' 'milliseconds/s' apps.plugin netdata.apps_cpu stacked 140000 %1$d\n"
+ "DIMENSION user '' incremental 1 1000\n"
+ "DIMENSION system '' incremental 1 1000\n"
+ "CHART netdata.apps_files '' 'Apps Plugin Files' 'files/s' apps.plugin netdata.apps_files line 140001 %1$d\n"
+ "DIMENSION calls '' incremental 1 1\n"
+ "DIMENSION files '' incremental 1 1\n"
+ "DIMENSION pids '' absolute 1 1\n"
+ "DIMENSION fds '' absolute 1 1\n"
+ "DIMENSION targets '' absolute 1 1\n"
+ "CHART netdata.apps_fix '' 'Apps Plugin Normalization Ratios' 'percentage' apps.plugin netdata.apps_fix line 140002 %1$d\n"
+ "DIMENSION utime '' absolute 1 %2$llu\n"
+ "DIMENSION stime '' absolute 1 %2$llu\n"
+ "DIMENSION gtime '' absolute 1 %2$llu\n"
+ "DIMENSION minflt '' absolute 1 %2$llu\n"
+ "DIMENSION majflt '' absolute 1 %2$llu\n"
+ , update_every
+ , RATES_DETAIL
+ );
+
+ if(include_exited_childs)
+ fprintf(stdout
+ , "CHART netdata.apps_children_fix '' 'Apps Plugin Exited Children Normalization Ratios' 'percentage' apps.plugin netdata.apps_children_fix line 140003 %1$d\n"
+ "DIMENSION cutime '' absolute 1 %2$llu\n"
+ "DIMENSION cstime '' absolute 1 %2$llu\n"
+ "DIMENSION cgtime '' absolute 1 %2$llu\n"
+ "DIMENSION cminflt '' absolute 1 %2$llu\n"
+ "DIMENSION cmajflt '' absolute 1 %2$llu\n"
+ , update_every
+ , RATES_DETAIL
+ );
+ }
+
+ fprintf(stdout,
+ "BEGIN netdata.apps_cpu %llu\n"
+ "SET user = %llu\n"
+ "SET system = %llu\n"
+ "END\n"
+ "BEGIN netdata.apps_files %llu\n"
+ "SET calls = %zu\n"
+ "SET files = %zu\n"
+ "SET pids = %zu\n"
+ "SET fds = %d\n"
+ "SET targets = %zu\n"
+ "END\n"
+ "BEGIN netdata.apps_fix %llu\n"
+ "SET utime = %u\n"
+ "SET stime = %u\n"
+ "SET gtime = %u\n"
+ "SET minflt = %u\n"
+ "SET majflt = %u\n"
+ "END\n"
+ , usec
+ , cpuuser
+ , cpusyst
+ , usec
+ , calls_counter
+ , file_counter
+ , all_pids_count
+ , all_files_len
+ , apps_groups_targets_count
+ , usec
+ , (unsigned int)(utime_fix_ratio * 100 * RATES_DETAIL)
+ , (unsigned int)(stime_fix_ratio * 100 * RATES_DETAIL)
+ , (unsigned int)(gtime_fix_ratio * 100 * RATES_DETAIL)
+ , (unsigned int)(minflt_fix_ratio * 100 * RATES_DETAIL)
+ , (unsigned int)(majflt_fix_ratio * 100 * RATES_DETAIL)
+ );
+
+ if(include_exited_childs)
+ fprintf(stdout,
+ "BEGIN netdata.apps_children_fix %llu\n"
+ "SET cutime = %u\n"
+ "SET cstime = %u\n"
+ "SET cgtime = %u\n"
+ "SET cminflt = %u\n"
+ "SET cmajflt = %u\n"
+ "END\n"
+ , usec
+ , (unsigned int)(cutime_fix_ratio * 100 * RATES_DETAIL)
+ , (unsigned int)(cstime_fix_ratio * 100 * RATES_DETAIL)
+ , (unsigned int)(cgtime_fix_ratio * 100 * RATES_DETAIL)
+ , (unsigned int)(cminflt_fix_ratio * 100 * RATES_DETAIL)
+ , (unsigned int)(cmajflt_fix_ratio * 100 * RATES_DETAIL)
+ );
+
+ return usec;
+}
- fprintf(stdout, "DIMENSION %s '' incremental 100 %ld noreset\n", w->name, hz * processors);
- }
+static void normalize_utilization(struct target *root) {
+ struct target *w;
+
+ // childs processing introduces spikes
+ // here we try to eliminate them by disabling childs processing either for specific dimensions
+ // or entirely. Of course, either way, we disable it just a single iteration.
+
+ kernel_uint_t max_time = processors * hz * RATES_DETAIL;
+ kernel_uint_t utime = 0, cutime = 0, stime = 0, cstime = 0, gtime = 0, cgtime = 0, minflt = 0, cminflt = 0, majflt = 0, cmajflt = 0;
+
+ if(global_utime > max_time) global_utime = max_time;
+ if(global_stime > max_time) global_stime = max_time;
+ if(global_gtime > max_time) global_gtime = max_time;
+
+ for(w = root; w ; w = w->next) {
+ if(w->target || (!w->processes && !w->exposed)) continue;
+
+ utime += w->utime;
+ stime += w->stime;
+ gtime += w->gtime;
+ cutime += w->cutime;
+ cstime += w->cstime;
+ cgtime += w->cgtime;
+
+ minflt += w->minflt;
+ majflt += w->majflt;
+ cminflt += w->cminflt;
+ cmajflt += w->cmajflt;
+ }
+
+ if((global_utime || global_stime || global_gtime) && (utime || stime || gtime)) {
+ if(global_utime + global_stime + global_gtime > utime + cutime + stime + cstime + gtime + cgtime) {
+ // everything we collected fits
+ utime_fix_ratio =
+ stime_fix_ratio =
+ gtime_fix_ratio =
+ cutime_fix_ratio =
+ cstime_fix_ratio =
+ cgtime_fix_ratio = 1.0; //(double)(global_utime + global_stime) / (double)(utime + cutime + stime + cstime);
+ }
+ else if(global_utime + global_stime > utime + stime) {
+ // childrens resources are too high
+ // lower only the children resources
+ utime_fix_ratio =
+ stime_fix_ratio =
+ gtime_fix_ratio = 1.0;
+ cutime_fix_ratio =
+ cstime_fix_ratio =
+ cgtime_fix_ratio = (double)((global_utime + global_stime) - (utime + stime)) / (double)(cutime + cstime);
+ }
+ else {
+ // even running processes are unrealistic
+ // zero the children resources
+ // lower the running processes resources
+ utime_fix_ratio =
+ stime_fix_ratio =
+ gtime_fix_ratio = (double)(global_utime + global_stime) / (double)(utime + stime);
+ cutime_fix_ratio =
+ cstime_fix_ratio =
+ cgtime_fix_ratio = 0.0;
+ }
+ }
+ else {
+ utime_fix_ratio =
+ stime_fix_ratio =
+ gtime_fix_ratio =
+ cutime_fix_ratio =
+ cstime_fix_ratio =
+ cgtime_fix_ratio = 0.0;
+ }
+
+ if(utime_fix_ratio > 1.0) utime_fix_ratio = 1.0;
+ if(cutime_fix_ratio > 1.0) cutime_fix_ratio = 1.0;
+ if(stime_fix_ratio > 1.0) stime_fix_ratio = 1.0;
+ if(cstime_fix_ratio > 1.0) cstime_fix_ratio = 1.0;
+ if(gtime_fix_ratio > 1.0) gtime_fix_ratio = 1.0;
+ if(cgtime_fix_ratio > 1.0) cgtime_fix_ratio = 1.0;
+
+ // if(utime_fix_ratio < 0.0) utime_fix_ratio = 0.0;
+ // if(cutime_fix_ratio < 0.0) cutime_fix_ratio = 0.0;
+ // if(stime_fix_ratio < 0.0) stime_fix_ratio = 0.0;
+ // if(cstime_fix_ratio < 0.0) cstime_fix_ratio = 0.0;
+ // if(gtime_fix_ratio < 0.0) gtime_fix_ratio = 0.0;
+ // if(cgtime_fix_ratio < 0.0) cgtime_fix_ratio = 0.0;
+
+ // FIXME
+ // we use cpu time to normalize page faults
+ // the problem is that to find the proper max values
+ // for page faults we have to parse /proc/vmstat
+ // which is quite big to do it again (netdata does it already)
+ //
+ // a better solution could be to somehow have netdata
+ // do this normalization for us
+
+ if(utime || stime || gtime)
+ majflt_fix_ratio =
+ minflt_fix_ratio = (double)(utime * utime_fix_ratio + stime * stime_fix_ratio + gtime * gtime_fix_ratio) / (double)(utime + stime + gtime);
+ else
+ minflt_fix_ratio =
+ majflt_fix_ratio = 1.0;
+
+ if(cutime || cstime || cgtime)
+ cmajflt_fix_ratio =
+ cminflt_fix_ratio = (double)(cutime * cutime_fix_ratio + cstime * cstime_fix_ratio + cgtime * cgtime_fix_ratio) / (double)(cutime + cstime + cgtime);
+ else
+ cminflt_fix_ratio =
+ cmajflt_fix_ratio = 1.0;
+
+ // the report
+
+ if(unlikely(debug)) {
+ fprintf(stderr,
+ "SYSTEM: u=" KERNEL_UINT_FORMAT " s=" KERNEL_UINT_FORMAT " g=" KERNEL_UINT_FORMAT " "
+ "COLLECTED: u=" KERNEL_UINT_FORMAT " s=" KERNEL_UINT_FORMAT " g=" KERNEL_UINT_FORMAT " cu=" KERNEL_UINT_FORMAT " cs=" KERNEL_UINT_FORMAT " cg=" KERNEL_UINT_FORMAT " "
+ "DELTA: u=" KERNEL_UINT_FORMAT " s=" KERNEL_UINT_FORMAT " g=" KERNEL_UINT_FORMAT " "
+ "FIX: u=%0.2f s=%0.2f g=%0.2f cu=%0.2f cs=%0.2f cg=%0.2f "
+ "FINALLY: u=" KERNEL_UINT_FORMAT " s=" KERNEL_UINT_FORMAT " g=" KERNEL_UINT_FORMAT " cu=" KERNEL_UINT_FORMAT " cs=" KERNEL_UINT_FORMAT " cg=" KERNEL_UINT_FORMAT " "
+ "\n"
+ , global_utime
+ , global_stime
+ , global_gtime
+ , utime
+ , stime
+ , gtime
+ , cutime
+ , cstime
+ , cgtime
+ , utime + cutime - global_utime
+ , stime + cstime - global_stime
+ , gtime + cgtime - global_gtime
+ , utime_fix_ratio
+ , stime_fix_ratio
+ , gtime_fix_ratio
+ , cutime_fix_ratio
+ , cstime_fix_ratio
+ , cgtime_fix_ratio
+ , (kernel_uint_t)(utime * utime_fix_ratio)
+ , (kernel_uint_t)(stime * stime_fix_ratio)
+ , (kernel_uint_t)(gtime * gtime_fix_ratio)
+ , (kernel_uint_t)(cutime * cutime_fix_ratio)
+ , (kernel_uint_t)(cstime * cstime_fix_ratio)
+ , (kernel_uint_t)(cgtime * cgtime_fix_ratio)
+ );
+ }
+}
- fprintf(stdout, "CHART %s.major_faults '' '%s Major Page Faults (swap read)' 'page faults/s' swap %s.major_faults stacked 20010 %d\n", type, title, type, update_every);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
+static void send_collected_data_to_netdata(struct target *root, const char *type, usec_t usec) {
+ struct target *w;
+
+ send_BEGIN(type, "cpu", usec);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ send_SET(w->name, (kernel_uint_t)(w->utime * utime_fix_ratio) + (kernel_uint_t)(w->stime * stime_fix_ratio) + (kernel_uint_t)(w->gtime * gtime_fix_ratio) + (include_exited_childs?((kernel_uint_t)(w->cutime * cutime_fix_ratio) + (kernel_uint_t)(w->cstime * cstime_fix_ratio) + (kernel_uint_t)(w->cgtime * cgtime_fix_ratio)):0ULL));
+ }
+ send_END();
+
+ send_BEGIN(type, "cpu_user", usec);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ send_SET(w->name, (kernel_uint_t)(w->utime * utime_fix_ratio) + (include_exited_childs?((kernel_uint_t)(w->cutime * cutime_fix_ratio)):0ULL));
+ }
+ send_END();
+
+ send_BEGIN(type, "cpu_system", usec);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ send_SET(w->name, (kernel_uint_t)(w->stime * stime_fix_ratio) + (include_exited_childs?((kernel_uint_t)(w->cstime * cstime_fix_ratio)):0ULL));
+ }
+ send_END();
+
+ if(show_guest_time) {
+ send_BEGIN(type, "cpu_guest", usec);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ send_SET(w->name, (kernel_uint_t)(w->gtime * gtime_fix_ratio) + (include_exited_childs?((kernel_uint_t)(w->cgtime * cgtime_fix_ratio)):0ULL));
+ }
+ send_END();
+ }
+
+ send_BEGIN(type, "threads", usec);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ send_SET(w->name, w->num_threads);
+ }
+ send_END();
+
+ send_BEGIN(type, "processes", usec);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ send_SET(w->name, w->processes);
+ }
+ send_END();
+
+ send_BEGIN(type, "mem", usec);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ send_SET(w->name, (w->statm_resident > w->statm_share)?(w->statm_resident - w->statm_share):0ULL);
+ }
+ send_END();
+
+ send_BEGIN(type, "vmem", usec);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ send_SET(w->name, w->statm_size);
+ }
+ send_END();
+
+ send_BEGIN(type, "minor_faults", usec);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ send_SET(w->name, (kernel_uint_t)(w->minflt * minflt_fix_ratio) + (include_exited_childs?((kernel_uint_t)(w->cminflt * cminflt_fix_ratio)):0ULL));
+ }
+ send_END();
+
+ send_BEGIN(type, "major_faults", usec);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ send_SET(w->name, (kernel_uint_t)(w->majflt * majflt_fix_ratio) + (include_exited_childs?((kernel_uint_t)(w->cmajflt * cmajflt_fix_ratio)):0ULL));
+ }
+ send_END();
+
+#ifndef __FreeBSD__
+ send_BEGIN(type, "lreads", usec);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ send_SET(w->name, w->io_logical_bytes_read);
+ }
+ send_END();
+
+ send_BEGIN(type, "lwrites", usec);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ send_SET(w->name, w->io_logical_bytes_written);
+ }
+ send_END();
+#endif
- fprintf(stdout, "DIMENSION %s '' incremental 1 1 noreset\n", w->name);
- }
+ send_BEGIN(type, "preads", usec);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ send_SET(w->name, w->io_storage_bytes_read);
+ }
+ send_END();
+
+ send_BEGIN(type, "pwrites", usec);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ send_SET(w->name, w->io_storage_bytes_written);
+ }
+ send_END();
+
+ if(enable_file_charts) {
+ send_BEGIN(type, "files", usec);
+ for (w = root; w; w = w->next) {
+ if (unlikely(w->exposed))
+ send_SET(w->name, w->openfiles);
+ }
+ send_END();
+
+ send_BEGIN(type, "sockets", usec);
+ for (w = root; w; w = w->next) {
+ if (unlikely(w->exposed))
+ send_SET(w->name, w->opensockets);
+ }
+ send_END();
+
+ send_BEGIN(type, "pipes", usec);
+ for (w = root; w; w = w->next) {
+ if (unlikely(w->exposed))
+ send_SET(w->name, w->openpipes);
+ }
+ send_END();
+ }
+}
- fprintf(stdout, "CHART %s.minor_faults '' '%s Minor Page Faults' 'page faults/s' mem %s.minor_faults stacked 20011 %d\n", type, title, type, update_every);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
- fprintf(stdout, "DIMENSION %s '' incremental 1 1 noreset\n", w->name);
- }
+// ----------------------------------------------------------------------------
+// generate the charts
- fprintf(stdout, "CHART %s.lreads '' '%s Disk Logical Reads' 'kilobytes/s' disk %s.lreads stacked 20042 %d\n", type, title, type, update_every);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
+static void send_charts_updates_to_netdata(struct target *root, const char *type, const char *title)
+{
+ struct target *w;
+ int newly_added = 0;
+
+ for(w = root ; w ; w = w->next) {
+ if (w->target) continue;
+
+ if (!w->exposed && w->processes) {
+ newly_added++;
+ w->exposed = 1;
+ if (debug || w->debug) fprintf(stderr, "apps.plugin: %s just added - regenerating charts.\n", w->name);
+ }
+ }
+
+ // nothing more to show
+ if(!newly_added && show_guest_time == show_guest_time_old) return;
+
+ // we have something new to show
+ // update the charts
+ fprintf(stdout, "CHART %s.cpu '' '%s CPU Time (%d%% = %d core%s)' 'cpu time %%' cpu %s.cpu stacked 20001 %d\n", type, title, (processors * 100), processors, (processors>1)?"s":"", type, update_every);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ fprintf(stdout, "DIMENSION %s '' absolute 1 %llu %s\n", w->name, hz * RATES_DETAIL / 100, w->hidden ? "hidden" : "");
+ }
+
+ fprintf(stdout, "CHART %s.mem '' '%s Real Memory (w/o shared)' 'MB' mem %s.mem stacked 20003 %d\n", type, title, type, update_every);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ fprintf(stdout, "DIMENSION %s '' absolute %ld %ld\n", w->name, sysconf(_SC_PAGESIZE), 1024L*1024L);
+ }
+
+ fprintf(stdout, "CHART %s.vmem '' '%s Virtual Memory Size' 'MB' mem %s.vmem stacked 20004 %d\n", type, title, type, update_every);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ fprintf(stdout, "DIMENSION %s '' absolute %ld %ld\n", w->name, sysconf(_SC_PAGESIZE), 1024L*1024L);
+ }
+
+ fprintf(stdout, "CHART %s.threads '' '%s Threads' 'threads' processes %s.threads stacked 20005 %d\n", type, title, type, update_every);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name);
+ }
+
+ fprintf(stdout, "CHART %s.processes '' '%s Processes' 'processes' processes %s.processes stacked 20004 %d\n", type, title, type, update_every);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name);
+ }
+
+ fprintf(stdout, "CHART %s.cpu_user '' '%s CPU User Time (%d%% = %d core%s)' 'cpu time %%' cpu %s.cpu_user stacked 20020 %d\n", type, title, (processors * 100), processors, (processors>1)?"s":"", type, update_every);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, hz * RATES_DETAIL / 100LLU);
+ }
+
+ fprintf(stdout, "CHART %s.cpu_system '' '%s CPU System Time (%d%% = %d core%s)' 'cpu time %%' cpu %s.cpu_system stacked 20021 %d\n", type, title, (processors * 100), processors, (processors>1)?"s":"", type, update_every);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, hz * RATES_DETAIL / 100LLU);
+ }
+
+ if(show_guest_time) {
+ fprintf(stdout, "CHART %s.cpu_guest '' '%s CPU Guest Time (%d%% = %d core%s)' 'cpu time %%' cpu %s.cpu_system stacked 20022 %d\n", type, title, (processors * 100), processors, (processors > 1) ? "s" : "", type, update_every);
+ for (w = root; w; w = w->next) {
+ if(unlikely(w->exposed))
+ fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, hz * RATES_DETAIL / 100LLU);
+ }
+ }
+
+ fprintf(stdout, "CHART %s.major_faults '' '%s Major Page Faults (swap read)' 'page faults/s' swap %s.major_faults stacked 20010 %d\n", type, title, type, update_every);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, RATES_DETAIL);
+ }
+
+ fprintf(stdout, "CHART %s.minor_faults '' '%s Minor Page Faults' 'page faults/s' mem %s.minor_faults stacked 20011 %d\n", type, title, type, update_every);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, RATES_DETAIL);
+ }
+
+#ifndef __FreeBSD__
+ fprintf(stdout, "CHART %s.lreads '' '%s Disk Logical Reads' 'kilobytes/s' disk %s.lreads stacked 20042 %d\n", type, title, type, update_every);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, 1024LLU * RATES_DETAIL);
+ }
+
+ fprintf(stdout, "CHART %s.lwrites '' '%s I/O Logical Writes' 'kilobytes/s' disk %s.lwrites stacked 20042 %d\n", type, title, type, update_every);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, 1024LLU * RATES_DETAIL);
+ }
+#endif
- fprintf(stdout, "DIMENSION %s '' incremental 1 %d noreset\n", w->name, 1024);
- }
+#ifdef __FreeBSD__
+ fprintf(stdout, "CHART %s.preads '' '%s Disk Reads' 'blocks/s' disk %s.preads stacked 20002 %d\n", type, title, type, update_every);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, RATES_DETAIL);
+ }
+
+ fprintf(stdout, "CHART %s.pwrites '' '%s Disk Writes' 'blocks/s' disk %s.pwrites stacked 20002 %d\n", type, title, type, update_every);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, RATES_DETAIL);
+ }
+#else
+ fprintf(stdout, "CHART %s.preads '' '%s Disk Reads' 'kilobytes/s' disk %s.preads stacked 20002 %d\n", type, title, type, update_every);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, 1024LLU * RATES_DETAIL);
+ }
+
+ fprintf(stdout, "CHART %s.pwrites '' '%s Disk Writes' 'kilobytes/s' disk %s.pwrites stacked 20002 %d\n", type, title, type, update_every);
+ for (w = root; w ; w = w->next) {
+ if(unlikely(w->exposed))
+ fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, 1024LLU * RATES_DETAIL);
+ }
+#endif
- fprintf(stdout, "CHART %s.lwrites '' '%s I/O Logical Writes' 'kilobytes/s' disk %s.lwrites stacked 20042 %d\n", type, title, type, update_every);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
+ if(enable_file_charts) {
+ fprintf(stdout, "CHART %s.files '' '%s Open Files' 'open files' disk %s.files stacked 20050 %d\n", type,
+ title, type, update_every);
+ for (w = root; w; w = w->next) {
+ if (unlikely(w->exposed))
+ fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name);
+ }
+
+ fprintf(stdout, "CHART %s.sockets '' '%s Open Sockets' 'open sockets' net %s.sockets stacked 20051 %d\n",
+ type, title, type, update_every);
+ for (w = root; w; w = w->next) {
+ if (unlikely(w->exposed))
+ fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name);
+ }
+
+ fprintf(stdout, "CHART %s.pipes '' '%s Pipes' 'open pipes' processes %s.pipes stacked 20053 %d\n", type,
+ title, type, update_every);
+ for (w = root; w; w = w->next) {
+ if (unlikely(w->exposed))
+ fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name);
+ }
+ }
+}
- fprintf(stdout, "DIMENSION %s '' incremental 1 %d noreset\n", w->name, 1024);
- }
- fprintf(stdout, "CHART %s.preads '' '%s Disk Reads' 'kilobytes/s' disk %s.preads stacked 20002 %d\n", type, title, type, update_every);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
+// ----------------------------------------------------------------------------
+// parse command line arguments
- fprintf(stdout, "DIMENSION %s '' incremental 1 %d noreset\n", w->name, 1024);
- }
+int check_proc_1_io() {
+ int ret = 0;
- fprintf(stdout, "CHART %s.pwrites '' '%s Disk Writes' 'kilobytes/s' disk %s.pwrites stacked 20002 %d\n", type, title, type, update_every);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
+ procfile *ff = procfile_open("/proc/1/io", NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
+ if(!ff) goto cleanup;
- fprintf(stdout, "DIMENSION %s '' incremental 1 %d noreset\n", w->name, 1024);
- }
+ ff = procfile_readall(ff);
+ if(!ff) goto cleanup;
- fprintf(stdout, "CHART %s.files '' '%s Open Files' 'open files' disk %s.files stacked 20050 %d\n", type, title, type, update_every);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
+ ret = 1;
- fprintf(stdout, "DIMENSION %s '' absolute 1 1 noreset\n", w->name);
- }
+cleanup:
+ procfile_close(ff);
+ return ret;
+}
- fprintf(stdout, "CHART %s.sockets '' '%s Open Sockets' 'open sockets' net %s.sockets stacked 20051 %d\n", type, title, type, update_every);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
+static void parse_args(int argc, char **argv)
+{
+ int i, freq = 0;
+ char *name = NULL;
+
+ for(i = 1; i < argc; i++) {
+ if(!freq) {
+ int n = (int)str2l(argv[i]);
+ if(n > 0) {
+ freq = n;
+ continue;
+ }
+ }
+
+ if(strcmp("version", argv[i]) == 0 || strcmp("-v", argv[i]) == 0) {
+ printf("apps.plugin %s\n", VERSION);
+ exit(0);
+ }
+
+ if(strcmp("test-permissions", argv[i]) == 0 || strcmp("-t", argv[i]) == 0) {
+ if(!check_proc_1_io()) {
+ perror("Tried to read /proc/1/io and it failed");
+ exit(1);
+ }
+ printf("OK\n");
+ exit(0);
+ }
+
+ if(strcmp("debug", argv[i]) == 0) {
+ debug = 1;
+ // debug_flags = 0xffffffff;
+ continue;
+ }
+
+ if(strcmp("no-childs", argv[i]) == 0 || strcmp("without-childs", argv[i]) == 0) {
+ include_exited_childs = 0;
+ continue;
+ }
+
+ if(strcmp("with-childs", argv[i]) == 0) {
+ include_exited_childs = 1;
+ continue;
+ }
+
+ if(strcmp("with-guest", argv[i]) == 0) {
+ enable_guest_charts = 1;
+ continue;
+ }
+
+ if(strcmp("no-guest", argv[i]) == 0 || strcmp("without-guest", argv[i]) == 0) {
+ enable_guest_charts = 0;
+ continue;
+ }
+
+ if(strcmp("with-files", argv[i]) == 0) {
+ enable_file_charts = 1;
+ continue;
+ }
+
+ if(strcmp("no-files", argv[i]) == 0 || strcmp("without-files", argv[i]) == 0) {
+ enable_file_charts = 0;
+ continue;
+ }
+
+ if(strcmp("no-users", argv[i]) == 0 || strcmp("without-users", argv[i]) == 0) {
+ enable_users_charts = 0;
+ continue;
+ }
+
+ if(strcmp("no-groups", argv[i]) == 0 || strcmp("without-groups", argv[i]) == 0) {
+ enable_groups_charts = 0;
+ continue;
+ }
+
+ if(strcmp("-h", argv[i]) == 0 || strcmp("--help", argv[i]) == 0) {
+ fprintf(stderr,
+ "\n"
+ " netdata apps.plugin %s\n"
+ " Copyright (C) 2016-2017 Costa Tsaousis <costa@tsaousis.gr>\n"
+ " Released under GNU General Public License v3 or later.\n"
+ " All rights reserved.\n"
+ "\n"
+ " This program is a data collector plugin for netdata.\n"
+ "\n"
+ " Valid command line options:\n"
+ "\n"
+ " SECONDS set the data collection frequency\n"
+ "\n"
+ " debug enable debugging (lot of output)\n"
+ "\n"
+ " with-childs\n"
+ " without-childs enable / disable aggregating exited\n"
+ " children resources into parents\n"
+ " (default is enabled)\n"
+ "\n"
+ " with-guest\n"
+ " without-guest enable / disable reporting guest charts\n"
+ " (default is disabled)\n"
+ "\n"
+ " with-files\n"
+ " without-files enable / disable reporting files, sockets, pipes\n"
+ " (default is enabled)\n"
+ "\n"
+ " NAME read apps_NAME.conf instead of\n"
+ " apps_groups.conf\n"
+ " (default NAME=groups)\n"
+ "\n"
+ " version print program version and exit\n"
+ "\n"
+ , VERSION
+ );
+ exit(1);
+ }
+
+ if(!name) {
+ name = argv[i];
+ continue;
+ }
+
+ error("Cannot understand option %s", argv[i]);
+ exit(1);
+ }
+
+ if(freq > 0) update_every = freq;
+ if(!name) name = "groups";
+
+ if(read_apps_groups_conf(name)) {
+ error("Cannot read process groups '%s/apps_%s.conf'. There are no internal defaults. Failing.", config_dir, name);
+ exit(1);
+ }
+}
- fprintf(stdout, "DIMENSION %s '' absolute 1 1 noreset\n", w->name);
- }
+static int am_i_running_as_root() {
+ uid_t uid = getuid(), euid = geteuid();
- fprintf(stdout, "CHART %s.pipes '' '%s Pipes' 'open pipes' processes %s.pipes stacked 20053 %d\n", type, title, type, update_every);
- for (w = root; w ; w = w->next) {
- if(w->target || (!w->processes && !w->exposed)) continue;
+ if(uid == 0 || euid == 0) {
+ if(debug) info("I am running with escalated privileges, uid = %u, euid = %u.", uid, euid);
+ return 1;
+ }
- fprintf(stdout, "DIMENSION %s '' absolute 1 1 noreset\n", w->name);
- }
+ if(debug) info("I am not running with escalated privileges, uid = %u, euid = %u.", uid, euid);
+ return 0;
}
-
-// ----------------------------------------------------------------------------
-// parse command line arguments
-
-void parse_args(int argc, char **argv)
-{
- int i, freq = 0;
- char *name = NULL;
-
- for(i = 1; i < argc; i++) {
- if(!freq) {
- int n = atoi(argv[i]);
- if(n > 0) {
- freq = n;
- continue;
- }
- }
-
- if(strcmp("debug", argv[i]) == 0) {
- debug = 1;
- debug_flags = 0xffffffff;
- continue;
- }
-
- if(!name) {
- name = argv[i];
- continue;
- }
-
- error("Cannot understand option %s", argv[i]);
- exit(1);
- }
-
- if(freq > 0) update_every = freq;
- if(!name) name = "groups";
-
- if(read_apps_groups_conf(name)) {
- error("Cannot read process groups %s", name);
- exit(1);
- }
+#ifdef HAVE_CAPABILITY
+static int check_capabilities() {
+ cap_t caps = cap_get_proc();
+ if(!caps) {
+ error("Cannot get current capabilities.");
+ return 0;
+ }
+ else if(debug)
+ info("Received my capabilities from the system.");
+
+ int ret = 1;
+
+ cap_flag_value_t cfv = CAP_CLEAR;
+ if(cap_get_flag(caps, CAP_DAC_READ_SEARCH, CAP_EFFECTIVE, &cfv) == -1) {
+ error("Cannot find if CAP_DAC_READ_SEARCH is effective.");
+ ret = 0;
+ }
+ else {
+ if(cfv != CAP_SET) {
+ error("apps.plugin should run with CAP_DAC_READ_SEARCH.");
+ ret = 0;
+ }
+ else if(debug)
+ info("apps.plugin runs with CAP_DAC_READ_SEARCH.");
+ }
+
+ cfv = CAP_CLEAR;
+ if(cap_get_flag(caps, CAP_SYS_PTRACE, CAP_EFFECTIVE, &cfv) == -1) {
+ error("Cannot find if CAP_SYS_PTRACE is effective.");
+ ret = 0;
+ }
+ else {
+ if(cfv != CAP_SET) {
+ error("apps.plugin should run with CAP_SYS_PTRACE.");
+ ret = 0;
+ }
+ else if(debug)
+ info("apps.plugin runs with CAP_SYS_PTRACE.");
+ }
+
+ cap_free(caps);
+
+ return ret;
}
+#else
+static int check_capabilities() {
+ return 0;
+}
+#endif
-int main(int argc, char **argv)
-{
- // debug_flags = D_PROCFILE;
+int main(int argc, char **argv) {
+ // debug_flags = D_PROCFILE;
- // set the name for logging
- program_name = "apps.plugin";
+ // set the name for logging
+ program_name = "apps.plugin";
- // disable syslog for apps.plugin
- error_log_syslog = 0;
+ // disable syslog for apps.plugin
+ error_log_syslog = 0;
- // set errors flood protection to 100 logs per hour
- error_log_errors_per_period = 100;
- error_log_throttle_period = 3600;
+ // set errors flood protection to 100 logs per hour
+ error_log_errors_per_period = 100;
+ error_log_throttle_period = 3600;
- host_prefix = getenv("NETDATA_HOST_PREFIX");
- if(host_prefix == NULL) {
- info("NETDATA_HOST_PREFIX is not passed from netdata");
- host_prefix = "";
- }
- else info("Found NETDATA_HOST_PREFIX='%s'", host_prefix);
+ netdata_configured_host_prefix = getenv("NETDATA_HOST_PREFIX");
+ if(netdata_configured_host_prefix == NULL) {
+ // info("NETDATA_HOST_PREFIX is not passed from netdata");
+ netdata_configured_host_prefix = "";
+ }
+ // else info("Found NETDATA_HOST_PREFIX='%s'", netdata_configured_host_prefix);
- config_dir = getenv("NETDATA_CONFIG_DIR");
- if(config_dir == NULL) {
- info("NETDATA_CONFIG_DIR is not passed from netdata");
- config_dir = CONFIG_DIR;
- }
- else info("Found NETDATA_CONFIG_DIR='%s'", config_dir);
+ config_dir = getenv("NETDATA_CONFIG_DIR");
+ if(config_dir == NULL) {
+ // info("NETDATA_CONFIG_DIR is not passed from netdata");
+ config_dir = CONFIG_DIR;
+ }
+ // else info("Found NETDATA_CONFIG_DIR='%s'", config_dir);
#ifdef NETDATA_INTERNAL_CHECKS
- if(debug_flags != 0) {
- struct rlimit rl = { RLIM_INFINITY, RLIM_INFINITY };
- if(setrlimit(RLIMIT_CORE, &rl) != 0)
- info("Cannot request unlimited core dumps for debugging... Proceeding anyway...");
- prctl(PR_SET_DUMPABLE, 1, 0, 0, 0);
- }
+ if(debug_flags != 0) {
+ struct rlimit rl = { RLIM_INFINITY, RLIM_INFINITY };
+ if(setrlimit(RLIMIT_CORE, &rl) != 0)
+ info("Cannot request unlimited core dumps for debugging... Proceeding anyway...");
+#ifdef HAVE_SYS_PRCTL_H
+ prctl(PR_SET_DUMPABLE, 1, 0, 0, 0);
+#endif
+ }
#endif /* NETDATA_INTERNAL_CHECKS */
- info("starting...");
+ procfile_adaptive_initial_allocation = 1;
- procfile_adaptive_initial_allocation = 1;
+ time_t started_t = now_monotonic_sec();
+ get_system_HZ();
+ get_system_pid_max();
+ get_system_cpus();
- time_t started_t = time(NULL);
- time_t current_t;
- get_HZ();
- pid_max = get_system_pid_max();
- processors = get_system_cpus();
+ parse_args(argc, argv);
- parse_args(argc, argv);
+ if(!check_capabilities() && !am_i_running_as_root() && !check_proc_1_io()) {
+ uid_t uid = getuid(), euid = geteuid();
+#ifdef HAVE_CAPABILITY
+ error("apps.plugin should either run as root (now running with uid %u, euid %u) or have special capabilities. "
+ "Without these, apps.plugin cannot report disk I/O utilization of other processes. "
+ "To enable capabilities run: sudo setcap cap_dac_read_search,cap_sys_ptrace+ep %s; "
+ "To enable setuid to root run: sudo chown root %s; sudo chmod 4755 %s; "
+ , uid, euid, argv[0], argv[0], argv[0]
+ );
+#else
+ error("apps.plugin should either run as root (now running with uid %u, euid %u) or have special capabilities. "
+ "Without these, apps.plugin cannot report disk I/O utilization of other processes. "
+ "Your system does not support capabilities. "
+ "To enable setuid to root run: sudo chown root %s; sudo chmod 4755 %s; "
+ , uid, euid, argv[0], argv[0]
+ );
+#endif
+ }
- all_pids = calloc(sizeof(struct pid_stat *), (size_t) pid_max);
- if(!all_pids) {
- error("Cannot allocate %lu bytes of memory.", sizeof(struct pid_stat *) * pid_max);
- printf("DISABLE\n");
- exit(1);
- }
+ info("started on pid %d", getpid());
- fprintf(stdout, "CHART netdata.apps_cpu '' 'Apps Plugin CPU' 'milliseconds/s' apps.plugin netdata.apps_cpu stacked 140000 %d\n", update_every);
- fprintf(stdout, "DIMENSION user '' incremental 1 %d\n", 1000);
- fprintf(stdout, "DIMENSION system '' incremental 1 %d\n", 1000);
+#if (ALL_PIDS_ARE_READ_INSTANTLY == 0)
+ all_pids_sortlist = callocz(sizeof(pid_t), (size_t)pid_max);
+#endif
- fprintf(stdout, "CHART netdata.apps_files '' 'Apps Plugin Files' 'files/s' apps.plugin netdata.apps_files line 140001 %d\n", update_every);
- fprintf(stdout, "DIMENSION files '' incremental 1 1\n");
- fprintf(stdout, "DIMENSION pids '' absolute 1 1\n");
- fprintf(stdout, "DIMENSION fds '' absolute 1 1\n");
- fprintf(stdout, "DIMENSION targets '' absolute 1 1\n");
+ all_pids = callocz(sizeof(struct pid_stat *), (size_t) pid_max);
+ usec_t step = update_every * USEC_PER_SEC;
+ global_iterations_counter = 1;
+ heartbeat_t hb;
+ heartbeat_init(&hb);
+ for(;1; global_iterations_counter++) {
-#ifndef PROFILING_MODE
- unsigned long long sunext = (time(NULL) - (time(NULL) % update_every) + update_every) * 1000000ULL;
- unsigned long long sunow;
-#endif /* PROFILING_MODE */
+#ifdef NETDATA_PROFILING
+#warning "compiling for profiling"
+ static int profiling_count=0;
+ profiling_count++;
+ if(unlikely(profiling_count > 1000)) exit(0);
+#else
+ heartbeat_next(&hb, step);
+#endif
- unsigned long long counter = 1;
- for(;1; counter++) {
-#ifndef PROFILING_MODE
- // delay until it is our time to run
- while((sunow = timems()) < sunext)
- usleep((useconds_t)(sunext - sunow));
+ if(!collect_data_for_all_processes()) {
+ error("Cannot collect /proc data for running processes. Disabling apps.plugin...");
+ printf("DISABLE\n");
+ exit(1);
+ }
- // find the next time we need to run
- while(timems() > sunext)
- sunext += update_every * 1000000ULL;
-#endif /* PROFILING_MODE */
+ calculate_netdata_statistics();
+ normalize_utilization(apps_groups_root_target);
- if(!collect_data_for_all_processes_from_proc()) {
- error("Cannot collect /proc data for running processes. Disabling apps.plugin...");
- printf("DISABLE\n");
- exit(1);
- }
+ usec_t dt = send_resource_usage_to_netdata();
- calculate_netdata_statistics();
+ // this is smart enough to show only newly added apps, when needed
+ send_charts_updates_to_netdata(apps_groups_root_target, "apps", "Apps");
- unsigned long long dt = send_resource_usage_to_netdata();
+ if(likely(enable_users_charts))
+ send_charts_updates_to_netdata(users_root_target, "users", "Users");
- // this is smart enough to show only newly added apps, when needed
- send_charts_updates_to_netdata(apps_groups_root_target, "apps", "Apps");
- send_charts_updates_to_netdata(users_root_target, "users", "Users");
- send_charts_updates_to_netdata(groups_root_target, "groups", "User Groups");
+ if(likely(enable_groups_charts))
+ send_charts_updates_to_netdata(groups_root_target, "groups", "User Groups");
- send_collected_data_to_netdata(apps_groups_root_target, "apps", dt);
- send_collected_data_to_netdata(users_root_target, "users", dt);
- send_collected_data_to_netdata(groups_root_target, "groups", dt);
+ send_collected_data_to_netdata(apps_groups_root_target, "apps", dt);
- if(debug) fprintf(stderr, "apps.plugin: done Loop No %llu\n", counter);
- fflush(NULL);
+ if(likely(enable_users_charts))
+ send_collected_data_to_netdata(users_root_target, "users", dt);
- current_t = time(NULL);
+ if(likely(enable_groups_charts))
+ send_collected_data_to_netdata(groups_root_target, "groups", dt);
-#ifndef PROFILING_MODE
- // restart check (14400 seconds)
- if(current_t - started_t > 14400) exit(0);
-#else
- if(current_t - started_t > 10) exit(0);
-#endif /* PROFILING_MODE */
- }
+ fflush(stdout);
+
+ show_guest_time_old = show_guest_time;
+
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: done Loop No %zu\n", global_iterations_counter);
+
+ // restart check (14400 seconds)
+ if(now_monotonic_sec() - started_t > 14400) exit(0);
+ }
}