3 // 1. support -ch option to set global /proc and /sys prefix
4 // 2. disable RESET_OR_OVERFLOW check in charts
14 #include <sys/types.h>
18 #include <sys/resource.h>
29 #include <arpa/inet.h>
36 #define MAX_COMPARE_NAME 15
39 unsigned long long Hertz = 1;
46 unsigned long long file_counter = 0;
48 #define PROC_BUFFER 4096
50 // ----------------------------------------------------------------------------
57 } allocations = { 0, 0, 0 };
59 #define MALLOC_MARK (uint32_t)(0x0BADCAFE)
60 #define MALLOC_PREFIX (sizeof(uint32_t) * 2)
61 #define MALLOC_SUFFIX (sizeof(uint32_t))
62 #define MALLOC_OVERHEAD (MALLOC_PREFIX + MALLOC_SUFFIX)
64 void *mark_allocation(void *allocated_ptr, size_t size_without_overheads) {
65 uint32_t *real_ptr = (uint32_t *)allocated_ptr;
66 real_ptr[0] = MALLOC_MARK;
67 real_ptr[1] = size_without_overheads;
69 uint32_t *end_ptr = (uint32_t *)(allocated_ptr + MALLOC_PREFIX + size_without_overheads);
70 end_ptr[0] = MALLOC_MARK;
72 // fprintf(stderr, "MEMORY_POINTER: Allocated at %p, returning %p.\n", allocated_ptr, (void *)(allocated_ptr + MALLOC_PREFIX));
74 return (void *)(allocated_ptr + MALLOC_PREFIX);
77 void *check_allocation(const char *file, int line, const char *function, void *marked_ptr, size_t *size_without_overheads_ptr) {
78 uint32_t *real_ptr = (uint32_t *)(marked_ptr - MALLOC_PREFIX);
80 // 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);
82 if(real_ptr[0] != MALLOC_MARK) fatal("MEMORY: prefix MARK is not valid for %s/%u@%s.", function, line, file);
84 size_t size = real_ptr[1];
86 uint32_t *end_ptr = (uint32_t *)(marked_ptr + size);
87 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);
89 if(size_without_overheads_ptr) *size_without_overheads_ptr = size;
94 void *malloc_debug(const char *file, int line, const char *function, size_t size) {
95 void *ptr = malloc(size + MALLOC_OVERHEAD);
96 if(!ptr) fatal("MEMORY: Cannot allocate %zu bytes for %s/%u@%s.", size, function, line, file);
98 allocations.allocated += size;
99 allocations.allocations++;
101 debug(D_MEMORY, "MEMORY: Allocated %zu bytes for %s/%u@%s."
102 " Status: allocated %zu in %zu allocs."
104 , function, line, file
105 , allocations.allocated
106 , allocations.allocations
109 if(allocations.allocated > allocations.allocated_max) {
110 debug(D_MEMORY, "MEMORY: total allocation peak increased from %zu to %zu", allocations.allocated_max, allocations.allocated);
111 allocations.allocated_max = allocations.allocated;
115 check_allocation(file, line, function, mark_allocation(ptr, size), &csize);
117 fatal("Invalid size.");
120 return mark_allocation(ptr, size);
123 void *calloc_debug(const char *file, int line, const char *function, size_t nmemb, size_t size) {
124 void *ptr = malloc_debug(file, line, function, (nmemb * size));
125 bzero(ptr, nmemb * size);
129 void free_debug(const char *file, int line, const char *function, void *ptr) {
131 void *real_ptr = check_allocation(file, line, function, ptr, &size);
133 bzero(real_ptr, size + MALLOC_OVERHEAD);
136 allocations.allocated -= size;
137 allocations.allocations--;
139 debug(D_MEMORY, "MEMORY: freed %zu bytes for %s/%u@%s."
140 " Status: allocated %zu in %zu allocs."
142 , function, line, file
143 , allocations.allocated
144 , allocations.allocations
148 void *realloc_debug(const char *file, int line, const char *function, void *ptr, size_t size) {
149 if(!ptr) return malloc_debug(file, line, function, size);
150 if(!size) { free_debug(file, line, function, ptr); return NULL; }
153 void *real_ptr = check_allocation(file, line, function, ptr, &old_size);
155 void *new_ptr = realloc(real_ptr, size + MALLOC_OVERHEAD);
156 if(!new_ptr) fatal("MEMORY: Cannot allocate %zu bytes for %s/%u@%s.", size, function, line, file);
158 allocations.allocated += size;
159 allocations.allocated -= old_size;
161 debug(D_MEMORY, "MEMORY: Re-allocated from %zu to %zu bytes for %s/%u@%s."
162 " Status: allocated %z in %zu allocs."
164 , function, line, file
165 , allocations.allocated
166 , allocations.allocations
169 if(allocations.allocated > allocations.allocated_max) {
170 debug(D_MEMORY, "MEMORY: total allocation peak increased from %zu to %zu", allocations.allocated_max, allocations.allocated);
171 allocations.allocated_max = allocations.allocated;
174 return mark_allocation(new_ptr, size);
177 char *strdup_debug(const char *file, int line, const char *function, const char *ptr) {
184 char *p = malloc_debug(file, line, function, size);
185 if(!p) fatal("Cannot allocate %zu bytes.", size);
187 memcpy(p, ptr, size);
191 #define malloc(size) malloc_debug(__FILE__, __LINE__, __FUNCTION__, (size))
192 #define calloc(nmemb, size) calloc_debug(__FILE__, __LINE__, __FUNCTION__, (nmemb), (size))
193 #define realloc(ptr, size) realloc_debug(__FILE__, __LINE__, __FUNCTION__, (ptr), (size))
194 #define free(ptr) free_debug(__FILE__, __LINE__, __FUNCTION__, (ptr))
199 #define strdup(ptr) strdup_debug(__FILE__, __LINE__, __FUNCTION__, (ptr))
201 // ----------------------------------------------------------------------------
206 long get_processors(void) {
209 ff = procfile_reopen(ff, "/proc/stat", "", PROCFILE_FLAG_DEFAULT);
212 ff = procfile_readall(ff);
214 // procfile_close(ff);
219 for(i = 0; i < procfile_lines(ff); i++) {
220 if(!procfile_linewords(ff, i)) continue;
222 if(strncmp(procfile_lineword(ff, i, 0), "cpu", 3) == 0) processors++;
225 if(processors < 1) processors = 1;
227 // procfile_close(ff);
231 long get_pid_max(void) {
234 ff = procfile_reopen(ff, "/proc/sys/kernel/pid_max", "", PROCFILE_FLAG_DEFAULT);
237 ff = procfile_readall(ff);
239 // procfile_close(ff);
243 mpid = atol(procfile_lineword(ff, 0, 0));
244 if(!mpid) mpid = 32768;
246 // procfile_close(ff);
250 unsigned long long get_hertz(void)
252 unsigned long long hz = 1;
255 if((hz = sysconf(_SC_CLK_TCK)) > 0) {
261 hz = (unsigned long long)HZ; /* <asm/param.h> */
263 /* If 32-bit or big-endian (not Alpha or ia64), assume HZ is 100. */
264 hz = (sizeof(long)==sizeof(int) || htons(999)==999) ? 100UL : 1024UL;
267 error("Unknown HZ value. Assuming %llu.", hz);
272 // ----------------------------------------------------------------------------
274 // target is the point to aggregate a process tree values
277 char compare[MAX_COMPARE_NAME + 1];
278 char id[MAX_NAME + 1];
279 char name[MAX_NAME + 1];
281 unsigned long long minflt;
282 unsigned long long cminflt;
283 unsigned long long majflt;
284 unsigned long long cmajflt;
285 unsigned long long utime;
286 unsigned long long stime;
287 unsigned long long cutime;
288 unsigned long long cstime;
289 unsigned long long num_threads;
290 unsigned long long rss;
292 unsigned long long fix_minflt;
293 unsigned long long fix_cminflt;
294 unsigned long long fix_majflt;
295 unsigned long long fix_cmajflt;
296 unsigned long long fix_utime;
297 unsigned long long fix_stime;
298 unsigned long long fix_cutime;
299 unsigned long long fix_cstime;
301 unsigned long long statm_size;
302 unsigned long long statm_resident;
303 unsigned long long statm_share;
304 unsigned long long statm_text;
305 unsigned long long statm_lib;
306 unsigned long long statm_data;
307 unsigned long long statm_dirty;
309 unsigned long long io_logical_bytes_read;
310 unsigned long long io_logical_bytes_written;
311 unsigned long long io_read_calls;
312 unsigned long long io_write_calls;
313 unsigned long long io_storage_bytes_read;
314 unsigned long long io_storage_bytes_written;
315 unsigned long long io_cancelled_write_bytes;
317 unsigned long long fix_io_logical_bytes_read;
318 unsigned long long fix_io_logical_bytes_written;
319 unsigned long long fix_io_read_calls;
320 unsigned long long fix_io_write_calls;
321 unsigned long long fix_io_storage_bytes_read;
322 unsigned long long fix_io_storage_bytes_written;
323 unsigned long long fix_io_cancelled_write_bytes;
326 unsigned long long openfiles;
327 unsigned long long openpipes;
328 unsigned long long opensockets;
329 unsigned long long openinotifies;
330 unsigned long long openeventfds;
331 unsigned long long opentimerfds;
332 unsigned long long opensignalfds;
333 unsigned long long openeventpolls;
334 unsigned long long openother;
336 unsigned long processes; // how many processes have been merged to this
337 int exposed; // if set, we have sent this to netdata
338 int hidden; // if set, we set the hidden flag on the dimension
341 struct target *target; // the one that will be reported to netdata
343 } *target_root = NULL, *default_target = NULL;
347 // find or create a target
348 // there are targets that are just agregated to other target (the second argument)
349 struct target *get_target(const char *id, struct target *target)
351 const char *nid = id;
352 if(nid[0] == '-') nid++;
355 for(w = target_root ; w ; w = w->next)
356 if(strncmp(nid, w->id, MAX_NAME) == 0) return w;
358 w = calloc(sizeof(struct target), 1);
360 error("Cannot allocate %lu bytes of memory", (unsigned long)sizeof(struct target));
364 strncpy(w->id, nid, MAX_NAME);
365 strncpy(w->name, nid, MAX_NAME);
366 strncpy(w->compare, nid, MAX_COMPARE_NAME);
367 if(id[0] == '-') w->hidden = 1;
371 w->next = target_root;
374 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:"");
379 // read the process groups file
380 int read_process_groups(const char *name)
383 char filename[FILENAME_MAX + 1];
385 snprintf(filename, FILENAME_MAX, "%s/apps_%s.conf", CONFIG_DIR, name);
387 if(debug) fprintf(stderr, "apps.plugin: process groups file: '%s'\n", filename);
388 FILE *fp = fopen(filename, "r");
390 error("Cannot open file '%s'", filename);
395 while(fgets(buffer, 4096, fp) != NULL) {
396 int whidden = 0, wdebug = 0;
399 // if(debug) fprintf(stderr, "apps.plugin: \tread %s\n", buffer);
401 char *s = buffer, *t, *p;
403 if(!s || !*s || *s == '#') continue;
405 if(debug) fprintf(stderr, "apps.plugin: \tread %s\n", s);
410 if(!t || !*t) continue;
432 if(debug) fprintf(stderr, "apps.plugin: \t\ttarget %s\n", t);
434 struct target *w = NULL;
438 while((p = strsep(&s, " "))) {
440 if(!p || !*p) continue;
442 struct target *n = get_target(p, w);
450 if(w) strncpy(w->name, t, MAX_NAME);
451 if(!count) error("The line %ld on file '%s', for group '%s' does not state any process names.", line, filename, t);
455 default_target = get_target("+p!o@w#e$i^r&7*5(-i)l-o_", NULL); // match nothing
456 strncpy(default_target->name, "other", MAX_NAME);
462 // ----------------------------------------------------------------------------
463 // data to store for each pid
468 char comm[MAX_COMPARE_NAME + 1];
476 unsigned long long minflt;
477 unsigned long long cminflt;
478 unsigned long long majflt;
479 unsigned long long cmajflt;
480 unsigned long long utime;
481 unsigned long long stime;
482 unsigned long long cutime;
483 unsigned long long cstime;
487 // int64_t itrealvalue;
488 // unsigned long long starttime;
489 // unsigned long long vsize;
490 unsigned long long rss;
491 // unsigned long long rsslim;
492 // unsigned long long starcode;
493 // unsigned long long endcode;
494 // unsigned long long startstack;
495 // unsigned long long kstkesp;
496 // unsigned long long kstkeip;
499 // uint64_t sigignore;
500 // uint64_t sigcatch;
504 // int32_t exit_signal;
505 // int32_t processor;
506 // uint32_t rt_priority;
508 // unsigned long long delayacct_blkio_ticks;
509 // uint64_t guest_time;
510 // int64_t cguest_time;
512 unsigned long long statm_size;
513 unsigned long long statm_resident;
514 unsigned long long statm_share;
515 unsigned long long statm_text;
516 unsigned long long statm_lib;
517 unsigned long long statm_data;
518 unsigned long long statm_dirty;
520 unsigned long long io_logical_bytes_read;
521 unsigned long long io_logical_bytes_written;
522 unsigned long long io_read_calls;
523 unsigned long long io_write_calls;
524 unsigned long long io_storage_bytes_read;
525 unsigned long long io_storage_bytes_written;
526 unsigned long long io_cancelled_write_bytes;
528 #ifdef INCLUDE_CHILDS
529 unsigned long long old_utime;
530 unsigned long long old_stime;
531 unsigned long long old_minflt;
532 unsigned long long old_majflt;
534 unsigned long long old_cutime;
535 unsigned long long old_cstime;
536 unsigned long long old_cminflt;
537 unsigned long long old_cmajflt;
539 unsigned long long fix_cutime;
540 unsigned long long fix_cstime;
541 unsigned long long fix_cminflt;
542 unsigned long long fix_cmajflt;
544 unsigned long long diff_cutime;
545 unsigned long long diff_cstime;
546 unsigned long long diff_cminflt;
547 unsigned long long diff_cmajflt;
550 int *fds; // array of fds it uses
551 int fds_size; // the size of the fds array
553 int childs; // number of processes directly referencing this
554 int updated; // 1 when update
555 int merged; // 1 when it has been merged to its parent
557 struct target *target;
558 struct pid_stat *parent;
559 struct pid_stat *prev;
560 struct pid_stat *next;
561 } *root_of_pids = NULL, **all_pids;
563 long all_pids_count = 0;
565 struct pid_stat *get_pid_entry(pid_t pid)
568 all_pids[pid]->new_entry = 0;
569 return all_pids[pid];
572 all_pids[pid] = calloc(sizeof(struct pid_stat), 1);
574 error("Cannot allocate %lu bytes of memory", (unsigned long)sizeof(struct pid_stat));
578 all_pids[pid]->fds = calloc(sizeof(int), 100);
579 if(!all_pids[pid]->fds)
580 error("Cannot allocate %ld bytes of memory", (unsigned long)(sizeof(int) * 100));
581 else all_pids[pid]->fds_size = 100;
583 if(root_of_pids) root_of_pids->prev = all_pids[pid];
584 all_pids[pid]->next = root_of_pids;
585 root_of_pids = all_pids[pid];
587 all_pids[pid]->pid = pid;
588 all_pids[pid]->new_entry = 1;
590 return all_pids[pid];
593 void del_pid_entry(pid_t pid)
595 if(!all_pids[pid]) return;
597 if(debug) fprintf(stderr, "apps.plugin: process %d %s exited, deleting it.\n", pid, all_pids[pid]->comm);
599 if(root_of_pids == all_pids[pid]) root_of_pids = all_pids[pid]->next;
600 if(all_pids[pid]->next) all_pids[pid]->next->prev = all_pids[pid]->prev;
601 if(all_pids[pid]->prev) all_pids[pid]->prev->next = all_pids[pid]->next;
603 if(all_pids[pid]->fds) free(all_pids[pid]->fds);
605 all_pids[pid] = NULL;
609 // ----------------------------------------------------------------------------
610 // update pids from proc
612 int read_proc_pid_stat(struct pid_stat *p) {
613 char filename[FILENAME_MAX + 1];
615 snprintf(filename, FILENAME_MAX, "/proc/%d/stat", p->pid);
617 ff = procfile_reopen(ff, filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
620 ff = procfile_readall(ff);
622 // procfile_close(ff);
629 p->comm[MAX_COMPARE_NAME] = '\0';
632 char *s = procfile_lineword(ff, 0, 1);
634 size_t len = strlen(s);
636 while(len && s[len - 1] != ')') {
637 if(blen < MAX_COMPARE_NAME) {
638 strncpy(&p->comm[blen], s, MAX_COMPARE_NAME - blen);
639 blen = strlen(p->comm);
643 s = procfile_lineword(ff, 0, 1+i);
646 if(len && s[len - 1] == ')') s[len - 1] = '\0';
647 if(blen < MAX_COMPARE_NAME)
648 strncpy(&p->comm[blen], s, MAX_COMPARE_NAME - blen);
650 // p->pid = atol(procfile_lineword(ff, 0, 0+i));
652 // p->state = *(procfile_lineword(ff, 0, 2+i));
653 p->ppid = atol(procfile_lineword(ff, 0, 3+i));
654 // p->pgrp = atol(procfile_lineword(ff, 0, 4+i));
655 // p->session = atol(procfile_lineword(ff, 0, 5+i));
656 // p->tty_nr = atol(procfile_lineword(ff, 0, 6+i));
657 // p->tpgid = atol(procfile_lineword(ff, 0, 7+i));
658 // p->flags = strtoull(procfile_lineword(ff, 0, 8+i), NULL, 10);
659 p->minflt = strtoull(procfile_lineword(ff, 0, 9+i), NULL, 10);
660 p->cminflt = strtoull(procfile_lineword(ff, 0, 10+i), NULL, 10);
661 p->majflt = strtoull(procfile_lineword(ff, 0, 11+i), NULL, 10);
662 p->cmajflt = strtoull(procfile_lineword(ff, 0, 12+i), NULL, 10);
663 p->utime = strtoull(procfile_lineword(ff, 0, 13+i), NULL, 10);
664 p->stime = strtoull(procfile_lineword(ff, 0, 14+i), NULL, 10);
665 p->cutime = strtoull(procfile_lineword(ff, 0, 15+i), NULL, 10);
666 p->cstime = strtoull(procfile_lineword(ff, 0, 16+i), NULL, 10);
667 // p->priority = strtoull(procfile_lineword(ff, 0, 17+i), NULL, 10);
668 // p->nice = strtoull(procfile_lineword(ff, 0, 18+i), NULL, 10);
669 p->num_threads = atol(procfile_lineword(ff, 0, 19+i));
670 // p->itrealvalue = strtoull(procfile_lineword(ff, 0, 20+i), NULL, 10);
671 // p->starttime = strtoull(procfile_lineword(ff, 0, 21+i), NULL, 10);
672 // p->vsize = strtoull(procfile_lineword(ff, 0, 22+i), NULL, 10);
673 p->rss = strtoull(procfile_lineword(ff, 0, 23+i), NULL, 10);
674 // p->rsslim = strtoull(procfile_lineword(ff, 0, 24+i), NULL, 10);
675 // p->starcode = strtoull(procfile_lineword(ff, 0, 25+i), NULL, 10);
676 // p->endcode = strtoull(procfile_lineword(ff, 0, 26+i), NULL, 10);
677 // p->startstack = strtoull(procfile_lineword(ff, 0, 27+i), NULL, 10);
678 // p->kstkesp = strtoull(procfile_lineword(ff, 0, 28+i), NULL, 10);
679 // p->kstkeip = strtoull(procfile_lineword(ff, 0, 29+i), NULL, 10);
680 // p->signal = strtoull(procfile_lineword(ff, 0, 30+i), NULL, 10);
681 // p->blocked = strtoull(procfile_lineword(ff, 0, 31+i), NULL, 10);
682 // p->sigignore = strtoull(procfile_lineword(ff, 0, 32+i), NULL, 10);
683 // p->sigcatch = strtoull(procfile_lineword(ff, 0, 33+i), NULL, 10);
684 // p->wchan = strtoull(procfile_lineword(ff, 0, 34+i), NULL, 10);
685 // p->nswap = strtoull(procfile_lineword(ff, 0, 35+i), NULL, 10);
686 // p->cnswap = strtoull(procfile_lineword(ff, 0, 36+i), NULL, 10);
687 // p->exit_signal = atol(procfile_lineword(ff, 0, 37+i));
688 // p->processor = atol(procfile_lineword(ff, 0, 38+i));
689 // p->rt_priority = strtoul(procfile_lineword(ff, 0, 39+i), NULL, 10);
690 // p->policy = strtoul(procfile_lineword(ff, 0, 40+i), NULL, 10);
691 // p->delayacct_blkio_ticks = strtoull(procfile_lineword(ff, 0, 41+i), NULL, 10);
692 // p->guest_time = strtoull(procfile_lineword(ff, 0, 42+i), NULL, 10);
693 // p->cguest_time = strtoull(procfile_lineword(ff, 0, 43), NULL, 10);
695 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);
697 // procfile_close(ff);
701 int read_proc_pid_statm(struct pid_stat *p) {
702 char filename[FILENAME_MAX + 1];
704 snprintf(filename, FILENAME_MAX, "/proc/%d/statm", p->pid);
706 ff = procfile_reopen(ff, filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
709 ff = procfile_readall(ff);
711 // procfile_close(ff);
717 p->statm_size = strtoull(procfile_lineword(ff, 0, 0), NULL, 10);
718 p->statm_resident = strtoull(procfile_lineword(ff, 0, 1), NULL, 10);
719 p->statm_share = strtoull(procfile_lineword(ff, 0, 2), NULL, 10);
720 p->statm_text = strtoull(procfile_lineword(ff, 0, 3), NULL, 10);
721 p->statm_lib = strtoull(procfile_lineword(ff, 0, 4), NULL, 10);
722 p->statm_data = strtoull(procfile_lineword(ff, 0, 5), NULL, 10);
723 p->statm_dirty = strtoull(procfile_lineword(ff, 0, 6), NULL, 10);
725 // procfile_close(ff);
729 int read_proc_pid_io(struct pid_stat *p) {
730 char filename[FILENAME_MAX + 1];
732 snprintf(filename, FILENAME_MAX, "/proc/%d/io", p->pid);
734 ff = procfile_reopen(ff, filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
737 ff = procfile_readall(ff);
739 // procfile_close(ff);
745 p->io_logical_bytes_read = strtoull(procfile_lineword(ff, 0, 1), NULL, 10);
746 p->io_logical_bytes_written = strtoull(procfile_lineword(ff, 1, 1), NULL, 10);
747 p->io_read_calls = strtoull(procfile_lineword(ff, 2, 1), NULL, 10);
748 p->io_write_calls = strtoull(procfile_lineword(ff, 3, 1), NULL, 10);
749 p->io_storage_bytes_read = strtoull(procfile_lineword(ff, 4, 1), NULL, 10);
750 p->io_storage_bytes_written = strtoull(procfile_lineword(ff, 5, 1), NULL, 10);
751 p->io_cancelled_write_bytes = strtoull(procfile_lineword(ff, 6, 1), NULL, 10);
753 // procfile_close(ff);
758 // ----------------------------------------------------------------------------
760 #ifdef INCLUDE_CHILDS
761 // print a tree view of all processes
762 int walk_down(pid_t pid, int level) {
763 struct pid_stat *p = NULL;
767 for(i = 0; i < level; i++) b[i] = '\t';
772 for(p = root_of_pids; p ; p = p->next) {
774 ret += walk_down(p->pid, level+1);
780 if(!p->updated) ret += 1;
781 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"
782 , b, p->comm, p->pid, p->updated?"OK":"KILLED", p->target->name, p->childs
783 , p->utime, p->utime - p->old_utime
784 , p->stime, p->stime - p->old_stime
785 , p->cutime, p->cutime - p->old_cutime
786 , p->cstime, p->cstime - p->old_cstime
787 , p->minflt, p->minflt - p->old_minflt
788 , p->majflt, p->majflt - p->old_majflt
789 , p->cminflt, p->cminflt - p->old_cminflt
790 , p->cmajflt, p->cmajflt - p->old_cmajflt
799 // ----------------------------------------------------------------------------
801 // this is used to keep a global list of all open files of the system
802 // it is needed in order to figure out the unique files a process tree has open
804 #define FILE_DESCRIPTORS_INCREASE_STEP 100
806 struct file_descriptor {
816 int all_files_len = 0;
817 int all_files_size = 0;
819 int file_descriptor_compare(void* a, void* b) {
820 if(((struct file_descriptor *)a)->magic != 0x0BADCAFE || ((struct file_descriptor *)b)->magic != 0x0BADCAFE)
821 error("Corrupted index data detected. Please report this.");
823 if(((struct file_descriptor *)a)->hash < ((struct file_descriptor *)b)->hash)
825 else if(((struct file_descriptor *)a)->hash > ((struct file_descriptor *)b)->hash)
827 else return strcmp(((struct file_descriptor *)a)->name, ((struct file_descriptor *)b)->name);
829 int file_descriptor_iterator(avl *a) { if(a) {}; return 0; }
831 avl_tree all_files_index = {
833 file_descriptor_compare,
834 #ifdef AVL_LOCK_WITH_MUTEX
835 PTHREAD_MUTEX_INITIALIZER
837 PTHREAD_RWLOCK_INITIALIZER
841 static struct file_descriptor *file_descriptor_find(const char *name, uint32_t hash) {
842 struct file_descriptor *result = NULL, tmp;
843 tmp.hash = (hash)?hash:simple_hash(name);
847 tmp.magic = 0x0BADCAFE;
849 avl_search(&all_files_index, (avl *)&tmp, file_descriptor_iterator, (avl **)&result);
853 #define file_descriptor_add(fd) avl_insert(&all_files_index, (avl *)(fd))
854 #define file_descriptor_remove(fd) avl_remove(&all_files_index, (avl *)(fd))
856 #define FILETYPE_OTHER 0
857 #define FILETYPE_FILE 1
858 #define FILETYPE_PIPE 2
859 #define FILETYPE_SOCKET 3
860 #define FILETYPE_INOTIFY 4
861 #define FILETYPE_EVENTFD 5
862 #define FILETYPE_EVENTPOLL 6
863 #define FILETYPE_TIMERFD 7
864 #define FILETYPE_SIGNALFD 8
866 void file_descriptor_not_used(int id)
868 if(id > 0 && id < all_files_size) {
869 if(all_files[id].magic != 0x0BADCAFE) {
870 error("Ignoring request to remove empty file id %d.", id);
874 if(debug) fprintf(stderr, "apps.plugin: decreasing slot %d (count = %ld).\n", id, all_files[id].count);
876 if(all_files[id].count > 0) {
877 all_files[id].count--;
879 if(!all_files[id].count) {
880 if(debug) fprintf(stderr, "apps.plugin: >> slot %d is empty.\n", id);
881 file_descriptor_remove(&all_files[id]);
882 all_files[id].magic = 0x00000000;
887 error("Request to decrease counter of fd %d (%s), while the use counter is 0", id, all_files[id].name);
889 else error("Request to decrease counter of fd %d, which is outside the array size (1 to %d)", id, all_files_size);
892 unsigned long file_descriptor_find_or_add(const char *name)
894 static int last_pos = 0;
895 uint32_t hash = simple_hash(name);
897 if(debug) fprintf(stderr, "apps.plugin: adding or finding name '%s' with hash %u\n", name, hash);
899 struct file_descriptor *fd = file_descriptor_find(name, hash);
902 if(debug) fprintf(stderr, "apps.plugin: >> found on slot %ld\n", fd->pos);
908 // check we have enough memory to add it
909 if(!all_files || all_files_len == all_files_size) {
910 void *old = all_files;
913 // there is no empty slot
914 if(debug) fprintf(stderr, "apps.plugin: extending fd array to %d entries\n", all_files_size + FILE_DESCRIPTORS_INCREASE_STEP);
915 all_files = realloc(all_files, (all_files_size + FILE_DESCRIPTORS_INCREASE_STEP) * sizeof(struct file_descriptor));
917 // if the address changed, we have to rebuild the index
918 // since all pointers are now invalid
919 if(old && old != (void *)all_files) {
920 if(debug) fprintf(stderr, "apps.plugin: >> re-indexing.\n");
921 all_files_index.root = NULL;
922 for(i = 0; i < all_files_size; i++) {
923 if(!all_files[i].count) continue;
924 file_descriptor_add(&all_files[i]);
926 if(debug) fprintf(stderr, "apps.plugin: >> re-indexing done.\n");
929 for(i = all_files_size; i < (all_files_size + FILE_DESCRIPTORS_INCREASE_STEP); i++) {
930 all_files[i].count = 0;
931 all_files[i].name = NULL;
932 all_files[i].magic = 0x00000000;
933 all_files[i].pos = i;
936 if(!all_files_size) all_files_len = 1;
937 all_files_size += FILE_DESCRIPTORS_INCREASE_STEP;
940 if(debug) fprintf(stderr, "apps.plugin: >> searching for empty slot.\n");
942 // search for an empty slot
944 for(i = 0, c = last_pos ; i < all_files_size ; i++, c++) {
945 if(c >= all_files_size) c = 0;
948 if(!all_files[c].count) {
949 if(debug) fprintf(stderr, "apps.plugin: >> Examining slot %d.\n", c);
951 if(all_files[c].magic == 0x0BADCAFE && all_files[c].name && file_descriptor_find(all_files[c].name, all_files[c].hash))
952 error("fd on position %d is not cleared properly. It still has %s in it.\n", c, all_files[c].name);
954 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);
955 if(all_files[c].name) free((void *)all_files[c].name);
956 all_files[c].name = NULL;
961 if(i == all_files_size) {
962 fatal("We should find an empty slot, but there isn't any");
965 if(debug) fprintf(stderr, "apps.plugin: >> updating slot %d.\n", c);
969 // else we have an empty slot in 'c'
971 int type = FILETYPE_OTHER;
972 if(name[0] == '/') type = FILETYPE_FILE;
973 else if(strncmp(name, "pipe:", 5) == 0) type = FILETYPE_PIPE;
974 else if(strncmp(name, "socket:", 7) == 0) type = FILETYPE_SOCKET;
975 else if(strcmp(name, "anon_inode:inotify") == 0 || strcmp(name, "inotify") == 0) type = FILETYPE_INOTIFY;
976 else if(strcmp(name, "anon_inode:[eventfd]") == 0) type = FILETYPE_EVENTFD;
977 else if(strcmp(name, "anon_inode:[eventpoll]") == 0) type = FILETYPE_EVENTPOLL;
978 else if(strcmp(name, "anon_inode:[timerfd]") == 0) type = FILETYPE_TIMERFD;
979 else if(strcmp(name, "anon_inode:[signalfd]") == 0) type = FILETYPE_SIGNALFD;
980 else if(strncmp(name, "anon_inode:", 11) == 0) {
981 if(debug) fprintf(stderr, "apps.plugin: FIXME: unknown anonymous inode: %s\n", name);
982 type = FILETYPE_OTHER;
985 if(debug) fprintf(stderr, "apps.plugin: FIXME: cannot understand linkname: %s\n", name);
986 type = FILETYPE_OTHER;
989 all_files[c].name = strdup(name);
990 all_files[c].hash = hash;
991 all_files[c].type = type;
992 all_files[c].pos = c;
993 all_files[c].count = 1;
994 all_files[c].magic = 0x0BADCAFE;
996 file_descriptor_add(&all_files[c]);
998 if(debug) fprintf(stderr, "apps.plugin: using fd position %d (name: %s)\n", c, all_files[c].name);
1004 // 1. read all files in /proc
1005 // 2. for each numeric directory:
1006 // i. read /proc/pid/stat
1007 // ii. read /proc/pid/statm
1008 // iii. read /proc/pid/io (requires root access)
1009 // iii. read the entries in directory /proc/pid/fd (requires root access)
1011 // a. find or create a struct file_descriptor
1012 // b. cleanup any old/unused file_descriptors
1014 // after all these, some pids may be linked to targets, while others may not
1016 // in case of errors, only 1 every 1000 errors is printed
1017 // to avoid filling up all disk space
1018 // if debug is enabled, all errors are printed
1020 int update_from_proc(void)
1022 static long count_errors = 0;
1024 char filename[FILENAME_MAX+1];
1025 DIR *dir = opendir("/proc");
1028 struct dirent *file = NULL;
1029 struct pid_stat *p = NULL;
1031 // mark them all as un-updated
1033 for(p = root_of_pids; p ; p = p->next) {
1042 while((file = readdir(dir))) {
1043 char *endptr = file->d_name;
1044 pid_t pid = strtoul(file->d_name, &endptr, 10);
1045 if(pid <= 0 || pid > pid_max || endptr == file->d_name || *endptr != '\0') continue;
1047 p = get_pid_entry(pid);
1050 // --------------------------------------------------------------------
1053 if(read_proc_pid_stat(p)) {
1054 if(!count_errors++ || debug || (p->target && p->target->debug))
1055 error("Cannot process /proc/%d/stat", pid);
1059 if(p->ppid < 0 || p->ppid > pid_max) p->ppid = 0;
1062 // --------------------------------------------------------------------
1063 // /proc/<pid>/statm
1065 if(read_proc_pid_statm(p)) {
1066 if(!count_errors++ || debug || (p->target && p->target->debug))
1067 error("Cannot process /proc/%d/statm", pid);
1073 // --------------------------------------------------------------------
1076 if(read_proc_pid_io(p)) {
1077 if(!count_errors++ || debug || (p->target && p->target->debug))
1078 error("Cannot process /proc/%d/io", pid);
1083 // --------------------------------------------------------------------
1086 // check if it is target
1087 // we do this only once, the first time this pid is loaded
1089 if(debug) fprintf(stderr, "apps.plugin: \tJust added %s\n", p->comm);
1092 for(w = target_root; w ; w = w->next) {
1093 // if(debug || (p->target && p->target->debug)) fprintf(stderr, "apps.plugin: \t\tcomparing '%s' with '%s'\n", w->compare, p->comm);
1095 if(strcmp(w->compare, p->comm) == 0) {
1096 if(w->target) p->target = w->target;
1099 if(debug || (p->target && p->target->debug)) fprintf(stderr, "apps.plugin: \t\t%s linked to target %s\n", p->comm, p->target->name);
1104 // --------------------------------------------------------------------
1107 snprintf(filename, FILENAME_MAX, "/proc/%s/fd", file->d_name);
1108 DIR *fds = opendir(filename);
1112 char fdname[FILENAME_MAX + 1];
1113 char linkname[FILENAME_MAX + 1];
1115 // make the array negative
1116 for(c = 0 ; c < p->fds_size ; c++) p->fds[c] = -p->fds[c];
1118 while((de = readdir(fds))) {
1119 if(strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) continue;
1121 // check if the fds array is small
1122 int fdid = atol(de->d_name);
1123 if(fdid < 0) continue;
1124 if(fdid >= p->fds_size) {
1125 // it is small, extend it
1126 if(debug) fprintf(stderr, "apps.plugin: extending fd memory slots for %s from %d to %d\n", p->comm, p->fds_size, fdid + 100);
1127 p->fds = realloc(p->fds, (fdid + 100) * sizeof(int));
1129 error("Cannot re-allocate fds for %s", p->comm);
1133 // and initialize it
1134 for(c = p->fds_size ; c < (fdid + 100) ; c++) p->fds[c] = 0;
1135 p->fds_size = fdid + 100;
1138 if(p->fds[fdid] == 0) {
1139 // we don't know this fd, get it
1141 sprintf(fdname, "/proc/%s/fd/%s", file->d_name, de->d_name);
1142 int l = readlink(fdname, linkname, FILENAME_MAX);
1144 if(debug || (p->target && p->target->debug)) {
1145 if(!count_errors++ || debug || (p->target && p->target->debug))
1146 error("Cannot read link %s", fdname);
1153 // if another process already has this, we will get
1155 p->fds[fdid] = file_descriptor_find_or_add(linkname);
1158 // else make it positive again, we need it
1159 // of course, the actual file may have changed, but we don't care so much
1160 // FIXME: we could compare the inode as returned by readdir direct structure
1161 else p->fds[fdid] = -p->fds[fdid];
1165 // remove all the negative file descriptors
1166 for(c = 0 ; c < p->fds_size ; c++) if(p->fds[c] < 0) {
1167 file_descriptor_not_used(-p->fds[c]);
1172 // --------------------------------------------------------------------
1175 // mark it as updated
1178 if(count_errors > 1000) {
1179 error("%ld more errors encountered\n", count_errors - 1);
1189 // ----------------------------------------------------------------------------
1190 // update statistics on the targets
1192 // 1. link all childs to their parents
1193 // 2. go from bottom to top, marking as merged all childs to their parents
1194 // this step links all parents without a target to the child target, if any
1195 // 3. link all top level processes (the ones not merged) to the default target
1196 // 4. go from top to bottom, linking all childs without a target, to their parent target
1197 // after this step, all processes have a target
1198 // [5. for each killed pid (updated = 0), remove its usage from its target]
1199 // 6. zero all targets
1200 // 7. concentrate all values on the targets
1201 // 8. remove all killed processes
1202 // 9. find the unique file count for each target
1204 void update_statistics(void)
1207 struct pid_stat *p = NULL;
1209 // link all parents and update childs count
1210 for(p = root_of_pids; p ; p = p->next) {
1211 if(p->ppid > 0 && p->ppid <= pid_max && all_pids[p->ppid]) {
1212 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);
1214 p->parent = all_pids[p->ppid];
1215 p->parent->childs++;
1217 else if(p->ppid != 0) error("pid %d %s states parent %d, but the later does not exist.", p->pid, p->comm, p->ppid);
1220 // find all the procs with 0 childs and merge them to their parents
1221 // repeat, until nothing more can be done.
1225 for(p = root_of_pids; p ; p = p->next) {
1226 // if this process does not have any childs, and
1227 // is not already merged, and
1228 // its parent has childs waiting to be merged, and
1229 // 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
1230 // and its parent is not init
1231 // then... merge them!
1232 if(!p->childs && !p->merged && p->parent && p->parent->childs && (p->target == p->parent->target || !p->parent->target || !p->target) && p->ppid != 1) {
1233 p->parent->childs--;
1236 // the parent inherits the child's target, if it does not have a target itself
1237 if(p->target && !p->parent->target) {
1238 p->parent->target = p->target;
1239 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);
1245 if(debug) fprintf(stderr, "apps.plugin: merged %d processes\n", found);
1248 // give a default target on all top level processes
1249 // init goes always to default target
1250 if(all_pids[1]) all_pids[1]->target = default_target;
1252 for(p = root_of_pids; p ; p = p->next) {
1253 // if the process is not merged itself
1254 // then is is a top level process
1255 if(!p->merged && !p->target) p->target = default_target;
1257 #ifdef INCLUDE_CHILDS
1258 // by the way, update the diffs
1259 // will be used later for substracting killed process times
1260 p->diff_cutime = p->utime - p->cutime;
1261 p->diff_cstime = p->stime - p->cstime;
1262 p->diff_cminflt = p->minflt - p->cminflt;
1263 p->diff_cmajflt = p->majflt - p->cmajflt;
1267 // give a target to all merged child processes
1271 for(p = root_of_pids; p ; p = p->next) {
1272 if(!p->target && p->merged && p->parent && p->parent->target) {
1273 p->target = p->parent->target;
1279 #ifdef INCLUDE_CHILDS
1280 // for each killed process, remove its values from the parents
1281 // sums (we had already added them in a previous loop)
1282 for(p = root_of_pids; p ; p = p->next) {
1283 if(p->updated) continue;
1285 if(debug) fprintf(stderr, "apps.plugin: UNMERGING %d %s\n", p->pid, p->comm);
1287 unsigned long long diff_utime = p->utime + p->cutime + p->fix_cutime;
1288 unsigned long long diff_stime = p->stime + p->cstime + p->fix_cstime;
1289 unsigned long long diff_minflt = p->minflt + p->cminflt + p->fix_cminflt;
1290 unsigned long long diff_majflt = p->majflt + p->cmajflt + p->fix_cmajflt;
1292 struct pid_stat *t = p;
1293 while((t = t->parent)) {
1294 if(!t->updated) continue;
1296 unsigned long long x;
1297 if(diff_utime && t->diff_cutime) {
1298 x = (t->diff_cutime < diff_utime)?t->diff_cutime:diff_utime;
1300 t->diff_cutime -= x;
1302 if(debug) fprintf(stderr, "apps.plugin: \t cutime %llu from %d %s %s\n", x, t->pid, t->comm, t->target->name);
1304 if(diff_stime && t->diff_cstime) {
1305 x = (t->diff_cstime < diff_stime)?t->diff_cstime:diff_stime;
1307 t->diff_cstime -= x;
1309 if(debug) fprintf(stderr, "apps.plugin: \t cstime %llu from %d %s %s\n", x, t->pid, t->comm, t->target->name);
1311 if(diff_minflt && t->diff_cminflt) {
1312 x = (t->diff_cminflt < diff_minflt)?t->diff_cminflt:diff_minflt;
1314 t->diff_cminflt -= x;
1315 t->fix_cminflt += x;
1316 if(debug) fprintf(stderr, "apps.plugin: \t cminflt %llu from %d %s %s\n", x, t->pid, t->comm, t->target->name);
1318 if(diff_majflt && t->diff_cmajflt) {
1319 x = (t->diff_cmajflt < diff_majflt)?t->diff_cmajflt:diff_majflt;
1321 t->diff_cmajflt -= x;
1322 t->fix_cmajflt += x;
1323 if(debug) fprintf(stderr, "apps.plugin: \t cmajflt %llu from %d %s %s\n", x, t->pid, t->comm, t->target->name);
1327 if(diff_utime) error("Cannot fix up utime %llu", diff_utime);
1328 if(diff_stime) error("Cannot fix up stime %llu", diff_stime);
1329 if(diff_minflt) error("Cannot fix up minflt %llu", diff_minflt);
1330 if(diff_majflt) error("Cannot fix up majflt %llu", diff_majflt);
1334 // zero all the targets
1337 for (w = target_root; w ; w = w->next) {
1340 w->fds = calloc(sizeof(int), all_files_size);
1342 error("Cannot allocate memory for fds in %s", w->name);
1357 w->statm_resident = 0;
1364 w->io_logical_bytes_read = 0;
1365 w->io_logical_bytes_written = 0;
1366 w->io_read_calls = 0;
1367 w->io_write_calls = 0;
1368 w->io_storage_bytes_read = 0;
1369 w->io_storage_bytes_written = 0;
1370 w->io_cancelled_write_bytes = 0;
1373 #ifdef INCLUDE_CHILDS
1374 if(debug) walk_down(0, 1);
1377 // concentrate everything on the targets
1378 for(p = root_of_pids; p ; p = p->next) {
1380 error("pid %d %s was left without a target!", p->pid, p->comm);
1385 p->target->cutime += p->cutime; // - p->fix_cutime;
1386 p->target->cstime += p->cstime; // - p->fix_cstime;
1387 p->target->cminflt += p->cminflt; // - p->fix_cminflt;
1388 p->target->cmajflt += p->cmajflt; // - p->fix_cmajflt;
1390 p->target->utime += p->utime; //+ (p->pid != 1)?(p->cutime - p->fix_cutime):0;
1391 p->target->stime += p->stime; //+ (p->pid != 1)?(p->cstime - p->fix_cstime):0;
1392 p->target->minflt += p->minflt; //+ (p->pid != 1)?(p->cminflt - p->fix_cminflt):0;
1393 p->target->majflt += p->majflt; //+ (p->pid != 1)?(p->cmajflt - p->fix_cmajflt):0;
1395 //if(p->num_threads < 0)
1396 // error("Negative threads number for pid '%s' (%d): %d", p->comm, p->pid, p->num_threads);
1398 //if(p->num_threads > 10000)
1399 // error("Excessive threads number for pid '%s' (%d): %d", p->comm, p->pid, p->num_threads);
1401 p->target->num_threads += p->num_threads;
1402 p->target->rss += p->rss;
1404 p->target->statm_size += p->statm_size;
1405 p->target->statm_resident += p->statm_resident;
1406 p->target->statm_share += p->statm_share;
1407 p->target->statm_text += p->statm_text;
1408 p->target->statm_lib += p->statm_lib;
1409 p->target->statm_data += p->statm_data;
1410 p->target->statm_dirty += p->statm_dirty;
1412 p->target->io_logical_bytes_read += p->io_logical_bytes_read;
1413 p->target->io_logical_bytes_written += p->io_logical_bytes_written;
1414 p->target->io_read_calls += p->io_read_calls;
1415 p->target->io_write_calls += p->io_write_calls;
1416 p->target->io_storage_bytes_read += p->io_storage_bytes_read;
1417 p->target->io_storage_bytes_written += p->io_storage_bytes_written;
1418 p->target->io_cancelled_write_bytes += p->io_cancelled_write_bytes;
1420 p->target->processes++;
1422 for(c = 0; c < p->fds_size ;c++) {
1423 if(p->fds[c] == 0) continue;
1424 if(p->fds[c] < all_files_size) {
1425 if(p->target->fds) p->target->fds[p->fds[c]]++;
1428 error("Invalid fd number %d", p->fds[c]);
1431 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);
1433 /* 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);
1434 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);
1435 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);
1436 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);
1437 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);
1438 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);
1439 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);
1440 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);
1442 #ifdef INCLUDE_CHILDS
1443 p->old_utime = p->utime;
1444 p->old_cutime = p->cutime;
1445 p->old_stime = p->stime;
1446 p->old_cstime = p->cstime;
1447 p->old_minflt = p->minflt;
1448 p->old_majflt = p->majflt;
1449 p->old_cminflt = p->cminflt;
1450 p->old_cmajflt = p->cmajflt;
1454 // since the process has exited, the user
1455 // will see a drop in our charts, because the incremental
1456 // values of this process will not be there
1458 // add them to the fix_* values and they will be added to
1459 // the reported values, so that the report goes steady
1460 p->target->fix_minflt += p->minflt;
1461 p->target->fix_majflt += p->majflt;
1462 p->target->fix_utime += p->utime;
1463 p->target->fix_stime += p->stime;
1464 p->target->fix_cminflt += p->cminflt;
1465 p->target->fix_cmajflt += p->cmajflt;
1466 p->target->fix_cutime += p->cutime;
1467 p->target->fix_cstime += p->cstime;
1469 p->target->fix_io_logical_bytes_read += p->io_logical_bytes_read;
1470 p->target->fix_io_logical_bytes_written += p->io_logical_bytes_written;
1471 p->target->fix_io_read_calls += p->io_read_calls;
1472 p->target->fix_io_write_calls += p->io_write_calls;
1473 p->target->fix_io_storage_bytes_read += p->io_storage_bytes_read;
1474 p->target->fix_io_storage_bytes_written += p->io_storage_bytes_written;
1475 p->target->fix_io_cancelled_write_bytes += p->io_cancelled_write_bytes;
1479 // fprintf(stderr, "\n");
1480 // cleanup all un-updated processed (exited, killed, etc)
1481 for(p = root_of_pids; p ;) {
1483 // 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);
1485 for(c = 0 ; c < p->fds_size ; c++) if(p->fds[c] > 0) {
1486 file_descriptor_not_used(p->fds[c]);
1497 for (w = target_root; w ; w = w->next) {
1501 w->openinotifies = 0;
1502 w->openeventfds = 0;
1503 w->opentimerfds = 0;
1504 w->opensignalfds = 0;
1505 w->openeventpolls = 0;
1508 for(c = 1; c < all_files_size ;c++) {
1509 if(w->fds && w->fds[c] > 0) switch(all_files[c].type) {
1518 case FILETYPE_SOCKET:
1522 case FILETYPE_INOTIFY:
1526 case FILETYPE_EVENTFD:
1530 case FILETYPE_TIMERFD:
1534 case FILETYPE_SIGNALFD:
1538 case FILETYPE_EVENTPOLL:
1539 w->openeventpolls++;
1552 // ----------------------------------------------------------------------------
1553 // update chart dimensions
1555 void show_dimensions(void)
1557 static struct timeval last = { 0, 0 };
1558 static struct rusage me_last;
1564 unsigned long long usec;
1565 unsigned long long cpuuser;
1566 unsigned long long cpusyst;
1569 gettimeofday(&last, NULL);
1570 getrusage(RUSAGE_SELF, &me_last);
1572 // the first time, give a zero to allow
1573 // netdata calibrate to the current time
1574 // usec = update_every * 1000000ULL;
1580 gettimeofday(&now, NULL);
1581 getrusage(RUSAGE_SELF, &me);
1583 usec = usecdiff(&now, &last);
1584 cpuuser = me.ru_utime.tv_sec * 1000000ULL + me.ru_utime.tv_usec;
1585 cpusyst = me.ru_stime.tv_sec * 1000000ULL + me.ru_stime.tv_usec;
1587 bcopy(&now, &last, sizeof(struct timeval));
1588 bcopy(&me, &me_last, sizeof(struct rusage));
1591 fprintf(stdout, "BEGIN apps.cpu %llu\n", usec);
1592 for (w = target_root; w ; w = w->next) {
1593 if(w->target || (!w->processes && !w->exposed)) continue;
1595 fprintf(stdout, "SET %s = %llu\n", w->name, w->utime + w->stime + w->fix_utime + w->fix_stime);
1597 fprintf(stdout, "END\n");
1599 fprintf(stdout, "BEGIN apps.cpu_user %llu\n", usec);
1600 for (w = target_root; w ; w = w->next) {
1601 if(w->target || (!w->processes && !w->exposed)) continue;
1603 fprintf(stdout, "SET %s = %llu\n", w->name, w->utime + w->fix_utime);
1605 fprintf(stdout, "END\n");
1607 fprintf(stdout, "BEGIN apps.cpu_system %llu\n", usec);
1608 for (w = target_root; w ; w = w->next) {
1609 if(w->target || (!w->processes && !w->exposed)) continue;
1611 fprintf(stdout, "SET %s = %llu\n", w->name, w->stime + w->fix_stime);
1613 fprintf(stdout, "END\n");
1615 fprintf(stdout, "BEGIN apps.threads %llu\n", usec);
1616 for (w = target_root; w ; w = w->next) {
1617 if(w->target || (!w->processes && !w->exposed)) continue;
1619 fprintf(stdout, "SET %s = %llu\n", w->name, w->num_threads);
1621 fprintf(stdout, "END\n");
1623 fprintf(stdout, "BEGIN apps.processes %llu\n", usec);
1624 for (w = target_root; w ; w = w->next) {
1625 if(w->target || (!w->processes && !w->exposed)) continue;
1627 fprintf(stdout, "SET %s = %lu\n", w->name, w->processes);
1629 fprintf(stdout, "END\n");
1631 fprintf(stdout, "BEGIN apps.mem %llu\n", usec);
1632 for (w = target_root; w ; w = w->next) {
1633 if(w->target || (!w->processes && !w->exposed)) continue;
1635 fprintf(stdout, "SET %s = %lld\n", w->name, (long long)w->statm_resident - (long long)w->statm_share);
1637 fprintf(stdout, "END\n");
1639 fprintf(stdout, "BEGIN apps.minor_faults %llu\n", usec);
1640 for (w = target_root; w ; w = w->next) {
1641 if(w->target || (!w->processes && !w->exposed)) continue;
1643 fprintf(stdout, "SET %s = %llu\n", w->name, w->minflt + w->fix_minflt);
1645 fprintf(stdout, "END\n");
1647 fprintf(stdout, "BEGIN apps.major_faults %llu\n", usec);
1648 for (w = target_root; w ; w = w->next) {
1649 if(w->target || (!w->processes && !w->exposed)) continue;
1651 fprintf(stdout, "SET %s = %llu\n", w->name, w->majflt + w->fix_majflt);
1653 fprintf(stdout, "END\n");
1655 fprintf(stdout, "BEGIN apps.lreads %llu\n", usec);
1656 for (w = target_root; w ; w = w->next) {
1657 if(w->target || (!w->processes && !w->exposed)) continue;
1659 fprintf(stdout, "SET %s = %llu\n", w->name, w->io_logical_bytes_read);
1661 fprintf(stdout, "END\n");
1663 fprintf(stdout, "BEGIN apps.lwrites %llu\n", usec);
1664 for (w = target_root; w ; w = w->next) {
1665 if(w->target || (!w->processes && !w->exposed)) continue;
1667 fprintf(stdout, "SET %s = %llu\n", w->name, w->io_logical_bytes_written);
1669 fprintf(stdout, "END\n");
1671 fprintf(stdout, "BEGIN apps.preads %llu\n", usec);
1672 for (w = target_root; w ; w = w->next) {
1673 if(w->target || (!w->processes && !w->exposed)) continue;
1675 fprintf(stdout, "SET %s = %llu\n", w->name, w->io_storage_bytes_read);
1677 fprintf(stdout, "END\n");
1679 fprintf(stdout, "BEGIN apps.pwrites %llu\n", usec);
1680 for (w = target_root; w ; w = w->next) {
1681 if(w->target || (!w->processes && !w->exposed)) continue;
1683 fprintf(stdout, "SET %s = %llu\n", w->name, w->io_storage_bytes_written);
1685 fprintf(stdout, "END\n");
1687 fprintf(stdout, "BEGIN apps.files %llu\n", usec);
1688 for (w = target_root; w ; w = w->next) {
1689 if(w->target || (!w->processes && !w->exposed)) continue;
1691 fprintf(stdout, "SET %s = %llu\n", w->name, w->openfiles);
1693 fprintf(stdout, "END\n");
1695 fprintf(stdout, "BEGIN apps.sockets %llu\n", usec);
1696 for (w = target_root; w ; w = w->next) {
1697 if(w->target || (!w->processes && !w->exposed)) continue;
1699 fprintf(stdout, "SET %s = %llu\n", w->name, w->opensockets);
1701 fprintf(stdout, "END\n");
1703 fprintf(stdout, "BEGIN apps.pipes %llu\n", usec);
1704 for (w = target_root; w ; w = w->next) {
1705 if(w->target || (!w->processes && !w->exposed)) continue;
1707 fprintf(stdout, "SET %s = %llu\n", w->name, w->openpipes);
1709 fprintf(stdout, "END\n");
1711 fprintf(stdout, "BEGIN netdata.apps_cpu %llu\n", usec);
1712 fprintf(stdout, "SET user = %llu\n", cpuuser);
1713 fprintf(stdout, "SET system = %llu\n", cpusyst);
1714 fprintf(stdout, "END\n");
1716 fprintf(stdout, "BEGIN netdata.apps_files %llu\n", usec);
1717 fprintf(stdout, "SET files = %llu\n", file_counter);
1718 fprintf(stdout, "SET pids = %ld\n", all_pids_count);
1719 fprintf(stdout, "SET fds = %d\n", all_files_len);
1720 fprintf(stdout, "SET targets = %ld\n", targets);
1721 fprintf(stdout, "END\n");
1727 // ----------------------------------------------------------------------------
1728 // generate the charts
1730 void show_charts(void)
1733 int newly_added = 0;
1735 for(w = target_root ; w ; w = w->next)
1736 if(!w->exposed && w->processes) {
1739 if(debug || w->debug) fprintf(stderr, "apps.plugin: %s just added - regenerating charts.\n", w->name);
1742 // nothing more to show
1743 if(!newly_added) return;
1745 // we have something new to show
1746 // update the charts
1747 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);
1748 for (w = target_root; w ; w = w->next) {
1749 if(w->target || (!w->processes && !w->exposed)) continue;
1751 fprintf(stdout, "DIMENSION %s '' incremental 100 %llu %s\n", w->name, (unsigned long long)(Hertz), w->hidden?"hidden,noreset":"noreset");
1754 fprintf(stdout, "CHART apps.mem '' 'Apps Dedicated Memory (w/o shared)' 'MB' apps apps stacked 20003 %d\n", update_every);
1755 for (w = target_root; w ; w = w->next) {
1756 if(w->target || (!w->processes && !w->exposed)) continue;
1758 fprintf(stdout, "DIMENSION %s '' absolute %ld %ld noreset\n", w->name, sysconf(_SC_PAGESIZE), 1024L*1024L);
1761 fprintf(stdout, "CHART apps.threads '' 'Apps Threads' 'threads' apps apps stacked 20005 %d\n", update_every);
1762 for (w = target_root; w ; w = w->next) {
1763 if(w->target || (!w->processes && !w->exposed)) continue;
1765 fprintf(stdout, "DIMENSION %s '' absolute 1 1 noreset\n", w->name);
1768 fprintf(stdout, "CHART apps.processes '' 'Apps Processes' 'processes' apps apps stacked 20004 %d\n", update_every);
1769 for (w = target_root; w ; w = w->next) {
1770 if(w->target || (!w->processes && !w->exposed)) continue;
1772 fprintf(stdout, "DIMENSION %s '' absolute 1 1 noreset\n", w->name);
1775 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);
1776 for (w = target_root; w ; w = w->next) {
1777 if(w->target || (!w->processes && !w->exposed)) continue;
1779 fprintf(stdout, "DIMENSION %s '' incremental 100 %llu noreset\n", w->name, Hertz * processors);
1782 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);
1783 for (w = target_root; w ; w = w->next) {
1784 if(w->target || (!w->processes && !w->exposed)) continue;
1786 fprintf(stdout, "DIMENSION %s '' incremental 100 %llu noreset\n", w->name, Hertz * processors);
1789 fprintf(stdout, "CHART apps.major_faults '' 'Apps Major Page Faults (swaps in)' 'page faults/s' apps apps stacked 20010 %d\n", update_every);
1790 for (w = target_root; w ; w = w->next) {
1791 if(w->target || (!w->processes && !w->exposed)) continue;
1793 fprintf(stdout, "DIMENSION %s '' incremental 1 1 noreset\n", w->name);
1796 fprintf(stdout, "CHART apps.minor_faults '' 'Apps Minor Page Faults' 'page faults/s' apps none stacked 20011 %d\n", update_every);
1797 for (w = target_root; w ; w = w->next) {
1798 if(w->target || (!w->processes && !w->exposed)) continue;
1800 fprintf(stdout, "DIMENSION %s '' incremental 1 1 noreset\n", w->name);
1803 fprintf(stdout, "CHART apps.lreads '' 'Apps Disk Logical Reads' 'kilobytes/s' apps none stacked 20042 %d\n", update_every);
1804 for (w = target_root; w ; w = w->next) {
1805 if(w->target || (!w->processes && !w->exposed)) continue;
1807 fprintf(stdout, "DIMENSION %s '' incremental 1 %d noreset\n", w->name, 1024);
1810 fprintf(stdout, "CHART apps.lwrites '' 'Apps I/O Logical Writes' 'kilobytes/s' apps none stacked 20042 %d\n", update_every);
1811 for (w = target_root; w ; w = w->next) {
1812 if(w->target || (!w->processes && !w->exposed)) continue;
1814 fprintf(stdout, "DIMENSION %s '' incremental 1 %d noreset\n", w->name, 1024);
1817 fprintf(stdout, "CHART apps.preads '' 'Apps Disk Reads' 'kilobytes/s' apps apps stacked 20002 %d\n", update_every);
1818 for (w = target_root; w ; w = w->next) {
1819 if(w->target || (!w->processes && !w->exposed)) continue;
1821 fprintf(stdout, "DIMENSION %s '' incremental 1 %d noreset\n", w->name, 1024);
1824 fprintf(stdout, "CHART apps.pwrites '' 'Apps Disk Writes' 'kilobytes/s' apps apps stacked 20002 %d\n", update_every);
1825 for (w = target_root; w ; w = w->next) {
1826 if(w->target || (!w->processes && !w->exposed)) continue;
1828 fprintf(stdout, "DIMENSION %s '' incremental 1 %d noreset\n", w->name, 1024);
1831 fprintf(stdout, "CHART apps.files '' 'Apps Open Files' 'open files' apps apps stacked 20050 %d\n", update_every);
1832 for (w = target_root; w ; w = w->next) {
1833 if(w->target || (!w->processes && !w->exposed)) continue;
1835 fprintf(stdout, "DIMENSION %s '' absolute 1 1 noreset\n", w->name);
1838 fprintf(stdout, "CHART apps.sockets '' 'Apps Open Sockets' 'open sockets' apps apps stacked 20051 %d\n", update_every);
1839 for (w = target_root; w ; w = w->next) {
1840 if(w->target || (!w->processes && !w->exposed)) continue;
1842 fprintf(stdout, "DIMENSION %s '' absolute 1 1 noreset\n", w->name);
1845 fprintf(stdout, "CHART apps.pipes '' 'Apps Pipes' 'open pipes' apps none stacked 20053 %d\n", update_every);
1846 for (w = target_root; w ; w = w->next) {
1847 if(w->target || (!w->processes && !w->exposed)) continue;
1849 fprintf(stdout, "DIMENSION %s '' absolute 1 1 noreset\n", w->name);
1852 fprintf(stdout, "CHART netdata.apps_cpu '' 'Apps Plugin CPU' 'milliseconds/s' netdata netdata stacked 10000 %d\n", update_every);
1853 fprintf(stdout, "DIMENSION user '' incremental 1 %d\n", 1000);
1854 fprintf(stdout, "DIMENSION system '' incremental 1 %d\n", 1000);
1856 fprintf(stdout, "CHART netdata.apps_files '' 'Apps Plugin Files' 'files/s' netdata netdata line 10001 %d\n", update_every);
1857 fprintf(stdout, "DIMENSION files '' incremental 1 1\n");
1858 fprintf(stdout, "DIMENSION pids '' absolute 1 1\n");
1859 fprintf(stdout, "DIMENSION fds '' absolute 1 1\n");
1860 fprintf(stdout, "DIMENSION targets '' absolute 1 1\n");
1866 // ----------------------------------------------------------------------------
1867 // parse command line arguments
1869 void parse_args(int argc, char **argv)
1874 for(i = 1; i < argc; i++) {
1876 int n = atoi(argv[i]);
1883 if(strcmp("debug", argv[i]) == 0) {
1885 debug_flags = 0xffffffff;
1894 error("Cannot understand option %s", argv[i]);
1898 if(freq > 0) update_every = freq;
1899 if(!name) name = "groups";
1901 if(read_process_groups(name)) {
1902 error("Cannot read process groups %s", name);
1907 int main(int argc, char **argv)
1909 // debug_flags = D_PROCFILE;
1911 // set the name for logging
1912 program_name = "apps.plugin";
1914 info("starting...");
1916 procfile_adaptive_initial_allocation = 1;
1918 unsigned long started_t = time(NULL), current_t;
1919 Hertz = get_hertz();
1920 pid_max = get_pid_max();
1921 processors = get_processors();
1923 parse_args(argc, argv);
1925 all_pids = calloc(sizeof(struct pid_stat *), pid_max);
1927 error("Cannot allocate %lu bytes of memory.", sizeof(struct pid_stat *) * pid_max);
1928 printf("DISABLE\n");
1932 unsigned long long counter = 1;
1933 unsigned long long usec = 0, susec = 0;
1934 struct timeval last, now;
1935 gettimeofday(&last, NULL);
1937 for(;1; counter++) {
1938 if(!update_from_proc()) {
1939 error("Cannot allocate %lu bytes of memory.", sizeof(struct pid_stat *) * pid_max);
1940 printf("DISABLE\n");
1944 update_statistics();
1945 show_charts(); // this is smart enough to show only newly added apps, when needed
1948 if(debug) fprintf(stderr, "apps.plugin: done Loop No %llu\n", counter);
1951 gettimeofday(&now, NULL);
1952 usec = usecdiff(&now, &last) - susec;
1953 if(debug) fprintf(stderr, "apps.plugin: last loop took %llu usec (worked for %llu, sleeped for %llu).\n", usec + susec, usec, susec);
1955 // if the last loop took less than half the time
1956 // wait the rest of the time
1957 if(usec < (update_every * 1000000ULL / 2)) susec = (update_every * 1000000ULL) - usec;
1958 else susec = update_every * 1000000ULL / 2;
1961 bcopy(&now, &last, sizeof(struct timeval));
1963 current_t = time(NULL);
1964 if(current_t - started_t > 3600) exit(0);