3 // 1. support -ch option to set global /proc and /sys prefix
4 // 2. disable RESET_OR_OVERFLOW check in charts
7 #define __STDC_FORMAT_MACROS
15 #include <sys/types.h>
19 #include <sys/resource.h>
31 #include <arpa/inet.h>
33 #include "../common.h"
36 #include "../procfile.h"
38 #define MAX_COMPARE_NAME 15
41 unsigned long long Hertz = 1;
48 unsigned long long file_counter = 0;
50 #define PROC_BUFFER 4096
52 // ----------------------------------------------------------------------------
59 } allocations = { 0, 0, 0 };
61 #define MALLOC_MARK (uint32_t)(0x0BADCAFE)
62 #define MALLOC_PREFIX (sizeof(uint32_t) * 2)
63 #define MALLOC_SUFFIX (sizeof(uint32_t))
64 #define MALLOC_OVERHEAD (MALLOC_PREFIX + MALLOC_SUFFIX)
66 void *mark_allocation(void *allocated_ptr, size_t size_without_overheads) {
67 uint32_t *real_ptr = (uint32_t *)allocated_ptr;
68 real_ptr[0] = MALLOC_MARK;
69 real_ptr[1] = size_without_overheads;
71 uint32_t *end_ptr = (uint32_t *)(allocated_ptr + MALLOC_PREFIX + size_without_overheads);
72 end_ptr[0] = MALLOC_MARK;
74 // fprintf(stderr, "MEMORY_POINTER: Allocated at %p, returning %p.\n", allocated_ptr, (void *)(allocated_ptr + MALLOC_PREFIX));
76 return (void *)(allocated_ptr + MALLOC_PREFIX);
79 void *check_allocation(const char *file, int line, const char *function, void *marked_ptr, size_t *size_without_overheads_ptr) {
80 uint32_t *real_ptr = (uint32_t *)(marked_ptr - MALLOC_PREFIX);
82 // 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);
84 if(real_ptr[0] != MALLOC_MARK) fatal("MEMORY: prefix MARK is not valid for %s/%u@%s.", function, line, file);
86 size_t size = real_ptr[1];
88 uint32_t *end_ptr = (uint32_t *)(marked_ptr + size);
89 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);
91 if(size_without_overheads_ptr) *size_without_overheads_ptr = size;
96 void *malloc_debug(const char *file, int line, const char *function, size_t size) {
97 void *ptr = malloc(size + MALLOC_OVERHEAD);
98 if(!ptr) fatal("MEMORY: Cannot allocate %zu bytes for %s/%u@%s.", size, function, line, file);
100 allocations.allocated += size;
101 allocations.allocations++;
103 debug(D_MEMORY, "MEMORY: Allocated %zu bytes for %s/%u@%s."
104 " Status: allocated %zu in %zu allocs."
106 , function, line, file
107 , allocations.allocated
108 , allocations.allocations
111 if(allocations.allocated > allocations.allocated_max) {
112 debug(D_MEMORY, "MEMORY: total allocation peak increased from %zu to %zu", allocations.allocated_max, allocations.allocated);
113 allocations.allocated_max = allocations.allocated;
117 check_allocation(file, line, function, mark_allocation(ptr, size), &csize);
119 fatal("Invalid size.");
122 return mark_allocation(ptr, size);
125 void *calloc_debug(const char *file, int line, const char *function, size_t nmemb, size_t size) {
126 void *ptr = malloc_debug(file, line, function, (nmemb * size));
127 bzero(ptr, nmemb * size);
131 void free_debug(const char *file, int line, const char *function, void *ptr) {
133 void *real_ptr = check_allocation(file, line, function, ptr, &size);
135 bzero(real_ptr, size + MALLOC_OVERHEAD);
138 allocations.allocated -= size;
139 allocations.allocations--;
141 debug(D_MEMORY, "MEMORY: freed %zu bytes for %s/%u@%s."
142 " Status: allocated %zu in %zu allocs."
144 , function, line, file
145 , allocations.allocated
146 , allocations.allocations
150 void *realloc_debug(const char *file, int line, const char *function, void *ptr, size_t size) {
151 if(!ptr) return malloc_debug(file, line, function, size);
152 if(!size) { free_debug(file, line, function, ptr); return NULL; }
155 void *real_ptr = check_allocation(file, line, function, ptr, &old_size);
157 void *new_ptr = realloc(real_ptr, size + MALLOC_OVERHEAD);
158 if(!new_ptr) fatal("MEMORY: Cannot allocate %zu bytes for %s/%u@%s.", size, function, line, file);
160 allocations.allocated += size;
161 allocations.allocated -= old_size;
163 debug(D_MEMORY, "MEMORY: Re-allocated from %zu to %zu bytes for %s/%u@%s."
164 " Status: allocated %z in %zu allocs."
166 , function, line, file
167 , allocations.allocated
168 , allocations.allocations
171 if(allocations.allocated > allocations.allocated_max) {
172 debug(D_MEMORY, "MEMORY: total allocation peak increased from %zu to %zu", allocations.allocated_max, allocations.allocated);
173 allocations.allocated_max = allocations.allocated;
176 return mark_allocation(new_ptr, size);
179 char *strdup_debug(const char *file, int line, const char *function, const char *ptr) {
186 char *p = malloc_debug(file, line, function, size);
187 if(!p) fatal("Cannot allocate %zu bytes.", size);
189 memcpy(p, ptr, size);
193 #define malloc(size) malloc_debug(__FILE__, __LINE__, __FUNCTION__, (size))
194 #define calloc(nmemb, size) calloc_debug(__FILE__, __LINE__, __FUNCTION__, (nmemb), (size))
195 #define realloc(ptr, size) realloc_debug(__FILE__, __LINE__, __FUNCTION__, (ptr), (size))
196 #define free(ptr) free_debug(__FILE__, __LINE__, __FUNCTION__, (ptr))
201 #define strdup(ptr) strdup_debug(__FILE__, __LINE__, __FUNCTION__, (ptr))
203 // ----------------------------------------------------------------------------
208 long get_processors(void) {
211 ff = procfile_reopen(ff, "/proc/stat", "", PROCFILE_FLAG_DEFAULT);
214 ff = procfile_readall(ff);
216 // procfile_close(ff);
221 for(i = 0; i < procfile_lines(ff); i++) {
222 if(!procfile_linewords(ff, i)) continue;
224 if(strncmp(procfile_lineword(ff, i, 0), "cpu", 3) == 0) processors++;
227 if(processors < 1) processors = 1;
229 // procfile_close(ff);
233 long get_pid_max(void) {
236 ff = procfile_reopen(ff, "/proc/sys/kernel/pid_max", "", PROCFILE_FLAG_DEFAULT);
239 ff = procfile_readall(ff);
241 // procfile_close(ff);
245 mpid = atol(procfile_lineword(ff, 0, 0));
246 if(mpid) mpid = 32768;
248 // procfile_close(ff);
252 unsigned long long get_hertz(void)
254 unsigned long long hz = 1;
257 if((hz = sysconf(_SC_CLK_TCK)) > 0) {
263 hz = (unsigned long long)HZ; /* <asm/param.h> */
265 /* If 32-bit or big-endian (not Alpha or ia64), assume HZ is 100. */
266 hz = (sizeof(long)==sizeof(int) || htons(999)==999) ? 100UL : 1024UL;
269 error("Unknown HZ value. Assuming %llu.", hz);
274 // ----------------------------------------------------------------------------
276 // target is the point to aggregate a process tree values
279 char compare[MAX_COMPARE_NAME + 1];
280 char id[MAX_NAME + 1];
281 char name[MAX_NAME + 1];
283 unsigned long long minflt;
284 unsigned long long cminflt;
285 unsigned long long majflt;
286 unsigned long long cmajflt;
287 unsigned long long utime;
288 unsigned long long stime;
289 unsigned long long cutime;
290 unsigned long long cstime;
291 unsigned long long num_threads;
292 unsigned long long rss;
294 unsigned long long fix_minflt;
295 unsigned long long fix_cminflt;
296 unsigned long long fix_majflt;
297 unsigned long long fix_cmajflt;
298 unsigned long long fix_utime;
299 unsigned long long fix_stime;
300 unsigned long long fix_cutime;
301 unsigned long long fix_cstime;
303 unsigned long long statm_size;
304 unsigned long long statm_resident;
305 unsigned long long statm_share;
306 unsigned long long statm_text;
307 unsigned long long statm_lib;
308 unsigned long long statm_data;
309 unsigned long long statm_dirty;
311 unsigned long long io_logical_bytes_read;
312 unsigned long long io_logical_bytes_written;
313 unsigned long long io_read_calls;
314 unsigned long long io_write_calls;
315 unsigned long long io_storage_bytes_read;
316 unsigned long long io_storage_bytes_written;
317 unsigned long long io_cancelled_write_bytes;
319 unsigned long long fix_io_logical_bytes_read;
320 unsigned long long fix_io_logical_bytes_written;
321 unsigned long long fix_io_read_calls;
322 unsigned long long fix_io_write_calls;
323 unsigned long long fix_io_storage_bytes_read;
324 unsigned long long fix_io_storage_bytes_written;
325 unsigned long long fix_io_cancelled_write_bytes;
328 unsigned long long openfiles;
329 unsigned long long openpipes;
330 unsigned long long opensockets;
331 unsigned long long openinotifies;
332 unsigned long long openeventfds;
333 unsigned long long opentimerfds;
334 unsigned long long opensignalfds;
335 unsigned long long openeventpolls;
336 unsigned long long openother;
338 unsigned long processes; // how many processes have been merged to this
339 int exposed; // if set, we have sent this to netdata
340 int hidden; // if set, we set the hidden flag on the dimension
343 struct target *target; // the one that will be reported to netdata
345 } *target_root = NULL, *default_target = NULL;
349 // find or create a target
350 // there are targets that are just agregated to other target (the second argument)
351 struct target *get_target(const char *id, struct target *target)
353 const char *nid = id;
354 if(nid[0] == '-') nid++;
357 for(w = target_root ; w ; w = w->next)
358 if(strncmp(nid, w->id, MAX_NAME) == 0) return w;
360 w = calloc(sizeof(struct target), 1);
362 error("Cannot allocate %lu bytes of memory", (unsigned long)sizeof(struct target));
366 strncpy(w->id, nid, MAX_NAME);
367 strncpy(w->name, nid, MAX_NAME);
368 strncpy(w->compare, nid, MAX_COMPARE_NAME);
369 if(id[0] == '-') w->hidden = 1;
373 w->next = target_root;
376 if(debug) fprintf(stderr, "apps.plugin: adding hook for process '%s', compare '%s' on target '%s'\n", w->id, w->compare, w->target?w->target->id:"");
381 // read the process groups file
382 int read_process_groups(const char *name)
385 char filename[FILENAME_MAX + 1];
387 snprintf(filename, FILENAME_MAX, "%s/apps_%s.conf", CONFIG_DIR, name);
389 if(debug) fprintf(stderr, "apps.plugin: process groups file: '%s'\n", filename);
390 FILE *fp = fopen(filename, "r");
392 error("Cannot open file '%s' (%s)", filename, strerror(errno));
397 while(fgets(buffer, 4096, fp) != NULL) {
398 int whidden = 0, wdebug = 0;
401 // if(debug) fprintf(stderr, "apps.plugin: \tread %s\n", buffer);
403 char *s = buffer, *t, *p;
405 if(!s || !*s || *s == '#') continue;
407 if(debug) fprintf(stderr, "apps.plugin: \tread %s\n", s);
412 if(!t || !*t) continue;
434 if(debug) fprintf(stderr, "apps.plugin: \t\ttarget %s\n", t);
436 struct target *w = NULL;
440 while((p = strsep(&s, " "))) {
442 if(!p || !*p) continue;
444 struct target *n = get_target(p, w);
452 if(w) strncpy(w->name, t, MAX_NAME);
453 if(!count) error("The line %ld on file '%s', for group '%s' does not state any process names.", line, filename, t);
457 default_target = get_target("+p!o@w#e$i^r&7*5(-i)l-o_", NULL); // match nothing
458 strncpy(default_target->name, "other", MAX_NAME);
464 // ----------------------------------------------------------------------------
465 // data to store for each pid
470 char comm[MAX_COMPARE_NAME + 1];
478 unsigned long long minflt;
479 unsigned long long cminflt;
480 unsigned long long majflt;
481 unsigned long long cmajflt;
482 unsigned long long utime;
483 unsigned long long stime;
484 unsigned long long cutime;
485 unsigned long long cstime;
489 // int64_t itrealvalue;
490 // unsigned long long starttime;
491 // unsigned long long vsize;
492 unsigned long long rss;
493 // unsigned long long rsslim;
494 // unsigned long long starcode;
495 // unsigned long long endcode;
496 // unsigned long long startstack;
497 // unsigned long long kstkesp;
498 // unsigned long long kstkeip;
501 // uint64_t sigignore;
502 // uint64_t sigcatch;
506 // int32_t exit_signal;
507 // int32_t processor;
508 // uint32_t rt_priority;
510 // unsigned long long delayacct_blkio_ticks;
511 // uint64_t guest_time;
512 // int64_t cguest_time;
514 unsigned long long statm_size;
515 unsigned long long statm_resident;
516 unsigned long long statm_share;
517 unsigned long long statm_text;
518 unsigned long long statm_lib;
519 unsigned long long statm_data;
520 unsigned long long statm_dirty;
522 unsigned long long io_logical_bytes_read;
523 unsigned long long io_logical_bytes_written;
524 unsigned long long io_read_calls;
525 unsigned long long io_write_calls;
526 unsigned long long io_storage_bytes_read;
527 unsigned long long io_storage_bytes_written;
528 unsigned long long io_cancelled_write_bytes;
530 #ifdef INCLUDE_CHILDS
531 unsigned long long old_utime;
532 unsigned long long old_stime;
533 unsigned long long old_minflt;
534 unsigned long long old_majflt;
536 unsigned long long old_cutime;
537 unsigned long long old_cstime;
538 unsigned long long old_cminflt;
539 unsigned long long old_cmajflt;
541 unsigned long long fix_cutime;
542 unsigned long long fix_cstime;
543 unsigned long long fix_cminflt;
544 unsigned long long fix_cmajflt;
546 unsigned long long diff_cutime;
547 unsigned long long diff_cstime;
548 unsigned long long diff_cminflt;
549 unsigned long long diff_cmajflt;
552 int *fds; // array of fds it uses
553 int fds_size; // the size of the fds array
555 int childs; // number of processes directly referencing this
556 int updated; // 1 when update
557 int merged; // 1 when it has been merged to its parent
559 struct target *target;
560 struct pid_stat *parent;
561 struct pid_stat *prev;
562 struct pid_stat *next;
563 } *root_of_pids = NULL, **all_pids;
565 long all_pids_count = 0;
567 struct pid_stat *get_pid_entry(pid_t pid)
570 all_pids[pid]->new_entry = 0;
571 return all_pids[pid];
574 all_pids[pid] = calloc(sizeof(struct pid_stat), 1);
576 error("Cannot allocate %lu bytes of memory", (unsigned long)sizeof(struct pid_stat));
580 all_pids[pid]->fds = calloc(sizeof(int), 100);
581 if(!all_pids[pid]->fds)
582 error("Cannot allocate %ld bytes of memory", (unsigned long)(sizeof(int) * 100));
583 else all_pids[pid]->fds_size = 100;
585 if(root_of_pids) root_of_pids->prev = all_pids[pid];
586 all_pids[pid]->next = root_of_pids;
587 root_of_pids = all_pids[pid];
589 all_pids[pid]->pid = pid;
590 all_pids[pid]->new_entry = 1;
592 return all_pids[pid];
595 void del_pid_entry(pid_t pid)
597 if(!all_pids[pid]) return;
599 if(debug) fprintf(stderr, "apps.plugin: process %d %s exited, deleting it.\n", pid, all_pids[pid]->comm);
601 if(root_of_pids == all_pids[pid]) root_of_pids = all_pids[pid]->next;
602 if(all_pids[pid]->next) all_pids[pid]->next->prev = all_pids[pid]->prev;
603 if(all_pids[pid]->prev) all_pids[pid]->prev->next = all_pids[pid]->next;
605 if(all_pids[pid]->fds) free(all_pids[pid]->fds);
607 all_pids[pid] = NULL;
611 // ----------------------------------------------------------------------------
612 // update pids from proc
614 int read_proc_pid_stat(struct pid_stat *p) {
615 char filename[FILENAME_MAX + 1];
617 snprintf(filename, FILENAME_MAX, "/proc/%d/stat", p->pid);
619 ff = procfile_reopen(ff, filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
622 ff = procfile_readall(ff);
624 // procfile_close(ff);
631 p->comm[MAX_COMPARE_NAME] = '\0';
634 char *s = procfile_lineword(ff, 0, 1);
636 size_t len = strlen(s);
638 while(len && s[len - 1] != ')') {
639 if(blen < MAX_COMPARE_NAME) {
640 strncpy(&p->comm[blen], s, MAX_COMPARE_NAME - blen);
641 blen = strlen(p->comm);
645 s = procfile_lineword(ff, 0, 1+i);
648 if(len && s[len - 1] == ')') s[len - 1] = '\0';
649 if(blen < MAX_COMPARE_NAME)
650 strncpy(&p->comm[blen], s, MAX_COMPARE_NAME - blen);
652 // p->pid = atol(procfile_lineword(ff, 0, 0+i));
654 // p->state = *(procfile_lineword(ff, 0, 2+i));
655 p->ppid = atol(procfile_lineword(ff, 0, 3+i));
656 // p->pgrp = atol(procfile_lineword(ff, 0, 4+i));
657 // p->session = atol(procfile_lineword(ff, 0, 5+i));
658 // p->tty_nr = atol(procfile_lineword(ff, 0, 6+i));
659 // p->tpgid = atol(procfile_lineword(ff, 0, 7+i));
660 // p->flags = strtoull(procfile_lineword(ff, 0, 8+i), NULL, 10);
661 p->minflt = strtoull(procfile_lineword(ff, 0, 9+i), NULL, 10);
662 p->cminflt = strtoull(procfile_lineword(ff, 0, 10+i), NULL, 10);
663 p->majflt = strtoull(procfile_lineword(ff, 0, 11+i), NULL, 10);
664 p->cmajflt = strtoull(procfile_lineword(ff, 0, 12+i), NULL, 10);
665 p->utime = strtoull(procfile_lineword(ff, 0, 13+i), NULL, 10);
666 p->stime = strtoull(procfile_lineword(ff, 0, 14+i), NULL, 10);
667 p->cutime = strtoull(procfile_lineword(ff, 0, 15+i), NULL, 10);
668 p->cstime = strtoull(procfile_lineword(ff, 0, 16+i), NULL, 10);
669 // p->priority = strtoull(procfile_lineword(ff, 0, 17+i), NULL, 10);
670 // p->nice = strtoull(procfile_lineword(ff, 0, 18+i), NULL, 10);
671 p->num_threads = atol(procfile_lineword(ff, 0, 19+i));
672 // p->itrealvalue = strtoull(procfile_lineword(ff, 0, 20+i), NULL, 10);
673 // p->starttime = strtoull(procfile_lineword(ff, 0, 21+i), NULL, 10);
674 // p->vsize = strtoull(procfile_lineword(ff, 0, 22+i), NULL, 10);
675 p->rss = strtoull(procfile_lineword(ff, 0, 23+i), NULL, 10);
676 // p->rsslim = strtoull(procfile_lineword(ff, 0, 24+i), NULL, 10);
677 // p->starcode = strtoull(procfile_lineword(ff, 0, 25+i), NULL, 10);
678 // p->endcode = strtoull(procfile_lineword(ff, 0, 26+i), NULL, 10);
679 // p->startstack = strtoull(procfile_lineword(ff, 0, 27+i), NULL, 10);
680 // p->kstkesp = strtoull(procfile_lineword(ff, 0, 28+i), NULL, 10);
681 // p->kstkeip = strtoull(procfile_lineword(ff, 0, 29+i), NULL, 10);
682 // p->signal = strtoull(procfile_lineword(ff, 0, 30+i), NULL, 10);
683 // p->blocked = strtoull(procfile_lineword(ff, 0, 31+i), NULL, 10);
684 // p->sigignore = strtoull(procfile_lineword(ff, 0, 32+i), NULL, 10);
685 // p->sigcatch = strtoull(procfile_lineword(ff, 0, 33+i), NULL, 10);
686 // p->wchan = strtoull(procfile_lineword(ff, 0, 34+i), NULL, 10);
687 // p->nswap = strtoull(procfile_lineword(ff, 0, 35+i), NULL, 10);
688 // p->cnswap = strtoull(procfile_lineword(ff, 0, 36+i), NULL, 10);
689 // p->exit_signal = atol(procfile_lineword(ff, 0, 37+i));
690 // p->processor = atol(procfile_lineword(ff, 0, 38+i));
691 // p->rt_priority = strtoul(procfile_lineword(ff, 0, 39+i), NULL, 10);
692 // p->policy = strtoul(procfile_lineword(ff, 0, 40+i), NULL, 10);
693 // p->delayacct_blkio_ticks = strtoull(procfile_lineword(ff, 0, 41+i), NULL, 10);
694 // p->guest_time = strtoull(procfile_lineword(ff, 0, 42+i), NULL, 10);
695 // p->cguest_time = strtoull(procfile_lineword(ff, 0, 43), NULL, 10);
697 if(debug || (p->target && p->target->debug)) fprintf(stderr, "apps.plugin: VALUES: %s utime=%llu, stime=%llu, cutime=%llu, cstime=%llu, minflt=%llu, majflt=%llu, cminflt=%llu, cmajflt=%llu, threads=%d\n", p->comm, p->utime, p->stime, p->cutime, p->cstime, p->minflt, p->majflt, p->cminflt, p->cmajflt, p->num_threads);
699 // procfile_close(ff);
703 int read_proc_pid_statm(struct pid_stat *p) {
704 char filename[FILENAME_MAX + 1];
706 snprintf(filename, FILENAME_MAX, "/proc/%d/statm", p->pid);
708 ff = procfile_reopen(ff, filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
711 ff = procfile_readall(ff);
713 // procfile_close(ff);
719 p->statm_size = strtoull(procfile_lineword(ff, 0, 0), NULL, 10);
720 p->statm_resident = strtoull(procfile_lineword(ff, 0, 1), NULL, 10);
721 p->statm_share = strtoull(procfile_lineword(ff, 0, 2), NULL, 10);
722 p->statm_text = strtoull(procfile_lineword(ff, 0, 3), NULL, 10);
723 p->statm_lib = strtoull(procfile_lineword(ff, 0, 4), NULL, 10);
724 p->statm_data = strtoull(procfile_lineword(ff, 0, 5), NULL, 10);
725 p->statm_dirty = strtoull(procfile_lineword(ff, 0, 6), NULL, 10);
727 // procfile_close(ff);
731 int read_proc_pid_io(struct pid_stat *p) {
732 char filename[FILENAME_MAX + 1];
734 snprintf(filename, FILENAME_MAX, "/proc/%d/io", p->pid);
736 ff = procfile_reopen(ff, filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
739 ff = procfile_readall(ff);
741 // procfile_close(ff);
747 p->io_logical_bytes_read = strtoull(procfile_lineword(ff, 0, 1), NULL, 10);
748 p->io_logical_bytes_written = strtoull(procfile_lineword(ff, 1, 1), NULL, 10);
749 p->io_read_calls = strtoull(procfile_lineword(ff, 2, 1), NULL, 10);
750 p->io_write_calls = strtoull(procfile_lineword(ff, 3, 1), NULL, 10);
751 p->io_storage_bytes_read = strtoull(procfile_lineword(ff, 4, 1), NULL, 10);
752 p->io_storage_bytes_written = strtoull(procfile_lineword(ff, 5, 1), NULL, 10);
753 p->io_cancelled_write_bytes = strtoull(procfile_lineword(ff, 6, 1), NULL, 10);
755 // procfile_close(ff);
760 // ----------------------------------------------------------------------------
762 #ifdef INCLUDE_CHILDS
763 // print a tree view of all processes
764 int walk_down(pid_t pid, int level) {
765 struct pid_stat *p = NULL;
769 for(i = 0; i < level; i++) b[i] = '\t';
774 for(p = root_of_pids; p ; p = p->next) {
776 ret += walk_down(p->pid, level+1);
782 if(!p->updated) ret += 1;
783 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"
784 , b, p->comm, p->pid, p->updated?"OK":"KILLED", p->target->name, p->childs
785 , p->utime, p->utime - p->old_utime
786 , p->stime, p->stime - p->old_stime
787 , p->cutime, p->cutime - p->old_cutime
788 , p->cstime, p->cstime - p->old_cstime
789 , p->minflt, p->minflt - p->old_minflt
790 , p->majflt, p->majflt - p->old_majflt
791 , p->cminflt, p->cminflt - p->old_cminflt
792 , p->cmajflt, p->cmajflt - p->old_cmajflt
801 // ----------------------------------------------------------------------------
803 // this is used to keep a global list of all open files of the system
804 // it is needed in order to figure out the unique files a process tree has open
806 #define FILE_DESCRIPTORS_INCREASE_STEP 100
808 struct file_descriptor {
818 int all_files_len = 0;
819 int all_files_size = 0;
821 int file_descriptor_compare(void* a, void* b) {
822 if(((struct file_descriptor *)a)->magic != 0x0BADCAFE || ((struct file_descriptor *)b)->magic != 0x0BADCAFE)
823 error("Corrupted index data detected. Please report this.");
825 if(((struct file_descriptor *)a)->hash < ((struct file_descriptor *)b)->hash)
827 else if(((struct file_descriptor *)a)->hash > ((struct file_descriptor *)b)->hash)
829 else return strcmp(((struct file_descriptor *)a)->name, ((struct file_descriptor *)b)->name);
831 int file_descriptor_iterator(avl *a) { if(a) {}; return 0; }
833 avl_tree all_files_index = {
835 file_descriptor_compare
838 static struct file_descriptor *file_descriptor_find(const char *name, uint32_t hash) {
839 struct file_descriptor *result = NULL, tmp;
840 tmp.hash = (hash)?hash:simple_hash(name);
844 tmp.magic = 0x0BADCAFE;
846 avl_search(&all_files_index, (avl *)&tmp, file_descriptor_iterator, (avl **)&result);
850 #define file_descriptor_add(fd) avl_insert(&all_files_index, (avl *)(fd))
851 #define file_descriptor_remove(fd) avl_remove(&all_files_index, (avl *)(fd))
853 #define FILETYPE_OTHER 0
854 #define FILETYPE_FILE 1
855 #define FILETYPE_PIPE 2
856 #define FILETYPE_SOCKET 3
857 #define FILETYPE_INOTIFY 4
858 #define FILETYPE_EVENTFD 5
859 #define FILETYPE_EVENTPOLL 6
860 #define FILETYPE_TIMERFD 7
861 #define FILETYPE_SIGNALFD 8
863 void file_descriptor_not_used(int id)
865 if(id > 0 && id < all_files_size) {
866 if(all_files[id].magic != 0x0BADCAFE) {
867 error("Ignoring request to remove empty file id %d.", id);
871 if(debug) fprintf(stderr, "apps.plugin: decreasing slot %d (count = %ld).\n", id, all_files[id].count);
873 if(all_files[id].count > 0) {
874 all_files[id].count--;
876 if(!all_files[id].count) {
877 if(debug) fprintf(stderr, "apps.plugin: >> slot %d is empty.\n", id);
878 file_descriptor_remove(&all_files[id]);
879 all_files[id].magic = 0x00000000;
884 error("Request to decrease counter of fd %d (%s), while the use counter is 0", id, all_files[id].name);
886 else error("Request to decrease counter of fd %d, which is outside the array size (1 to %d)", id, all_files_size);
889 unsigned long file_descriptor_find_or_add(const char *name)
891 static int last_pos = 0;
892 uint32_t hash = simple_hash(name);
894 if(debug) fprintf(stderr, "apps.plugin: adding or finding name '%s' with hash %u\n", name, hash);
896 struct file_descriptor *fd = file_descriptor_find(name, hash);
899 if(debug) fprintf(stderr, "apps.plugin: >> found on slot %ld\n", fd->pos);
905 // check we have enough memory to add it
906 if(!all_files || all_files_len == all_files_size) {
907 void *old = all_files;
910 // there is no empty slot
911 if(debug) fprintf(stderr, "apps.plugin: extending fd array to %d entries\n", all_files_size + FILE_DESCRIPTORS_INCREASE_STEP);
912 all_files = realloc(all_files, (all_files_size + FILE_DESCRIPTORS_INCREASE_STEP) * sizeof(struct file_descriptor));
914 // if the address changed, we have to rebuild the index
915 // since all pointers are now invalid
916 if(old && old != (void *)all_files) {
917 if(debug) fprintf(stderr, "apps.plugin: >> re-indexing.\n");
918 all_files_index.root = NULL;
919 for(i = 0; i < all_files_size; i++) {
920 if(!all_files[i].count) continue;
921 file_descriptor_add(&all_files[i]);
923 if(debug) fprintf(stderr, "apps.plugin: >> re-indexing done.\n");
926 for(i = all_files_size; i < (all_files_size + FILE_DESCRIPTORS_INCREASE_STEP); i++) {
927 all_files[i].count = 0;
928 all_files[i].name = NULL;
929 all_files[i].magic = 0x00000000;
930 all_files[i].pos = i;
933 if(!all_files_size) all_files_len = 1;
934 all_files_size += FILE_DESCRIPTORS_INCREASE_STEP;
937 if(debug) fprintf(stderr, "apps.plugin: >> searching for empty slot.\n");
939 // search for an empty slot
941 for(i = 0, c = last_pos ; i < all_files_size ; i++, c++) {
942 if(c >= all_files_size) c = 0;
945 if(!all_files[c].count) {
946 if(debug) fprintf(stderr, "apps.plugin: >> Examining slot %d.\n", c);
948 if(all_files[c].magic == 0x0BADCAFE && all_files[c].name && file_descriptor_find(all_files[c].name, all_files[c].hash))
949 error("fd on position %d is not cleared properly. It still has %s in it.\n", c, all_files[c].name);
951 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);
952 if(all_files[c].name) free((void *)all_files[c].name);
953 all_files[c].name = NULL;
958 if(i == all_files_size) {
959 fatal("We should find an empty slot, but there isn't any");
962 if(debug) fprintf(stderr, "apps.plugin: >> updating slot %d.\n", c);
966 // else we have an empty slot in 'c'
968 int type = FILETYPE_OTHER;
969 if(name[0] == '/') type = FILETYPE_FILE;
970 else if(strncmp(name, "pipe:", 5) == 0) type = FILETYPE_PIPE;
971 else if(strncmp(name, "socket:", 7) == 0) type = FILETYPE_SOCKET;
972 else if(strcmp(name, "anon_inode:inotify") == 0 || strcmp(name, "inotify") == 0) type = FILETYPE_INOTIFY;
973 else if(strcmp(name, "anon_inode:[eventfd]") == 0) type = FILETYPE_EVENTFD;
974 else if(strcmp(name, "anon_inode:[eventpoll]") == 0) type = FILETYPE_EVENTPOLL;
975 else if(strcmp(name, "anon_inode:[timerfd]") == 0) type = FILETYPE_TIMERFD;
976 else if(strcmp(name, "anon_inode:[signalfd]") == 0) type = FILETYPE_SIGNALFD;
977 else if(strncmp(name, "anon_inode:", 11) == 0) {
978 if(debug) fprintf(stderr, "apps.plugin: FIXME: unknown anonymous inode: %s\n", name);
979 type = FILETYPE_OTHER;
982 if(debug) fprintf(stderr, "apps.plugin: FIXME: cannot understand linkname: %s\n", name);
983 type = FILETYPE_OTHER;
986 all_files[c].name = strdup(name);
987 all_files[c].hash = hash;
988 all_files[c].type = type;
989 all_files[c].pos = c;
990 all_files[c].count = 1;
991 all_files[c].magic = 0x0BADCAFE;
993 file_descriptor_add(&all_files[c]);
995 if(debug) fprintf(stderr, "apps.plugin: using fd position %d (name: %s)\n", c, all_files[c].name);
1001 // 1. read all files in /proc
1002 // 2. for each numeric directory:
1003 // i. read /proc/pid/stat
1004 // ii. read /proc/pid/statm
1005 // iii. read /proc/pid/io (requires root access)
1006 // iii. read the entries in directory /proc/pid/fd (requires root access)
1008 // a. find or create a struct file_descriptor
1009 // b. cleanup any old/unused file_descriptors
1011 // after all these, some pids may be linked to targets, while others may not
1013 // in case of errors, only 1 every 1000 errors is printed
1014 // to avoid filling up all disk space
1015 // if debug is enabled, all errors are printed
1017 int update_from_proc(void)
1019 static long count_errors = 0;
1021 char filename[FILENAME_MAX+1];
1022 DIR *dir = opendir("/proc");
1025 struct dirent *file = NULL;
1026 struct pid_stat *p = NULL;
1028 // mark them all as un-updated
1030 for(p = root_of_pids; p ; p = p->next) {
1039 while((file = readdir(dir))) {
1040 char *endptr = file->d_name;
1041 pid_t pid = strtoul(file->d_name, &endptr, 10);
1042 if(pid <= 0 || pid > pid_max || endptr == file->d_name || *endptr != '\0') continue;
1044 p = get_pid_entry(pid);
1047 // --------------------------------------------------------------------
1050 if(read_proc_pid_stat(p)) {
1051 if(!count_errors++ || debug || (p->target && p->target->debug))
1052 error("Cannot process /proc/%d/stat", pid);
1056 if(p->ppid < 0 || p->ppid > pid_max) p->ppid = 0;
1059 // --------------------------------------------------------------------
1060 // /proc/<pid>/statm
1062 if(read_proc_pid_statm(p)) {
1063 if(!count_errors++ || debug || (p->target && p->target->debug))
1064 error("Cannot process /proc/%d/statm", pid);
1070 // --------------------------------------------------------------------
1073 if(read_proc_pid_io(p)) {
1074 if(!count_errors++ || debug || (p->target && p->target->debug))
1075 error("Cannot process /proc/%d/io", pid);
1080 // --------------------------------------------------------------------
1083 // check if it is target
1084 // we do this only once, the first time this pid is loaded
1086 if(debug) fprintf(stderr, "apps.plugin: \tJust added %s\n", p->comm);
1089 for(w = target_root; w ; w = w->next) {
1090 // if(debug || (p->target && p->target->debug)) fprintf(stderr, "apps.plugin: \t\tcomparing '%s' with '%s'\n", w->compare, p->comm);
1092 if(strcmp(w->compare, p->comm) == 0) {
1093 if(w->target) p->target = w->target;
1096 if(debug || (p->target && p->target->debug)) fprintf(stderr, "apps.plugin: \t\t%s linked to target %s\n", p->comm, p->target->name);
1101 // --------------------------------------------------------------------
1104 snprintf(filename, FILENAME_MAX, "/proc/%s/fd", file->d_name);
1105 DIR *fds = opendir(filename);
1109 char fdname[FILENAME_MAX + 1];
1110 char linkname[FILENAME_MAX + 1];
1112 // make the array negative
1113 for(c = 0 ; c < p->fds_size ; c++) p->fds[c] = -p->fds[c];
1115 while((de = readdir(fds))) {
1116 if(strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) continue;
1118 // check if the fds array is small
1119 int fdid = atol(de->d_name);
1120 if(fdid < 0) continue;
1121 if(fdid >= p->fds_size) {
1122 // it is small, extend it
1123 if(debug) fprintf(stderr, "apps.plugin: extending fd memory slots for %s from %d to %d\n", p->comm, p->fds_size, fdid + 100);
1124 p->fds = realloc(p->fds, (fdid + 100) * sizeof(int));
1126 error("Cannot re-allocate fds for %s", p->comm);
1130 // and initialize it
1131 for(c = p->fds_size ; c < (fdid + 100) ; c++) p->fds[c] = 0;
1132 p->fds_size = fdid + 100;
1135 if(p->fds[fdid] == 0) {
1136 // we don't know this fd, get it
1138 sprintf(fdname, "/proc/%s/fd/%s", file->d_name, de->d_name);
1139 int l = readlink(fdname, linkname, FILENAME_MAX);
1141 if(debug || (p->target && p->target->debug)) {
1142 if(!count_errors++ || debug || (p->target && p->target->debug))
1143 error("Cannot read link %s", fdname);
1150 // if another process already has this, we will get
1152 p->fds[fdid] = file_descriptor_find_or_add(linkname);
1155 // else make it positive again, we need it
1156 // of course, the actual file may have changed, but we don't care so much
1157 // FIXME: we could compare the inode as returned by readdir direct structure
1158 else p->fds[fdid] = -p->fds[fdid];
1162 // remove all the negative file descriptors
1163 for(c = 0 ; c < p->fds_size ; c++) if(p->fds[c] < 0) {
1164 file_descriptor_not_used(-p->fds[c]);
1169 // --------------------------------------------------------------------
1172 // mark it as updated
1175 if(count_errors > 1000) {
1176 error("%ld more errors encountered\n", count_errors - 1);
1186 // ----------------------------------------------------------------------------
1187 // update statistics on the targets
1189 // 1. link all childs to their parents
1190 // 2. go from bottom to top, marking as merged all childs to their parents
1191 // this step links all parents without a target to the child target, if any
1192 // 3. link all top level processes (the ones not merged) to the default target
1193 // 4. go from top to bottom, linking all childs without a target, to their parent target
1194 // after this step, all processes have a target
1195 // [5. for each killed pid (updated = 0), remove its usage from its target]
1196 // 6. zero all targets
1197 // 7. concentrate all values on the targets
1198 // 8. remove all killed processes
1199 // 9. find the unique file count for each target
1201 void update_statistics(void)
1204 struct pid_stat *p = NULL;
1206 // link all parents and update childs count
1207 for(p = root_of_pids; p ; p = p->next) {
1208 if(p->ppid > 0 && p->ppid <= pid_max && all_pids[p->ppid]) {
1209 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);
1211 p->parent = all_pids[p->ppid];
1212 p->parent->childs++;
1214 else if(p->ppid != 0) error("pid %d %s states parent %d, but the later does not exist.", p->pid, p->comm, p->ppid);
1217 // find all the procs with 0 childs and merge them to their parents
1218 // repeat, until nothing more can be done.
1222 for(p = root_of_pids; p ; p = p->next) {
1223 // if this process does not have any childs, and
1224 // is not already merged, and
1225 // its parent has childs waiting to be merged, and
1226 // the target of this process and its parent is the same, or the parent does not have a target, or this process does not have a parent
1227 // and its parent is not init
1228 // then... merge them!
1229 if(!p->childs && !p->merged && p->parent && p->parent->childs && (p->target == p->parent->target || !p->parent->target || !p->target) && p->ppid != 1) {
1230 p->parent->childs--;
1233 // the parent inherits the child's target, if it does not have a target itself
1234 if(p->target && !p->parent->target) {
1235 p->parent->target = p->target;
1236 if(debug || (p->target && p->target->debug)) fprintf(stderr, "apps.plugin: \t\ttarget %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);
1242 if(debug) fprintf(stderr, "apps.plugin: merged %d processes\n", found);
1245 // give a default target on all top level processes
1246 // init goes always to default target
1247 if(all_pids[1]) all_pids[1]->target = default_target;
1249 for(p = root_of_pids; p ; p = p->next) {
1250 // if the process is not merged itself
1251 // then is is a top level process
1252 if(!p->merged && !p->target) p->target = default_target;
1254 #ifdef INCLUDE_CHILDS
1255 // by the way, update the diffs
1256 // will be used later for substracting killed process times
1257 p->diff_cutime = p->utime - p->cutime;
1258 p->diff_cstime = p->stime - p->cstime;
1259 p->diff_cminflt = p->minflt - p->cminflt;
1260 p->diff_cmajflt = p->majflt - p->cmajflt;
1264 // give a target to all merged child processes
1268 for(p = root_of_pids; p ; p = p->next) {
1269 if(!p->target && p->merged && p->parent && p->parent->target) {
1270 p->target = p->parent->target;
1276 #ifdef INCLUDE_CHILDS
1277 // for each killed process, remove its values from the parents
1278 // sums (we had already added them in a previous loop)
1279 for(p = root_of_pids; p ; p = p->next) {
1280 if(p->updated) continue;
1282 if(debug) fprintf(stderr, "apps.plugin: UNMERGING %d %s\n", p->pid, p->comm);
1284 unsigned long long diff_utime = p->utime + p->cutime + p->fix_cutime;
1285 unsigned long long diff_stime = p->stime + p->cstime + p->fix_cstime;
1286 unsigned long long diff_minflt = p->minflt + p->cminflt + p->fix_cminflt;
1287 unsigned long long diff_majflt = p->majflt + p->cmajflt + p->fix_cmajflt;
1289 struct pid_stat *t = p;
1290 while((t = t->parent)) {
1291 if(!t->updated) continue;
1293 unsigned long long x;
1294 if(diff_utime && t->diff_cutime) {
1295 x = (t->diff_cutime < diff_utime)?t->diff_cutime:diff_utime;
1297 t->diff_cutime -= x;
1299 if(debug) fprintf(stderr, "apps.plugin: \t cutime %llu from %d %s %s\n", x, t->pid, t->comm, t->target->name);
1301 if(diff_stime && t->diff_cstime) {
1302 x = (t->diff_cstime < diff_stime)?t->diff_cstime:diff_stime;
1304 t->diff_cstime -= x;
1306 if(debug) fprintf(stderr, "apps.plugin: \t cstime %llu from %d %s %s\n", x, t->pid, t->comm, t->target->name);
1308 if(diff_minflt && t->diff_cminflt) {
1309 x = (t->diff_cminflt < diff_minflt)?t->diff_cminflt:diff_minflt;
1311 t->diff_cminflt -= x;
1312 t->fix_cminflt += x;
1313 if(debug) fprintf(stderr, "apps.plugin: \t cminflt %llu from %d %s %s\n", x, t->pid, t->comm, t->target->name);
1315 if(diff_majflt && t->diff_cmajflt) {
1316 x = (t->diff_cmajflt < diff_majflt)?t->diff_cmajflt:diff_majflt;
1318 t->diff_cmajflt -= x;
1319 t->fix_cmajflt += x;
1320 if(debug) fprintf(stderr, "apps.plugin: \t cmajflt %llu from %d %s %s\n", x, t->pid, t->comm, t->target->name);
1324 if(diff_utime) error("Cannot fix up utime %llu", diff_utime);
1325 if(diff_stime) error("Cannot fix up stime %llu", diff_stime);
1326 if(diff_minflt) error("Cannot fix up minflt %llu", diff_minflt);
1327 if(diff_majflt) error("Cannot fix up majflt %llu", diff_majflt);
1331 // zero all the targets
1334 for (w = target_root; w ; w = w->next) {
1337 w->fds = calloc(sizeof(int), all_files_size);
1339 error("Cannot allocate memory for fds in %s", w->name);
1354 w->statm_resident = 0;
1361 w->io_logical_bytes_read = 0;
1362 w->io_logical_bytes_written = 0;
1363 w->io_read_calls = 0;
1364 w->io_write_calls = 0;
1365 w->io_storage_bytes_read = 0;
1366 w->io_storage_bytes_written = 0;
1367 w->io_cancelled_write_bytes = 0;
1370 #ifdef INCLUDE_CHILDS
1371 if(debug) walk_down(0, 1);
1374 // concentrate everything on the targets
1375 for(p = root_of_pids; p ; p = p->next) {
1377 error("pid %d %s was left without a target!", p->pid, p->comm);
1382 p->target->cutime += p->cutime; // - p->fix_cutime;
1383 p->target->cstime += p->cstime; // - p->fix_cstime;
1384 p->target->cminflt += p->cminflt; // - p->fix_cminflt;
1385 p->target->cmajflt += p->cmajflt; // - p->fix_cmajflt;
1387 p->target->utime += p->utime; //+ (p->pid != 1)?(p->cutime - p->fix_cutime):0;
1388 p->target->stime += p->stime; //+ (p->pid != 1)?(p->cstime - p->fix_cstime):0;
1389 p->target->minflt += p->minflt; //+ (p->pid != 1)?(p->cminflt - p->fix_cminflt):0;
1390 p->target->majflt += p->majflt; //+ (p->pid != 1)?(p->cmajflt - p->fix_cmajflt):0;
1392 //if(p->num_threads < 0)
1393 // error("Negative threads number for pid '%s' (%d): %d", p->comm, p->pid, p->num_threads);
1395 //if(p->num_threads > 10000)
1396 // error("Excessive threads number for pid '%s' (%d): %d", p->comm, p->pid, p->num_threads);
1398 p->target->num_threads += p->num_threads;
1399 p->target->rss += p->rss;
1401 p->target->statm_size += p->statm_size;
1402 p->target->statm_resident += p->statm_resident;
1403 p->target->statm_share += p->statm_share;
1404 p->target->statm_text += p->statm_text;
1405 p->target->statm_lib += p->statm_lib;
1406 p->target->statm_data += p->statm_data;
1407 p->target->statm_dirty += p->statm_dirty;
1409 p->target->io_logical_bytes_read += p->io_logical_bytes_read;
1410 p->target->io_logical_bytes_written += p->io_logical_bytes_written;
1411 p->target->io_read_calls += p->io_read_calls;
1412 p->target->io_write_calls += p->io_write_calls;
1413 p->target->io_storage_bytes_read += p->io_storage_bytes_read;
1414 p->target->io_storage_bytes_written += p->io_storage_bytes_written;
1415 p->target->io_cancelled_write_bytes += p->io_cancelled_write_bytes;
1417 p->target->processes++;
1419 for(c = 0; c < p->fds_size ;c++) {
1420 if(p->fds[c] == 0) continue;
1421 if(p->fds[c] < all_files_size) {
1422 if(p->target->fds) p->target->fds[p->fds[c]]++;
1425 error("Invalid fd number %d", p->fds[c]);
1428 if(debug || p->target->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, p->target->name, p->utime, p->stime, p->cutime, p->cstime, p->minflt, p->majflt, p->cminflt, p->cmajflt);
1430 /* 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);
1431 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);
1432 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);
1433 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);
1434 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);
1435 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);
1436 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);
1437 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);
1439 #ifdef INCLUDE_CHILDS
1440 p->old_utime = p->utime;
1441 p->old_cutime = p->cutime;
1442 p->old_stime = p->stime;
1443 p->old_cstime = p->cstime;
1444 p->old_minflt = p->minflt;
1445 p->old_majflt = p->majflt;
1446 p->old_cminflt = p->cminflt;
1447 p->old_cmajflt = p->cmajflt;
1451 // since the process has exited, the user
1452 // will see a drop in our charts, because the incremental
1453 // values of this process will not be there
1455 // add them to the fix_* values and they will be added to
1456 // the reported values, so that the report goes steady
1457 p->target->fix_minflt += p->minflt;
1458 p->target->fix_majflt += p->majflt;
1459 p->target->fix_utime += p->utime;
1460 p->target->fix_stime += p->stime;
1461 p->target->fix_cminflt += p->cminflt;
1462 p->target->fix_cmajflt += p->cmajflt;
1463 p->target->fix_cutime += p->cutime;
1464 p->target->fix_cstime += p->cstime;
1466 p->target->fix_io_logical_bytes_read += p->io_logical_bytes_read;
1467 p->target->fix_io_logical_bytes_written += p->io_logical_bytes_written;
1468 p->target->fix_io_read_calls += p->io_read_calls;
1469 p->target->fix_io_write_calls += p->io_write_calls;
1470 p->target->fix_io_storage_bytes_read += p->io_storage_bytes_read;
1471 p->target->fix_io_storage_bytes_written += p->io_storage_bytes_written;
1472 p->target->fix_io_cancelled_write_bytes += p->io_cancelled_write_bytes;
1476 // fprintf(stderr, "\n");
1477 // cleanup all un-updated processed (exited, killed, etc)
1478 for(p = root_of_pids; p ;) {
1480 // 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);
1482 for(c = 0 ; c < p->fds_size ; c++) if(p->fds[c] > 0) {
1483 file_descriptor_not_used(p->fds[c]);
1494 for (w = target_root; w ; w = w->next) {
1498 w->openinotifies = 0;
1499 w->openeventfds = 0;
1500 w->opentimerfds = 0;
1501 w->opensignalfds = 0;
1502 w->openeventpolls = 0;
1505 for(c = 1; c < all_files_size ;c++) {
1506 if(w->fds && w->fds[c] > 0) switch(all_files[c].type) {
1515 case FILETYPE_SOCKET:
1519 case FILETYPE_INOTIFY:
1523 case FILETYPE_EVENTFD:
1527 case FILETYPE_TIMERFD:
1531 case FILETYPE_SIGNALFD:
1535 case FILETYPE_EVENTPOLL:
1536 w->openeventpolls++;
1549 // ----------------------------------------------------------------------------
1550 // update chart dimensions
1552 void show_dimensions(void)
1554 static struct timeval last = { 0, 0 };
1555 static struct rusage me_last;
1561 unsigned long long usec;
1562 unsigned long long cpuuser;
1563 unsigned long long cpusyst;
1566 gettimeofday(&last, NULL);
1567 getrusage(RUSAGE_SELF, &me_last);
1569 // the first time, give a zero to allow
1570 // netdata calibrate to the current time
1571 // usec = update_every * 1000000ULL;
1577 gettimeofday(&now, NULL);
1578 getrusage(RUSAGE_SELF, &me);
1580 usec = usecdiff(&now, &last);
1581 cpuuser = me.ru_utime.tv_sec * 1000000ULL + me.ru_utime.tv_usec;
1582 cpusyst = me.ru_stime.tv_sec * 1000000ULL + me.ru_stime.tv_usec;
1584 bcopy(&now, &last, sizeof(struct timeval));
1585 bcopy(&me, &me_last, sizeof(struct rusage));
1588 fprintf(stdout, "BEGIN apps.cpu %llu\n", usec);
1589 for (w = target_root; w ; w = w->next) {
1590 if(w->target || (!w->processes && !w->exposed)) continue;
1592 fprintf(stdout, "SET %s = %llu\n", w->name, w->utime + w->stime + w->fix_utime + w->fix_stime);
1594 fprintf(stdout, "END\n");
1596 fprintf(stdout, "BEGIN apps.cpu_user %llu\n", usec);
1597 for (w = target_root; w ; w = w->next) {
1598 if(w->target || (!w->processes && !w->exposed)) continue;
1600 fprintf(stdout, "SET %s = %llu\n", w->name, w->utime + w->fix_utime);
1602 fprintf(stdout, "END\n");
1604 fprintf(stdout, "BEGIN apps.cpu_system %llu\n", usec);
1605 for (w = target_root; w ; w = w->next) {
1606 if(w->target || (!w->processes && !w->exposed)) continue;
1608 fprintf(stdout, "SET %s = %llu\n", w->name, w->stime + w->fix_stime);
1610 fprintf(stdout, "END\n");
1612 fprintf(stdout, "BEGIN apps.threads %llu\n", usec);
1613 for (w = target_root; w ; w = w->next) {
1614 if(w->target || (!w->processes && !w->exposed)) continue;
1616 fprintf(stdout, "SET %s = %llu\n", w->name, w->num_threads);
1618 fprintf(stdout, "END\n");
1620 fprintf(stdout, "BEGIN apps.processes %llu\n", usec);
1621 for (w = target_root; w ; w = w->next) {
1622 if(w->target || (!w->processes && !w->exposed)) continue;
1624 fprintf(stdout, "SET %s = %lu\n", w->name, w->processes);
1626 fprintf(stdout, "END\n");
1628 fprintf(stdout, "BEGIN apps.mem %llu\n", usec);
1629 for (w = target_root; w ; w = w->next) {
1630 if(w->target || (!w->processes && !w->exposed)) continue;
1632 fprintf(stdout, "SET %s = %lld\n", w->name, (long long)w->statm_resident - (long long)w->statm_share);
1634 fprintf(stdout, "END\n");
1636 fprintf(stdout, "BEGIN apps.minor_faults %llu\n", usec);
1637 for (w = target_root; w ; w = w->next) {
1638 if(w->target || (!w->processes && !w->exposed)) continue;
1640 fprintf(stdout, "SET %s = %llu\n", w->name, w->minflt + w->fix_minflt);
1642 fprintf(stdout, "END\n");
1644 fprintf(stdout, "BEGIN apps.major_faults %llu\n", usec);
1645 for (w = target_root; w ; w = w->next) {
1646 if(w->target || (!w->processes && !w->exposed)) continue;
1648 fprintf(stdout, "SET %s = %llu\n", w->name, w->majflt + w->fix_majflt);
1650 fprintf(stdout, "END\n");
1652 fprintf(stdout, "BEGIN apps.lreads %llu\n", usec);
1653 for (w = target_root; w ; w = w->next) {
1654 if(w->target || (!w->processes && !w->exposed)) continue;
1656 fprintf(stdout, "SET %s = %llu\n", w->name, w->io_logical_bytes_read);
1658 fprintf(stdout, "END\n");
1660 fprintf(stdout, "BEGIN apps.lwrites %llu\n", usec);
1661 for (w = target_root; w ; w = w->next) {
1662 if(w->target || (!w->processes && !w->exposed)) continue;
1664 fprintf(stdout, "SET %s = %llu\n", w->name, w->io_logical_bytes_written);
1666 fprintf(stdout, "END\n");
1668 fprintf(stdout, "BEGIN apps.preads %llu\n", usec);
1669 for (w = target_root; w ; w = w->next) {
1670 if(w->target || (!w->processes && !w->exposed)) continue;
1672 fprintf(stdout, "SET %s = %llu\n", w->name, w->io_storage_bytes_read);
1674 fprintf(stdout, "END\n");
1676 fprintf(stdout, "BEGIN apps.pwrites %llu\n", usec);
1677 for (w = target_root; w ; w = w->next) {
1678 if(w->target || (!w->processes && !w->exposed)) continue;
1680 fprintf(stdout, "SET %s = %llu\n", w->name, w->io_storage_bytes_written);
1682 fprintf(stdout, "END\n");
1684 fprintf(stdout, "BEGIN apps.files %llu\n", usec);
1685 for (w = target_root; w ; w = w->next) {
1686 if(w->target || (!w->processes && !w->exposed)) continue;
1688 fprintf(stdout, "SET %s = %llu\n", w->name, w->openfiles);
1690 fprintf(stdout, "END\n");
1692 fprintf(stdout, "BEGIN apps.sockets %llu\n", usec);
1693 for (w = target_root; w ; w = w->next) {
1694 if(w->target || (!w->processes && !w->exposed)) continue;
1696 fprintf(stdout, "SET %s = %llu\n", w->name, w->opensockets);
1698 fprintf(stdout, "END\n");
1700 fprintf(stdout, "BEGIN apps.pipes %llu\n", usec);
1701 for (w = target_root; w ; w = w->next) {
1702 if(w->target || (!w->processes && !w->exposed)) continue;
1704 fprintf(stdout, "SET %s = %llu\n", w->name, w->openpipes);
1706 fprintf(stdout, "END\n");
1708 fprintf(stdout, "BEGIN netdata.apps_cpu %llu\n", usec);
1709 fprintf(stdout, "SET user = %llu\n", cpuuser);
1710 fprintf(stdout, "SET system = %llu\n", cpusyst);
1711 fprintf(stdout, "END\n");
1713 fprintf(stdout, "BEGIN netdata.apps_files %llu\n", usec);
1714 fprintf(stdout, "SET files = %llu\n", file_counter);
1715 fprintf(stdout, "SET pids = %ld\n", all_pids_count);
1716 fprintf(stdout, "SET fds = %d\n", all_files_len);
1717 fprintf(stdout, "SET targets = %ld\n", targets);
1718 fprintf(stdout, "END\n");
1724 // ----------------------------------------------------------------------------
1725 // generate the charts
1727 void show_charts(void)
1730 int newly_added = 0;
1732 for(w = target_root ; w ; w = w->next)
1733 if(!w->exposed && w->processes) {
1736 if(debug || w->debug) fprintf(stderr, "apps.plugin: %s just added - regenerating charts.\n", w->name);
1739 // nothing more to show
1740 if(!newly_added) return;
1742 // we have something new to show
1743 // update the charts
1744 fprintf(stdout, "CHART apps.cpu '' 'Apps CPU Time (%ld%% = %ld core%s)' 'cpu time %%' apps apps stacked 20001 %d\n", (processors * 100), processors, (processors>1)?"s":"", update_every);
1745 for (w = target_root; w ; w = w->next) {
1746 if(w->target || (!w->processes && !w->exposed)) continue;
1748 fprintf(stdout, "DIMENSION %s '' incremental 100 %llu %s\n", w->name, (unsigned long long)(Hertz * update_every), w->hidden?"hidden":"");
1751 fprintf(stdout, "CHART apps.mem '' 'Apps Dedicated Memory (w/o shared)' 'MB' apps apps stacked 20003 %d\n", update_every);
1752 for (w = target_root; w ; w = w->next) {
1753 if(w->target || (!w->processes && !w->exposed)) continue;
1755 fprintf(stdout, "DIMENSION %s '' absolute %ld %ld\n", w->name, sysconf(_SC_PAGESIZE), 1024L*1024L);
1758 fprintf(stdout, "CHART apps.threads '' 'Apps Threads' 'threads' apps apps stacked 20005 %d\n", update_every);
1759 for (w = target_root; w ; w = w->next) {
1760 if(w->target || (!w->processes && !w->exposed)) continue;
1762 fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name);
1765 fprintf(stdout, "CHART apps.processes '' 'Apps Processes' 'processes' apps apps stacked 20004 %d\n", update_every);
1766 for (w = target_root; w ; w = w->next) {
1767 if(w->target || (!w->processes && !w->exposed)) continue;
1769 fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name);
1772 fprintf(stdout, "CHART apps.cpu_user '' 'Apps CPU User Time (%ld%% = %ld core%s)' 'cpu time %%' apps none stacked 20020 %d\n", (processors * 100), processors, (processors>1)?"s":"", update_every);
1773 for (w = target_root; w ; w = w->next) {
1774 if(w->target || (!w->processes && !w->exposed)) continue;
1776 fprintf(stdout, "DIMENSION %s '' incremental 100 %llu\n", w->name, Hertz * processors * update_every);
1779 fprintf(stdout, "CHART apps.cpu_system '' 'Apps CPU System Time (%ld%% = %ld core%s)' 'cpu time %%' apps none stacked 20021 %d\n", (processors * 100), processors, (processors>1)?"s":"", update_every);
1780 for (w = target_root; w ; w = w->next) {
1781 if(w->target || (!w->processes && !w->exposed)) continue;
1783 fprintf(stdout, "DIMENSION %s '' incremental 100 %llu\n", w->name, Hertz * processors * update_every);
1786 fprintf(stdout, "CHART apps.major_faults '' 'Apps Major Page Faults (swaps in)' 'page faults/s' apps apps stacked 20010 %d\n", update_every);
1787 for (w = target_root; w ; w = w->next) {
1788 if(w->target || (!w->processes && !w->exposed)) continue;
1790 fprintf(stdout, "DIMENSION %s '' incremental 1 %d\n", w->name, update_every);
1793 fprintf(stdout, "CHART apps.minor_faults '' 'Apps Minor Page Faults' 'page faults/s' apps none stacked 20011 %d\n", update_every);
1794 for (w = target_root; w ; w = w->next) {
1795 if(w->target || (!w->processes && !w->exposed)) continue;
1797 fprintf(stdout, "DIMENSION %s '' incremental 1 %d\n", w->name, update_every);
1800 fprintf(stdout, "CHART apps.lreads '' 'Apps Disk Logical Reads' 'kilobytes/s' apps none stacked 20042 %d\n", update_every);
1801 for (w = target_root; w ; w = w->next) {
1802 if(w->target || (!w->processes && !w->exposed)) continue;
1804 fprintf(stdout, "DIMENSION %s '' incremental 1 %d\n", w->name, 1024 * update_every);
1807 fprintf(stdout, "CHART apps.lwrites '' 'Apps I/O Logical Writes' 'kilobytes/s' apps none stacked 20042 %d\n", update_every);
1808 for (w = target_root; w ; w = w->next) {
1809 if(w->target || (!w->processes && !w->exposed)) continue;
1811 fprintf(stdout, "DIMENSION %s '' incremental 1 %d\n", w->name, 1024 * update_every);
1814 fprintf(stdout, "CHART apps.preads '' 'Apps Disk Reads' 'kilobytes/s' apps apps stacked 20002 %d\n", update_every);
1815 for (w = target_root; w ; w = w->next) {
1816 if(w->target || (!w->processes && !w->exposed)) continue;
1818 fprintf(stdout, "DIMENSION %s '' incremental 1 %d\n", w->name, 1024 * update_every);
1821 fprintf(stdout, "CHART apps.pwrites '' 'Apps Disk Writes' 'kilobytes/s' apps apps stacked 20002 %d\n", update_every);
1822 for (w = target_root; w ; w = w->next) {
1823 if(w->target || (!w->processes && !w->exposed)) continue;
1825 fprintf(stdout, "DIMENSION %s '' incremental 1 %d\n", w->name, 1024 * update_every);
1828 fprintf(stdout, "CHART apps.files '' 'Apps Open Files' 'open files' apps apps stacked 20050 %d\n", update_every);
1829 for (w = target_root; w ; w = w->next) {
1830 if(w->target || (!w->processes && !w->exposed)) continue;
1832 fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name);
1835 fprintf(stdout, "CHART apps.sockets '' 'Apps Open Sockets' 'open sockets' apps apps stacked 20051 %d\n", update_every);
1836 for (w = target_root; w ; w = w->next) {
1837 if(w->target || (!w->processes && !w->exposed)) continue;
1839 fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name);
1842 fprintf(stdout, "CHART apps.pipes '' 'Apps Pipes' 'open pipes' apps none stacked 20053 %d\n", update_every);
1843 for (w = target_root; w ; w = w->next) {
1844 if(w->target || (!w->processes && !w->exposed)) continue;
1846 fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name);
1849 fprintf(stdout, "CHART netdata.apps_cpu '' 'Apps Plugin CPU' 'milliseconds/s' netdata netdata stacked 10000 %d\n", update_every);
1850 fprintf(stdout, "DIMENSION user '' incremental 1 %d\n", 1000 * update_every);
1851 fprintf(stdout, "DIMENSION system '' incremental 1 %d\n", 1000 * update_every);
1853 fprintf(stdout, "CHART netdata.apps_files '' 'Apps Plugin Files' 'files/s' netdata netdata line 10001 %d\n", update_every);
1854 fprintf(stdout, "DIMENSION files '' incremental 1 %d\n", update_every);
1855 fprintf(stdout, "DIMENSION pids '' absolute 1 1\n");
1856 fprintf(stdout, "DIMENSION fds '' absolute 1 1\n");
1857 fprintf(stdout, "DIMENSION targets '' absolute 1 1\n");
1863 // ----------------------------------------------------------------------------
1864 // parse command line arguments
1866 void parse_args(int argc, char **argv)
1871 for(i = 1; i < argc; i++) {
1873 int n = atoi(argv[i]);
1880 if(strcmp("debug", argv[i]) == 0) {
1882 debug_flags = 0xffffffff;
1891 error("Cannot understand option %s", argv[i]);
1895 if(freq > 0) update_every = freq;
1896 if(!name) name = "groups";
1898 if(read_process_groups(name)) {
1899 error("Cannot read process groups %s", name);
1904 int main(int argc, char **argv)
1906 // debug_flags = D_PROCFILE;
1908 // set the name for logging
1909 program_name = "apps.plugin";
1911 info("starting...");
1913 procfile_adaptive_initial_allocation = 1;
1915 unsigned long started_t = time(NULL), current_t;
1916 Hertz = get_hertz();
1917 pid_max = get_pid_max();
1918 processors = get_processors();
1920 parse_args(argc, argv);
1922 all_pids = calloc(sizeof(struct pid_stat *), pid_max);
1924 error("Cannot allocate %lu bytes of memory.", sizeof(struct pid_stat *) * pid_max);
1925 printf("DISABLE\n");
1929 unsigned long long counter = 1;
1930 unsigned long long usec = 0, susec = 0;
1931 struct timeval last, now;
1932 gettimeofday(&last, NULL);
1934 for(;1; counter++) {
1935 if(!update_from_proc()) {
1936 error("Cannot allocate %lu bytes of memory.", sizeof(struct pid_stat *) * pid_max);
1937 printf("DISABLE\n");
1941 update_statistics();
1942 show_charts(); // this is smart enough to show only newly added apps, when needed
1945 if(debug) fprintf(stderr, "apps.plugin: done Loop No %llu\n", counter);
1948 gettimeofday(&now, NULL);
1949 usec = usecdiff(&now, &last) - susec;
1950 if(debug) fprintf(stderr, "apps.plugin: last loop took %llu usec (worked for %llu, sleeped for %llu).\n", usec + susec, usec, susec);
1952 // if the last loop took less than half the time
1953 // wait the rest of the time
1954 if(usec < (update_every * 1000000ULL / 2)) susec = (update_every * 1000000ULL) - usec;
1955 else susec = update_every * 1000000ULL / 2;
1958 bcopy(&now, &last, sizeof(struct timeval));
1960 current_t = time(NULL);
1961 if(current_t - started_t > 3600) exit(0);