3 // 1. disable RESET_OR_OVERFLOW check in charts
13 #include <sys/types.h>
17 #include <sys/resource.h>
28 #include <arpa/inet.h>
35 #define MAX_COMPARE_NAME 15
38 unsigned long long Hertz = 1;
45 unsigned long long file_counter = 0;
47 char *host_prefix = "";
48 char *config_dir = CONFIG_DIR;
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] = (uint32_t) 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 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 char filename[FILENAME_MAX + 1];
210 snprintf(filename, FILENAME_MAX, "%s/proc/stat", host_prefix);
212 ff = procfile_reopen(ff, filename, "", PROCFILE_FLAG_DEFAULT);
215 ff = procfile_readall(ff);
217 // procfile_close(ff);
222 for(i = 0; i < procfile_lines(ff); i++) {
223 if(!procfile_linewords(ff, i)) continue;
225 if(strncmp(procfile_lineword(ff, i, 0), "cpu", 3) == 0) processors++;
228 if(processors < 1) processors = 1;
230 // procfile_close(ff);
234 long get_pid_max(void) {
237 char filename[FILENAME_MAX + 1];
238 snprintf(filename, FILENAME_MAX, "%s/proc/sys/kernel/pid_max", host_prefix);
239 ff = procfile_reopen(ff, filename, "", PROCFILE_FLAG_DEFAULT);
242 ff = procfile_readall(ff);
244 // procfile_close(ff);
248 mpid = atol(procfile_lineword(ff, 0, 0));
249 if(!mpid) mpid = 32768;
251 // procfile_close(ff);
255 unsigned long long get_hertz(void)
257 unsigned long long myhz = 1;
260 if((myhz = (unsigned long long int) sysconf(_SC_CLK_TCK)) > 0) {
266 myhz = HZ; /* <asm/param.h> */
268 /* If 32-bit or big-endian (not Alpha or ia64), assume HZ is 100. */
269 hz = (sizeof(long)==sizeof(int) || htons(999)==999) ? 100UL : 1024UL;
272 error("Unknown HZ value. Assuming %llu.", myhz);
277 // ----------------------------------------------------------------------------
279 // target is the point to aggregate a process tree values
282 char compare[MAX_COMPARE_NAME + 1];
283 char id[MAX_NAME + 1];
284 char name[MAX_NAME + 1];
286 unsigned long long minflt;
287 unsigned long long cminflt;
288 unsigned long long majflt;
289 unsigned long long cmajflt;
290 unsigned long long utime;
291 unsigned long long stime;
292 unsigned long long cutime;
293 unsigned long long cstime;
294 unsigned long long num_threads;
295 unsigned long long rss;
297 unsigned long long fix_minflt;
298 unsigned long long fix_cminflt;
299 unsigned long long fix_majflt;
300 unsigned long long fix_cmajflt;
301 unsigned long long fix_utime;
302 unsigned long long fix_stime;
303 unsigned long long fix_cutime;
304 unsigned long long fix_cstime;
306 unsigned long long statm_size;
307 unsigned long long statm_resident;
308 unsigned long long statm_share;
309 unsigned long long statm_text;
310 unsigned long long statm_lib;
311 unsigned long long statm_data;
312 unsigned long long statm_dirty;
314 unsigned long long io_logical_bytes_read;
315 unsigned long long io_logical_bytes_written;
316 unsigned long long io_read_calls;
317 unsigned long long io_write_calls;
318 unsigned long long io_storage_bytes_read;
319 unsigned long long io_storage_bytes_written;
320 unsigned long long io_cancelled_write_bytes;
322 unsigned long long fix_io_logical_bytes_read;
323 unsigned long long fix_io_logical_bytes_written;
324 unsigned long long fix_io_read_calls;
325 unsigned long long fix_io_write_calls;
326 unsigned long long fix_io_storage_bytes_read;
327 unsigned long long fix_io_storage_bytes_written;
328 unsigned long long fix_io_cancelled_write_bytes;
331 unsigned long long openfiles;
332 unsigned long long openpipes;
333 unsigned long long opensockets;
334 unsigned long long openinotifies;
335 unsigned long long openeventfds;
336 unsigned long long opentimerfds;
337 unsigned long long opensignalfds;
338 unsigned long long openeventpolls;
339 unsigned long long openother;
341 unsigned long processes; // how many processes have been merged to this
342 int exposed; // if set, we have sent this to netdata
343 int hidden; // if set, we set the hidden flag on the dimension
346 struct target *target; // the one that will be reported to netdata
348 } *target_root = NULL, *default_target = NULL;
352 // find or create a target
353 // there are targets that are just agregated to other target (the second argument)
354 struct target *get_target(const char *id, struct target *target)
356 const char *nid = id;
357 if(nid[0] == '-') nid++;
360 for(w = target_root ; w ; w = w->next)
361 if(strncmp(nid, w->id, MAX_NAME) == 0) return w;
363 w = calloc(sizeof(struct target), 1);
365 error("Cannot allocate %lu bytes of memory", (unsigned long)sizeof(struct target));
369 strncpy(w->id, nid, MAX_NAME);
370 strncpy(w->name, nid, MAX_NAME);
371 strncpy(w->compare, nid, MAX_COMPARE_NAME);
372 if(id[0] == '-') w->hidden = 1;
376 w->next = target_root;
379 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:"");
384 // read the process groups file
385 int read_process_groups(const char *name)
388 char filename[FILENAME_MAX + 1];
390 snprintf(filename, FILENAME_MAX, "%s/apps_%s.conf", config_dir, name);
392 if(debug) fprintf(stderr, "apps.plugin: process groups file: '%s'\n", filename);
393 FILE *fp = fopen(filename, "r");
395 error("Cannot open file '%s'", filename);
400 while(fgets(buffer, 4096, fp) != NULL) {
401 int whidden = 0, wdebug = 0;
404 // if(debug) fprintf(stderr, "apps.plugin: \tread %s\n", buffer);
406 char *s = buffer, *t, *p;
408 if(!s || !*s || *s == '#') continue;
410 if(debug) fprintf(stderr, "apps.plugin: \tread %s\n", s);
415 if(!t || !*t) continue;
437 if(debug) fprintf(stderr, "apps.plugin: \t\ttarget %s\n", t);
439 struct target *w = NULL;
442 char buffer[4097] = "";
446 while((p = strsep(&s, " "))) {
448 if(!p || !*p) continue;
450 strncpy(&buffer[blen], p, 4096 - blen);
451 blen = strlen(buffer);
453 while(buffer[blen - 1] == '\\') {
454 buffer[blen - 1] = ' ';
456 if((p = strsep(&s, " ")))
459 if(!p || !*p) p = " ";
460 strncpy(&buffer[blen], p, 4096 - blen);
461 blen = strlen(buffer);
464 struct target *n = get_target(buffer, w);
475 if(w) strncpy(w->name, t, MAX_NAME);
476 if(!count) error("The line %ld on file '%s', for group '%s' does not state any process names.", line, filename, t);
480 default_target = get_target("+p!o@w#e$i^r&7*5(-i)l-o_", NULL); // match nothing
481 strncpy(default_target->name, "other", MAX_NAME);
487 // ----------------------------------------------------------------------------
488 // data to store for each pid
493 char comm[MAX_COMPARE_NAME + 1];
501 unsigned long long minflt;
502 unsigned long long cminflt;
503 unsigned long long majflt;
504 unsigned long long cmajflt;
505 unsigned long long utime;
506 unsigned long long stime;
507 unsigned long long cutime;
508 unsigned long long cstime;
512 // int64_t itrealvalue;
513 // unsigned long long starttime;
514 // unsigned long long vsize;
515 unsigned long long rss;
516 // unsigned long long rsslim;
517 // unsigned long long starcode;
518 // unsigned long long endcode;
519 // unsigned long long startstack;
520 // unsigned long long kstkesp;
521 // unsigned long long kstkeip;
524 // uint64_t sigignore;
525 // uint64_t sigcatch;
529 // int32_t exit_signal;
530 // int32_t processor;
531 // uint32_t rt_priority;
533 // unsigned long long delayacct_blkio_ticks;
534 // uint64_t guest_time;
535 // int64_t cguest_time;
537 unsigned long long statm_size;
538 unsigned long long statm_resident;
539 unsigned long long statm_share;
540 unsigned long long statm_text;
541 unsigned long long statm_lib;
542 unsigned long long statm_data;
543 unsigned long long statm_dirty;
545 unsigned long long io_logical_bytes_read;
546 unsigned long long io_logical_bytes_written;
547 unsigned long long io_read_calls;
548 unsigned long long io_write_calls;
549 unsigned long long io_storage_bytes_read;
550 unsigned long long io_storage_bytes_written;
551 unsigned long long io_cancelled_write_bytes;
553 #ifdef INCLUDE_CHILDS
554 unsigned long long old_utime;
555 unsigned long long old_stime;
556 unsigned long long old_minflt;
557 unsigned long long old_majflt;
559 unsigned long long old_cutime;
560 unsigned long long old_cstime;
561 unsigned long long old_cminflt;
562 unsigned long long old_cmajflt;
564 unsigned long long fix_cutime;
565 unsigned long long fix_cstime;
566 unsigned long long fix_cminflt;
567 unsigned long long fix_cmajflt;
569 unsigned long long diff_cutime;
570 unsigned long long diff_cstime;
571 unsigned long long diff_cminflt;
572 unsigned long long diff_cmajflt;
575 int *fds; // array of fds it uses
576 int fds_size; // the size of the fds array
578 int childs; // number of processes directly referencing this
579 int updated; // 1 when update
580 int merged; // 1 when it has been merged to its parent
582 struct target *target;
583 struct pid_stat *parent;
584 struct pid_stat *prev;
585 struct pid_stat *next;
586 } *root_of_pids = NULL, **all_pids;
588 long all_pids_count = 0;
590 struct pid_stat *get_pid_entry(pid_t pid)
593 all_pids[pid]->new_entry = 0;
594 return all_pids[pid];
597 all_pids[pid] = calloc(sizeof(struct pid_stat), 1);
599 error("Cannot allocate %lu bytes of memory", (unsigned long)sizeof(struct pid_stat));
603 all_pids[pid]->fds = calloc(sizeof(int), 100);
604 if(!all_pids[pid]->fds)
605 error("Cannot allocate %ld bytes of memory", (unsigned long)(sizeof(int) * 100));
606 else all_pids[pid]->fds_size = 100;
608 if(root_of_pids) root_of_pids->prev = all_pids[pid];
609 all_pids[pid]->next = root_of_pids;
610 root_of_pids = all_pids[pid];
612 all_pids[pid]->pid = pid;
613 all_pids[pid]->new_entry = 1;
615 return all_pids[pid];
618 void del_pid_entry(pid_t pid)
620 if(!all_pids[pid]) return;
622 if(debug) fprintf(stderr, "apps.plugin: process %d %s exited, deleting it.\n", pid, all_pids[pid]->comm);
624 if(root_of_pids == all_pids[pid]) root_of_pids = all_pids[pid]->next;
625 if(all_pids[pid]->next) all_pids[pid]->next->prev = all_pids[pid]->prev;
626 if(all_pids[pid]->prev) all_pids[pid]->prev->next = all_pids[pid]->next;
628 if(all_pids[pid]->fds) free(all_pids[pid]->fds);
630 all_pids[pid] = NULL;
634 // ----------------------------------------------------------------------------
635 // update pids from proc
637 int read_proc_pid_stat(struct pid_stat *p) {
638 char filename[FILENAME_MAX + 1];
640 snprintf(filename, FILENAME_MAX, "%s/proc/%d/stat", host_prefix, p->pid);
642 ff = procfile_reopen(ff, filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
645 ff = procfile_readall(ff);
647 // procfile_close(ff);
654 p->comm[MAX_COMPARE_NAME] = '\0';
657 char *s = procfile_lineword(ff, 0, 1);
659 size_t len = strlen(s);
661 while(len && s[len - 1] != ')') {
662 if(blen < MAX_COMPARE_NAME) {
663 strncpy(&p->comm[blen], s, MAX_COMPARE_NAME - blen);
664 blen = strlen(p->comm);
668 s = procfile_lineword(ff, 0, 1+i);
671 if(len && s[len - 1] == ')') s[len - 1] = '\0';
672 if(blen < MAX_COMPARE_NAME)
673 strncpy(&p->comm[blen], s, MAX_COMPARE_NAME - blen);
675 // p->pid = atol(procfile_lineword(ff, 0, 0+i));
677 // p->state = *(procfile_lineword(ff, 0, 2+i));
678 p->ppid = (int32_t) atol(procfile_lineword(ff, 0, 3 + i));
679 // p->pgrp = atol(procfile_lineword(ff, 0, 4+i));
680 // p->session = atol(procfile_lineword(ff, 0, 5+i));
681 // p->tty_nr = atol(procfile_lineword(ff, 0, 6+i));
682 // p->tpgid = atol(procfile_lineword(ff, 0, 7+i));
683 // p->flags = strtoull(procfile_lineword(ff, 0, 8+i), NULL, 10);
684 p->minflt = strtoull(procfile_lineword(ff, 0, 9+i), NULL, 10);
685 p->cminflt = strtoull(procfile_lineword(ff, 0, 10+i), NULL, 10);
686 p->majflt = strtoull(procfile_lineword(ff, 0, 11+i), NULL, 10);
687 p->cmajflt = strtoull(procfile_lineword(ff, 0, 12+i), NULL, 10);
688 p->utime = strtoull(procfile_lineword(ff, 0, 13+i), NULL, 10);
689 p->stime = strtoull(procfile_lineword(ff, 0, 14+i), NULL, 10);
690 p->cutime = strtoull(procfile_lineword(ff, 0, 15+i), NULL, 10);
691 p->cstime = strtoull(procfile_lineword(ff, 0, 16+i), NULL, 10);
692 // p->priority = strtoull(procfile_lineword(ff, 0, 17+i), NULL, 10);
693 // p->nice = strtoull(procfile_lineword(ff, 0, 18+i), NULL, 10);
694 p->num_threads = (int32_t) atol(procfile_lineword(ff, 0, 19 + i));
695 // p->itrealvalue = strtoull(procfile_lineword(ff, 0, 20+i), NULL, 10);
696 // p->starttime = strtoull(procfile_lineword(ff, 0, 21+i), NULL, 10);
697 // p->vsize = strtoull(procfile_lineword(ff, 0, 22+i), NULL, 10);
698 p->rss = strtoull(procfile_lineword(ff, 0, 23+i), NULL, 10);
699 // p->rsslim = strtoull(procfile_lineword(ff, 0, 24+i), NULL, 10);
700 // p->starcode = strtoull(procfile_lineword(ff, 0, 25+i), NULL, 10);
701 // p->endcode = strtoull(procfile_lineword(ff, 0, 26+i), NULL, 10);
702 // p->startstack = strtoull(procfile_lineword(ff, 0, 27+i), NULL, 10);
703 // p->kstkesp = strtoull(procfile_lineword(ff, 0, 28+i), NULL, 10);
704 // p->kstkeip = strtoull(procfile_lineword(ff, 0, 29+i), NULL, 10);
705 // p->signal = strtoull(procfile_lineword(ff, 0, 30+i), NULL, 10);
706 // p->blocked = strtoull(procfile_lineword(ff, 0, 31+i), NULL, 10);
707 // p->sigignore = strtoull(procfile_lineword(ff, 0, 32+i), NULL, 10);
708 // p->sigcatch = strtoull(procfile_lineword(ff, 0, 33+i), NULL, 10);
709 // p->wchan = strtoull(procfile_lineword(ff, 0, 34+i), NULL, 10);
710 // p->nswap = strtoull(procfile_lineword(ff, 0, 35+i), NULL, 10);
711 // p->cnswap = strtoull(procfile_lineword(ff, 0, 36+i), NULL, 10);
712 // p->exit_signal = atol(procfile_lineword(ff, 0, 37+i));
713 // p->processor = atol(procfile_lineword(ff, 0, 38+i));
714 // p->rt_priority = strtoul(procfile_lineword(ff, 0, 39+i), NULL, 10);
715 // p->policy = strtoul(procfile_lineword(ff, 0, 40+i), NULL, 10);
716 // p->delayacct_blkio_ticks = strtoull(procfile_lineword(ff, 0, 41+i), NULL, 10);
717 // p->guest_time = strtoull(procfile_lineword(ff, 0, 42+i), NULL, 10);
718 // p->cguest_time = strtoull(procfile_lineword(ff, 0, 43), NULL, 10);
720 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);
722 // procfile_close(ff);
726 int read_proc_pid_statm(struct pid_stat *p) {
727 char filename[FILENAME_MAX + 1];
729 snprintf(filename, FILENAME_MAX, "%s/proc/%d/statm", host_prefix, p->pid);
731 ff = procfile_reopen(ff, filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
734 ff = procfile_readall(ff);
736 // procfile_close(ff);
742 p->statm_size = strtoull(procfile_lineword(ff, 0, 0), NULL, 10);
743 p->statm_resident = strtoull(procfile_lineword(ff, 0, 1), NULL, 10);
744 p->statm_share = strtoull(procfile_lineword(ff, 0, 2), NULL, 10);
745 p->statm_text = strtoull(procfile_lineword(ff, 0, 3), NULL, 10);
746 p->statm_lib = strtoull(procfile_lineword(ff, 0, 4), NULL, 10);
747 p->statm_data = strtoull(procfile_lineword(ff, 0, 5), NULL, 10);
748 p->statm_dirty = strtoull(procfile_lineword(ff, 0, 6), NULL, 10);
750 // procfile_close(ff);
754 int read_proc_pid_io(struct pid_stat *p) {
755 char filename[FILENAME_MAX + 1];
757 snprintf(filename, FILENAME_MAX, "%s/proc/%d/io", host_prefix, p->pid);
759 ff = procfile_reopen(ff, filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
762 ff = procfile_readall(ff);
764 // procfile_close(ff);
770 p->io_logical_bytes_read = strtoull(procfile_lineword(ff, 0, 1), NULL, 10);
771 p->io_logical_bytes_written = strtoull(procfile_lineword(ff, 1, 1), NULL, 10);
772 p->io_read_calls = strtoull(procfile_lineword(ff, 2, 1), NULL, 10);
773 p->io_write_calls = strtoull(procfile_lineword(ff, 3, 1), NULL, 10);
774 p->io_storage_bytes_read = strtoull(procfile_lineword(ff, 4, 1), NULL, 10);
775 p->io_storage_bytes_written = strtoull(procfile_lineword(ff, 5, 1), NULL, 10);
776 p->io_cancelled_write_bytes = strtoull(procfile_lineword(ff, 6, 1), NULL, 10);
778 // procfile_close(ff);
783 // ----------------------------------------------------------------------------
785 #ifdef INCLUDE_CHILDS
786 // print a tree view of all processes
787 int walk_down(pid_t pid, int level) {
788 struct pid_stat *p = NULL;
792 for(i = 0; i < level; i++) b[i] = '\t';
797 for(p = root_of_pids; p ; p = p->next) {
799 ret += walk_down(p->pid, level+1);
805 if(!p->updated) ret += 1;
806 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"
807 , b, p->comm, p->pid, p->updated?"OK":"KILLED", p->target->name, p->childs
808 , p->utime, p->utime - p->old_utime
809 , p->stime, p->stime - p->old_stime
810 , p->cutime, p->cutime - p->old_cutime
811 , p->cstime, p->cstime - p->old_cstime
812 , p->minflt, p->minflt - p->old_minflt
813 , p->majflt, p->majflt - p->old_majflt
814 , p->cminflt, p->cminflt - p->old_cminflt
815 , p->cmajflt, p->cmajflt - p->old_cmajflt
824 // ----------------------------------------------------------------------------
826 // this is used to keep a global list of all open files of the system
827 // it is needed in order to figure out the unique files a process tree has open
829 #define FILE_DESCRIPTORS_INCREASE_STEP 100
831 struct file_descriptor {
841 int all_files_len = 0;
842 int all_files_size = 0;
844 int file_descriptor_compare(void* a, void* b) {
845 if(((struct file_descriptor *)a)->magic != 0x0BADCAFE || ((struct file_descriptor *)b)->magic != 0x0BADCAFE)
846 error("Corrupted index data detected. Please report this.");
848 if(((struct file_descriptor *)a)->hash < ((struct file_descriptor *)b)->hash)
850 else if(((struct file_descriptor *)a)->hash > ((struct file_descriptor *)b)->hash)
852 else return strcmp(((struct file_descriptor *)a)->name, ((struct file_descriptor *)b)->name);
854 int file_descriptor_iterator(avl *a) { if(a) {}; return 0; }
856 avl_tree all_files_index = {
858 file_descriptor_compare,
859 #ifdef AVL_LOCK_WITH_MUTEX
860 PTHREAD_MUTEX_INITIALIZER
862 PTHREAD_RWLOCK_INITIALIZER
866 static struct file_descriptor *file_descriptor_find(const char *name, uint32_t hash) {
867 struct file_descriptor *result = NULL, tmp;
868 tmp.hash = (hash)?hash:simple_hash(name);
872 tmp.magic = 0x0BADCAFE;
874 avl_search(&all_files_index, (avl *)&tmp, file_descriptor_iterator, (avl **)&result);
878 #define file_descriptor_add(fd) avl_insert(&all_files_index, (avl *)(fd))
879 #define file_descriptor_remove(fd) avl_remove(&all_files_index, (avl *)(fd))
881 #define FILETYPE_OTHER 0
882 #define FILETYPE_FILE 1
883 #define FILETYPE_PIPE 2
884 #define FILETYPE_SOCKET 3
885 #define FILETYPE_INOTIFY 4
886 #define FILETYPE_EVENTFD 5
887 #define FILETYPE_EVENTPOLL 6
888 #define FILETYPE_TIMERFD 7
889 #define FILETYPE_SIGNALFD 8
891 void file_descriptor_not_used(int id)
893 if(id > 0 && id < all_files_size) {
894 if(all_files[id].magic != 0x0BADCAFE) {
895 error("Ignoring request to remove empty file id %d.", id);
899 if(debug) fprintf(stderr, "apps.plugin: decreasing slot %d (count = %d).\n", id, all_files[id].count);
901 if(all_files[id].count > 0) {
902 all_files[id].count--;
904 if(!all_files[id].count) {
905 if(debug) fprintf(stderr, "apps.plugin: >> slot %d is empty.\n", id);
906 file_descriptor_remove(&all_files[id]);
907 all_files[id].magic = 0x00000000;
912 error("Request to decrease counter of fd %d (%s), while the use counter is 0", id, all_files[id].name);
914 else error("Request to decrease counter of fd %d, which is outside the array size (1 to %d)", id, all_files_size);
917 int file_descriptor_find_or_add(const char *name)
919 static int last_pos = 0;
920 uint32_t hash = simple_hash(name);
922 if(debug) fprintf(stderr, "apps.plugin: adding or finding name '%s' with hash %u\n", name, hash);
924 struct file_descriptor *fd = file_descriptor_find(name, hash);
927 if(debug) fprintf(stderr, "apps.plugin: >> found on slot %d\n", fd->pos);
933 // check we have enough memory to add it
934 if(!all_files || all_files_len == all_files_size) {
935 void *old = all_files;
938 // there is no empty slot
939 if(debug) fprintf(stderr, "apps.plugin: extending fd array to %d entries\n", all_files_size + FILE_DESCRIPTORS_INCREASE_STEP);
940 all_files = realloc(all_files, (all_files_size + FILE_DESCRIPTORS_INCREASE_STEP) * sizeof(struct file_descriptor));
942 // if the address changed, we have to rebuild the index
943 // since all pointers are now invalid
944 if(old && old != (void *)all_files) {
945 if(debug) fprintf(stderr, "apps.plugin: >> re-indexing.\n");
946 all_files_index.root = NULL;
947 for(i = 0; i < all_files_size; i++) {
948 if(!all_files[i].count) continue;
949 file_descriptor_add(&all_files[i]);
951 if(debug) fprintf(stderr, "apps.plugin: >> re-indexing done.\n");
954 for(i = all_files_size; i < (all_files_size + FILE_DESCRIPTORS_INCREASE_STEP); i++) {
955 all_files[i].count = 0;
956 all_files[i].name = NULL;
957 all_files[i].magic = 0x00000000;
958 all_files[i].pos = i;
961 if(!all_files_size) all_files_len = 1;
962 all_files_size += FILE_DESCRIPTORS_INCREASE_STEP;
965 if(debug) fprintf(stderr, "apps.plugin: >> searching for empty slot.\n");
967 // search for an empty slot
969 for(i = 0, c = last_pos ; i < all_files_size ; i++, c++) {
970 if(c >= all_files_size) c = 0;
973 if(!all_files[c].count) {
974 if(debug) fprintf(stderr, "apps.plugin: >> Examining slot %d.\n", c);
976 if(all_files[c].magic == 0x0BADCAFE && all_files[c].name && file_descriptor_find(all_files[c].name, all_files[c].hash))
977 error("fd on position %d is not cleared properly. It still has %s in it.\n", c, all_files[c].name);
979 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);
980 if(all_files[c].name) free((void *)all_files[c].name);
981 all_files[c].name = NULL;
986 if(i == all_files_size) {
987 fatal("We should find an empty slot, but there isn't any");
990 if(debug) fprintf(stderr, "apps.plugin: >> updating slot %d.\n", c);
994 // else we have an empty slot in 'c'
997 if(name[0] == '/') type = FILETYPE_FILE;
998 else if(strncmp(name, "pipe:", 5) == 0) type = FILETYPE_PIPE;
999 else if(strncmp(name, "socket:", 7) == 0) type = FILETYPE_SOCKET;
1000 else if(strcmp(name, "anon_inode:inotify") == 0 || strcmp(name, "inotify") == 0) type = FILETYPE_INOTIFY;
1001 else if(strcmp(name, "anon_inode:[eventfd]") == 0) type = FILETYPE_EVENTFD;
1002 else if(strcmp(name, "anon_inode:[eventpoll]") == 0) type = FILETYPE_EVENTPOLL;
1003 else if(strcmp(name, "anon_inode:[timerfd]") == 0) type = FILETYPE_TIMERFD;
1004 else if(strcmp(name, "anon_inode:[signalfd]") == 0) type = FILETYPE_SIGNALFD;
1005 else if(strncmp(name, "anon_inode:", 11) == 0) {
1006 if(debug) fprintf(stderr, "apps.plugin: FIXME: unknown anonymous inode: %s\n", name);
1007 type = FILETYPE_OTHER;
1010 if(debug) fprintf(stderr, "apps.plugin: FIXME: cannot understand linkname: %s\n", name);
1011 type = FILETYPE_OTHER;
1014 all_files[c].name = strdup(name);
1015 all_files[c].hash = hash;
1016 all_files[c].type = type;
1017 all_files[c].pos = c;
1018 all_files[c].count = 1;
1019 all_files[c].magic = 0x0BADCAFE;
1021 file_descriptor_add(&all_files[c]);
1023 if(debug) fprintf(stderr, "apps.plugin: using fd position %d (name: %s)\n", c, all_files[c].name);
1029 // 1. read all files in /proc
1030 // 2. for each numeric directory:
1031 // i. read /proc/pid/stat
1032 // ii. read /proc/pid/statm
1033 // iii. read /proc/pid/io (requires root access)
1034 // iii. read the entries in directory /proc/pid/fd (requires root access)
1036 // a. find or create a struct file_descriptor
1037 // b. cleanup any old/unused file_descriptors
1039 // after all these, some pids may be linked to targets, while others may not
1041 // in case of errors, only 1 every 1000 errors is printed
1042 // to avoid filling up all disk space
1043 // if debug is enabled, all errors are printed
1045 int update_from_proc(void)
1047 static long count_errors = 0;
1049 char filename[FILENAME_MAX+1];
1050 char dirname[FILENAME_MAX + 1];
1052 snprintf(dirname, FILENAME_MAX, "%s/proc", host_prefix);
1053 DIR *dir = opendir(dirname);
1056 struct dirent *file = NULL;
1057 struct pid_stat *p = NULL;
1059 // mark them all as un-updated
1061 for(p = root_of_pids; p ; p = p->next) {
1070 while((file = readdir(dir))) {
1071 char *endptr = file->d_name;
1072 pid_t pid = (pid_t) strtoul(file->d_name, &endptr, 10);
1073 if(pid <= 0 || pid > pid_max || endptr == file->d_name || *endptr != '\0') continue;
1075 p = get_pid_entry(pid);
1078 // --------------------------------------------------------------------
1081 if(read_proc_pid_stat(p)) {
1082 if(!count_errors++ || debug || (p->target && p->target->debug))
1083 error("Cannot process %s/proc/%d/stat", host_prefix, pid);
1087 if(p->ppid < 0 || p->ppid > pid_max) p->ppid = 0;
1090 // --------------------------------------------------------------------
1091 // /proc/<pid>/statm
1093 if(read_proc_pid_statm(p)) {
1094 if(!count_errors++ || debug || (p->target && p->target->debug))
1095 error("Cannot process %s/proc/%d/statm", host_prefix, pid);
1101 // --------------------------------------------------------------------
1104 if(read_proc_pid_io(p)) {
1105 if(!count_errors++ || debug || (p->target && p->target->debug))
1106 error("Cannot process %s/proc/%d/io", host_prefix, pid);
1108 // on systems without /proc/X/io
1109 // allow proceeding without I/O information
1113 // --------------------------------------------------------------------
1116 // check if it is target
1117 // we do this only once, the first time this pid is loaded
1119 if(debug) fprintf(stderr, "apps.plugin: \tJust added %s\n", p->comm);
1122 for(w = target_root; w ; w = w->next) {
1123 // if(debug || (p->target && p->target->debug)) fprintf(stderr, "apps.plugin: \t\tcomparing '%s' with '%s'\n", w->compare, p->comm);
1125 if(strcmp(w->compare, p->comm) == 0) {
1126 if(w->target) p->target = w->target;
1129 if(debug || (p->target && p->target->debug)) fprintf(stderr, "apps.plugin: \t\t%s linked to target %s\n", p->comm, p->target->name);
1134 // --------------------------------------------------------------------
1137 snprintf(filename, FILENAME_MAX, "%s/proc/%s/fd", host_prefix, file->d_name);
1138 DIR *fds = opendir(filename);
1142 char fdname[FILENAME_MAX + 1];
1143 char linkname[FILENAME_MAX + 1];
1145 // make the array negative
1146 for(c = 0 ; c < p->fds_size ; c++) p->fds[c] = -p->fds[c];
1148 while((de = readdir(fds))) {
1149 if(strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) continue;
1151 // check if the fds array is small
1152 int fdid = atoi(de->d_name);
1153 if(fdid < 0) continue;
1154 if(fdid >= p->fds_size) {
1155 // it is small, extend it
1156 if(debug) fprintf(stderr, "apps.plugin: extending fd memory slots for %s from %d to %d\n", p->comm, p->fds_size, fdid + 100);
1157 p->fds = realloc(p->fds, (fdid + 100) * sizeof(int));
1159 error("Cannot re-allocate fds for %s", p->comm);
1163 // and initialize it
1164 for(c = p->fds_size ; c < (fdid + 100) ; c++) p->fds[c] = 0;
1165 p->fds_size = fdid + 100;
1168 if(p->fds[fdid] == 0) {
1169 // we don't know this fd, get it
1171 sprintf(fdname, "%s/proc/%s/fd/%s", host_prefix, file->d_name, de->d_name);
1172 ssize_t l = readlink(fdname, linkname, FILENAME_MAX);
1174 if(debug || (p->target && p->target->debug)) {
1175 if(!count_errors++ || debug || (p->target && p->target->debug))
1176 error("Cannot read link %s", fdname);
1183 // if another process already has this, we will get
1185 p->fds[fdid] = file_descriptor_find_or_add(linkname);
1188 // else make it positive again, we need it
1189 // of course, the actual file may have changed, but we don't care so much
1190 // FIXME: we could compare the inode as returned by readdir direct structure
1191 else p->fds[fdid] = -p->fds[fdid];
1195 // remove all the negative file descriptors
1196 for(c = 0 ; c < p->fds_size ; c++) if(p->fds[c] < 0) {
1197 file_descriptor_not_used(-p->fds[c]);
1202 // --------------------------------------------------------------------
1205 // mark it as updated
1208 if(count_errors > 1000) {
1209 error("%ld more errors encountered\n", count_errors - 1);
1219 // ----------------------------------------------------------------------------
1220 // update statistics on the targets
1222 // 1. link all childs to their parents
1223 // 2. go from bottom to top, marking as merged all childs to their parents
1224 // this step links all parents without a target to the child target, if any
1225 // 3. link all top level processes (the ones not merged) to the default target
1226 // 4. go from top to bottom, linking all childs without a target, to their parent target
1227 // after this step, all processes have a target
1228 // [5. for each killed pid (updated = 0), remove its usage from its target]
1229 // 6. zero all targets
1230 // 7. concentrate all values on the targets
1231 // 8. remove all killed processes
1232 // 9. find the unique file count for each target
1234 void update_statistics(void)
1237 struct pid_stat *p = NULL;
1239 // link all parents and update childs count
1240 for(p = root_of_pids; p ; p = p->next) {
1241 if(p->ppid > 0 && p->ppid <= pid_max && all_pids[p->ppid]) {
1242 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);
1244 p->parent = all_pids[p->ppid];
1245 p->parent->childs++;
1247 else if(p->ppid != 0) error("pid %d %s states parent %d, but the later does not exist.", p->pid, p->comm, p->ppid);
1250 // find all the procs with 0 childs and merge them to their parents
1251 // repeat, until nothing more can be done.
1255 for(p = root_of_pids; p ; p = p->next) {
1256 // if this process does not have any childs, and
1257 // is not already merged, and
1258 // its parent has childs waiting to be merged, and
1259 // 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
1260 // and its parent is not init
1261 // then... merge them!
1262 if(!p->childs && !p->merged && p->parent && p->parent->childs && (p->target == p->parent->target || !p->parent->target || !p->target) && p->ppid != 1) {
1263 p->parent->childs--;
1266 // the parent inherits the child's target, if it does not have a target itself
1267 if(p->target && !p->parent->target) {
1268 p->parent->target = p->target;
1269 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);
1275 if(debug) fprintf(stderr, "apps.plugin: merged %d processes\n", found);
1278 // give a default target on all top level processes
1279 // init goes always to default target
1280 if(all_pids[1]) all_pids[1]->target = default_target;
1282 for(p = root_of_pids; p ; p = p->next) {
1283 // if the process is not merged itself
1284 // then is is a top level process
1285 if(!p->merged && !p->target) p->target = default_target;
1287 #ifdef INCLUDE_CHILDS
1288 // by the way, update the diffs
1289 // will be used later for substracting killed process times
1290 p->diff_cutime = p->utime - p->cutime;
1291 p->diff_cstime = p->stime - p->cstime;
1292 p->diff_cminflt = p->minflt - p->cminflt;
1293 p->diff_cmajflt = p->majflt - p->cmajflt;
1297 // give a target to all merged child processes
1301 for(p = root_of_pids; p ; p = p->next) {
1302 if(!p->target && p->merged && p->parent && p->parent->target) {
1303 p->target = p->parent->target;
1309 #ifdef INCLUDE_CHILDS
1310 // for each killed process, remove its values from the parents
1311 // sums (we had already added them in a previous loop)
1312 for(p = root_of_pids; p ; p = p->next) {
1313 if(p->updated) continue;
1315 if(debug) fprintf(stderr, "apps.plugin: UNMERGING %d %s\n", p->pid, p->comm);
1317 unsigned long long diff_utime = p->utime + p->cutime + p->fix_cutime;
1318 unsigned long long diff_stime = p->stime + p->cstime + p->fix_cstime;
1319 unsigned long long diff_minflt = p->minflt + p->cminflt + p->fix_cminflt;
1320 unsigned long long diff_majflt = p->majflt + p->cmajflt + p->fix_cmajflt;
1322 struct pid_stat *t = p;
1323 while((t = t->parent)) {
1324 if(!t->updated) continue;
1326 unsigned long long x;
1327 if(diff_utime && t->diff_cutime) {
1328 x = (t->diff_cutime < diff_utime)?t->diff_cutime:diff_utime;
1330 t->diff_cutime -= x;
1332 if(debug) fprintf(stderr, "apps.plugin: \t cutime %llu from %d %s %s\n", x, t->pid, t->comm, t->target->name);
1334 if(diff_stime && t->diff_cstime) {
1335 x = (t->diff_cstime < diff_stime)?t->diff_cstime:diff_stime;
1337 t->diff_cstime -= x;
1339 if(debug) fprintf(stderr, "apps.plugin: \t cstime %llu from %d %s %s\n", x, t->pid, t->comm, t->target->name);
1341 if(diff_minflt && t->diff_cminflt) {
1342 x = (t->diff_cminflt < diff_minflt)?t->diff_cminflt:diff_minflt;
1344 t->diff_cminflt -= x;
1345 t->fix_cminflt += x;
1346 if(debug) fprintf(stderr, "apps.plugin: \t cminflt %llu from %d %s %s\n", x, t->pid, t->comm, t->target->name);
1348 if(diff_majflt && t->diff_cmajflt) {
1349 x = (t->diff_cmajflt < diff_majflt)?t->diff_cmajflt:diff_majflt;
1351 t->diff_cmajflt -= x;
1352 t->fix_cmajflt += x;
1353 if(debug) fprintf(stderr, "apps.plugin: \t cmajflt %llu from %d %s %s\n", x, t->pid, t->comm, t->target->name);
1357 if(diff_utime) error("Cannot fix up utime %llu", diff_utime);
1358 if(diff_stime) error("Cannot fix up stime %llu", diff_stime);
1359 if(diff_minflt) error("Cannot fix up minflt %llu", diff_minflt);
1360 if(diff_majflt) error("Cannot fix up majflt %llu", diff_majflt);
1364 // zero all the targets
1367 for (w = target_root; w ; w = w->next) {
1370 w->fds = calloc(sizeof(int), (size_t) all_files_size);
1372 error("Cannot allocate memory for fds in %s", w->name);
1387 w->statm_resident = 0;
1394 w->io_logical_bytes_read = 0;
1395 w->io_logical_bytes_written = 0;
1396 w->io_read_calls = 0;
1397 w->io_write_calls = 0;
1398 w->io_storage_bytes_read = 0;
1399 w->io_storage_bytes_written = 0;
1400 w->io_cancelled_write_bytes = 0;
1403 #ifdef INCLUDE_CHILDS
1404 if(debug) walk_down(0, 1);
1407 // concentrate everything on the targets
1408 for(p = root_of_pids; p ; p = p->next) {
1410 error("pid %d %s was left without a target!", p->pid, p->comm);
1415 p->target->cutime += p->cutime; // - p->fix_cutime;
1416 p->target->cstime += p->cstime; // - p->fix_cstime;
1417 p->target->cminflt += p->cminflt; // - p->fix_cminflt;
1418 p->target->cmajflt += p->cmajflt; // - p->fix_cmajflt;
1420 p->target->utime += p->utime; //+ (p->pid != 1)?(p->cutime - p->fix_cutime):0;
1421 p->target->stime += p->stime; //+ (p->pid != 1)?(p->cstime - p->fix_cstime):0;
1422 p->target->minflt += p->minflt; //+ (p->pid != 1)?(p->cminflt - p->fix_cminflt):0;
1423 p->target->majflt += p->majflt; //+ (p->pid != 1)?(p->cmajflt - p->fix_cmajflt):0;
1425 //if(p->num_threads < 0)
1426 // error("Negative threads number for pid '%s' (%d): %d", p->comm, p->pid, p->num_threads);
1428 //if(p->num_threads > 10000)
1429 // error("Excessive threads number for pid '%s' (%d): %d", p->comm, p->pid, p->num_threads);
1431 p->target->num_threads += p->num_threads;
1432 p->target->rss += p->rss;
1434 p->target->statm_size += p->statm_size;
1435 p->target->statm_resident += p->statm_resident;
1436 p->target->statm_share += p->statm_share;
1437 p->target->statm_text += p->statm_text;
1438 p->target->statm_lib += p->statm_lib;
1439 p->target->statm_data += p->statm_data;
1440 p->target->statm_dirty += p->statm_dirty;
1442 p->target->io_logical_bytes_read += p->io_logical_bytes_read;
1443 p->target->io_logical_bytes_written += p->io_logical_bytes_written;
1444 p->target->io_read_calls += p->io_read_calls;
1445 p->target->io_write_calls += p->io_write_calls;
1446 p->target->io_storage_bytes_read += p->io_storage_bytes_read;
1447 p->target->io_storage_bytes_written += p->io_storage_bytes_written;
1448 p->target->io_cancelled_write_bytes += p->io_cancelled_write_bytes;
1450 p->target->processes++;
1452 for(c = 0; c < p->fds_size ;c++) {
1453 if(p->fds[c] == 0) continue;
1454 if(p->fds[c] < all_files_size) {
1455 if(p->target->fds) p->target->fds[p->fds[c]]++;
1458 error("Invalid fd number %d", p->fds[c]);
1461 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);
1463 /* 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);
1464 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);
1465 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);
1466 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);
1467 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);
1468 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);
1469 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);
1470 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);
1472 #ifdef INCLUDE_CHILDS
1473 p->old_utime = p->utime;
1474 p->old_cutime = p->cutime;
1475 p->old_stime = p->stime;
1476 p->old_cstime = p->cstime;
1477 p->old_minflt = p->minflt;
1478 p->old_majflt = p->majflt;
1479 p->old_cminflt = p->cminflt;
1480 p->old_cmajflt = p->cmajflt;
1484 // since the process has exited, the user
1485 // will see a drop in our charts, because the incremental
1486 // values of this process will not be there
1488 // add them to the fix_* values and they will be added to
1489 // the reported values, so that the report goes steady
1490 p->target->fix_minflt += p->minflt;
1491 p->target->fix_majflt += p->majflt;
1492 p->target->fix_utime += p->utime;
1493 p->target->fix_stime += p->stime;
1494 p->target->fix_cminflt += p->cminflt;
1495 p->target->fix_cmajflt += p->cmajflt;
1496 p->target->fix_cutime += p->cutime;
1497 p->target->fix_cstime += p->cstime;
1499 p->target->fix_io_logical_bytes_read += p->io_logical_bytes_read;
1500 p->target->fix_io_logical_bytes_written += p->io_logical_bytes_written;
1501 p->target->fix_io_read_calls += p->io_read_calls;
1502 p->target->fix_io_write_calls += p->io_write_calls;
1503 p->target->fix_io_storage_bytes_read += p->io_storage_bytes_read;
1504 p->target->fix_io_storage_bytes_written += p->io_storage_bytes_written;
1505 p->target->fix_io_cancelled_write_bytes += p->io_cancelled_write_bytes;
1509 // fprintf(stderr, "\n");
1510 // cleanup all un-updated processed (exited, killed, etc)
1511 for(p = root_of_pids; p ;) {
1513 // 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);
1515 for(c = 0 ; c < p->fds_size ; c++) if(p->fds[c] > 0) {
1516 file_descriptor_not_used(p->fds[c]);
1527 for (w = target_root; w ; w = w->next) {
1531 w->openinotifies = 0;
1532 w->openeventfds = 0;
1533 w->opentimerfds = 0;
1534 w->opensignalfds = 0;
1535 w->openeventpolls = 0;
1538 for(c = 1; c < all_files_size ;c++) {
1539 if(w->fds && w->fds[c] > 0) switch(all_files[c].type) {
1548 case FILETYPE_SOCKET:
1552 case FILETYPE_INOTIFY:
1556 case FILETYPE_EVENTFD:
1560 case FILETYPE_TIMERFD:
1564 case FILETYPE_SIGNALFD:
1568 case FILETYPE_EVENTPOLL:
1569 w->openeventpolls++;
1582 // ----------------------------------------------------------------------------
1583 // update chart dimensions
1585 void show_dimensions(void)
1587 static struct timeval last = { 0, 0 };
1588 static struct rusage me_last;
1594 unsigned long long usec;
1595 unsigned long long cpuuser;
1596 unsigned long long cpusyst;
1599 gettimeofday(&last, NULL);
1600 getrusage(RUSAGE_SELF, &me_last);
1602 // the first time, give a zero to allow
1603 // netdata calibrate to the current time
1604 // usec = update_every * 1000000ULL;
1610 gettimeofday(&now, NULL);
1611 getrusage(RUSAGE_SELF, &me);
1613 usec = usecdiff(&now, &last);
1614 cpuuser = me.ru_utime.tv_sec * 1000000ULL + me.ru_utime.tv_usec;
1615 cpusyst = me.ru_stime.tv_sec * 1000000ULL + me.ru_stime.tv_usec;
1617 bcopy(&now, &last, sizeof(struct timeval));
1618 bcopy(&me, &me_last, sizeof(struct rusage));
1621 fprintf(stdout, "BEGIN apps.cpu %llu\n", usec);
1622 for (w = target_root; w ; w = w->next) {
1623 if(w->target || (!w->processes && !w->exposed)) continue;
1625 fprintf(stdout, "SET %s = %llu\n", w->name, w->utime + w->stime + w->fix_utime + w->fix_stime);
1627 fprintf(stdout, "END\n");
1629 fprintf(stdout, "BEGIN apps.cpu_user %llu\n", usec);
1630 for (w = target_root; w ; w = w->next) {
1631 if(w->target || (!w->processes && !w->exposed)) continue;
1633 fprintf(stdout, "SET %s = %llu\n", w->name, w->utime + w->fix_utime);
1635 fprintf(stdout, "END\n");
1637 fprintf(stdout, "BEGIN apps.cpu_system %llu\n", usec);
1638 for (w = target_root; w ; w = w->next) {
1639 if(w->target || (!w->processes && !w->exposed)) continue;
1641 fprintf(stdout, "SET %s = %llu\n", w->name, w->stime + w->fix_stime);
1643 fprintf(stdout, "END\n");
1645 fprintf(stdout, "BEGIN apps.threads %llu\n", usec);
1646 for (w = target_root; w ; w = w->next) {
1647 if(w->target || (!w->processes && !w->exposed)) continue;
1649 fprintf(stdout, "SET %s = %llu\n", w->name, w->num_threads);
1651 fprintf(stdout, "END\n");
1653 fprintf(stdout, "BEGIN apps.processes %llu\n", usec);
1654 for (w = target_root; w ; w = w->next) {
1655 if(w->target || (!w->processes && !w->exposed)) continue;
1657 fprintf(stdout, "SET %s = %lu\n", w->name, w->processes);
1659 fprintf(stdout, "END\n");
1661 fprintf(stdout, "BEGIN apps.mem %llu\n", usec);
1662 for (w = target_root; w ; w = w->next) {
1663 if(w->target || (!w->processes && !w->exposed)) continue;
1665 fprintf(stdout, "SET %s = %lld\n", w->name, (long long)w->statm_resident - (long long)w->statm_share);
1667 fprintf(stdout, "END\n");
1669 fprintf(stdout, "BEGIN apps.minor_faults %llu\n", usec);
1670 for (w = target_root; w ; w = w->next) {
1671 if(w->target || (!w->processes && !w->exposed)) continue;
1673 fprintf(stdout, "SET %s = %llu\n", w->name, w->minflt + w->fix_minflt);
1675 fprintf(stdout, "END\n");
1677 fprintf(stdout, "BEGIN apps.major_faults %llu\n", usec);
1678 for (w = target_root; w ; w = w->next) {
1679 if(w->target || (!w->processes && !w->exposed)) continue;
1681 fprintf(stdout, "SET %s = %llu\n", w->name, w->majflt + w->fix_majflt);
1683 fprintf(stdout, "END\n");
1685 fprintf(stdout, "BEGIN apps.lreads %llu\n", usec);
1686 for (w = target_root; w ; w = w->next) {
1687 if(w->target || (!w->processes && !w->exposed)) continue;
1689 fprintf(stdout, "SET %s = %llu\n", w->name, w->io_logical_bytes_read);
1691 fprintf(stdout, "END\n");
1693 fprintf(stdout, "BEGIN apps.lwrites %llu\n", usec);
1694 for (w = target_root; w ; w = w->next) {
1695 if(w->target || (!w->processes && !w->exposed)) continue;
1697 fprintf(stdout, "SET %s = %llu\n", w->name, w->io_logical_bytes_written);
1699 fprintf(stdout, "END\n");
1701 fprintf(stdout, "BEGIN apps.preads %llu\n", usec);
1702 for (w = target_root; w ; w = w->next) {
1703 if(w->target || (!w->processes && !w->exposed)) continue;
1705 fprintf(stdout, "SET %s = %llu\n", w->name, w->io_storage_bytes_read);
1707 fprintf(stdout, "END\n");
1709 fprintf(stdout, "BEGIN apps.pwrites %llu\n", usec);
1710 for (w = target_root; w ; w = w->next) {
1711 if(w->target || (!w->processes && !w->exposed)) continue;
1713 fprintf(stdout, "SET %s = %llu\n", w->name, w->io_storage_bytes_written);
1715 fprintf(stdout, "END\n");
1717 fprintf(stdout, "BEGIN apps.files %llu\n", usec);
1718 for (w = target_root; w ; w = w->next) {
1719 if(w->target || (!w->processes && !w->exposed)) continue;
1721 fprintf(stdout, "SET %s = %llu\n", w->name, w->openfiles);
1723 fprintf(stdout, "END\n");
1725 fprintf(stdout, "BEGIN apps.sockets %llu\n", usec);
1726 for (w = target_root; w ; w = w->next) {
1727 if(w->target || (!w->processes && !w->exposed)) continue;
1729 fprintf(stdout, "SET %s = %llu\n", w->name, w->opensockets);
1731 fprintf(stdout, "END\n");
1733 fprintf(stdout, "BEGIN apps.pipes %llu\n", usec);
1734 for (w = target_root; w ; w = w->next) {
1735 if(w->target || (!w->processes && !w->exposed)) continue;
1737 fprintf(stdout, "SET %s = %llu\n", w->name, w->openpipes);
1739 fprintf(stdout, "END\n");
1741 fprintf(stdout, "BEGIN netdata.apps_cpu %llu\n", usec);
1742 fprintf(stdout, "SET user = %llu\n", cpuuser);
1743 fprintf(stdout, "SET system = %llu\n", cpusyst);
1744 fprintf(stdout, "END\n");
1746 fprintf(stdout, "BEGIN netdata.apps_files %llu\n", usec);
1747 fprintf(stdout, "SET files = %llu\n", file_counter);
1748 fprintf(stdout, "SET pids = %ld\n", all_pids_count);
1749 fprintf(stdout, "SET fds = %d\n", all_files_len);
1750 fprintf(stdout, "SET targets = %ld\n", targets);
1751 fprintf(stdout, "END\n");
1757 // ----------------------------------------------------------------------------
1758 // generate the charts
1760 void show_charts(void)
1763 int newly_added = 0;
1765 for(w = target_root ; w ; w = w->next)
1766 if(!w->exposed && w->processes) {
1769 if(debug || w->debug) fprintf(stderr, "apps.plugin: %s just added - regenerating charts.\n", w->name);
1772 // nothing more to show
1773 if(!newly_added) return;
1775 // we have something new to show
1776 // update the charts
1777 fprintf(stdout, "CHART apps.cpu '' 'Apps CPU Time (%ld%% = %ld core%s)' 'cpu time %%' cpu apps.cpu stacked 20001 %d\n", (processors * 100), processors, (processors>1)?"s":"", update_every);
1778 for (w = target_root; w ; w = w->next) {
1779 if(w->target || (!w->processes && !w->exposed)) continue;
1781 fprintf(stdout, "DIMENSION %s '' incremental 100 %llu %s\n", w->name, Hertz, w->hidden ? "hidden,noreset" : "noreset");
1784 fprintf(stdout, "CHART apps.mem '' 'Apps Dedicated Memory (w/o shared)' 'MB' mem apps.mem stacked 20003 %d\n", update_every);
1785 for (w = target_root; w ; w = w->next) {
1786 if(w->target || (!w->processes && !w->exposed)) continue;
1788 fprintf(stdout, "DIMENSION %s '' absolute %ld %ld noreset\n", w->name, sysconf(_SC_PAGESIZE), 1024L*1024L);
1791 fprintf(stdout, "CHART apps.threads '' 'Apps Threads' 'threads' processes apps.threads stacked 20005 %d\n", update_every);
1792 for (w = target_root; w ; w = w->next) {
1793 if(w->target || (!w->processes && !w->exposed)) continue;
1795 fprintf(stdout, "DIMENSION %s '' absolute 1 1 noreset\n", w->name);
1798 fprintf(stdout, "CHART apps.processes '' 'Apps Processes' 'processes' processes apps.processes stacked 20004 %d\n", update_every);
1799 for (w = target_root; w ; w = w->next) {
1800 if(w->target || (!w->processes && !w->exposed)) continue;
1802 fprintf(stdout, "DIMENSION %s '' absolute 1 1 noreset\n", w->name);
1805 fprintf(stdout, "CHART apps.cpu_user '' 'Apps CPU User Time (%ld%% = %ld core%s)' 'cpu time %%' cpu apps.cpu_user stacked 20020 %d\n", (processors * 100), processors, (processors>1)?"s":"", update_every);
1806 for (w = target_root; w ; w = w->next) {
1807 if(w->target || (!w->processes && !w->exposed)) continue;
1809 fprintf(stdout, "DIMENSION %s '' incremental 100 %llu noreset\n", w->name, Hertz * processors);
1812 fprintf(stdout, "CHART apps.cpu_system '' 'Apps CPU System Time (%ld%% = %ld core%s)' 'cpu time %%' cpu apps.cpu_system stacked 20021 %d\n", (processors * 100), processors, (processors>1)?"s":"", update_every);
1813 for (w = target_root; w ; w = w->next) {
1814 if(w->target || (!w->processes && !w->exposed)) continue;
1816 fprintf(stdout, "DIMENSION %s '' incremental 100 %llu noreset\n", w->name, Hertz * processors);
1819 fprintf(stdout, "CHART apps.major_faults '' 'Apps Major Page Faults (swap read)' 'page faults/s' swap apps.major_faults stacked 20010 %d\n", update_every);
1820 for (w = target_root; w ; w = w->next) {
1821 if(w->target || (!w->processes && !w->exposed)) continue;
1823 fprintf(stdout, "DIMENSION %s '' incremental 1 1 noreset\n", w->name);
1826 fprintf(stdout, "CHART apps.minor_faults '' 'Apps Minor Page Faults' 'page faults/s' mem apps.minor_faults stacked 20011 %d\n", update_every);
1827 for (w = target_root; w ; w = w->next) {
1828 if(w->target || (!w->processes && !w->exposed)) continue;
1830 fprintf(stdout, "DIMENSION %s '' incremental 1 1 noreset\n", w->name);
1833 fprintf(stdout, "CHART apps.lreads '' 'Apps Disk Logical Reads' 'kilobytes/s' disk apps.lreads stacked 20042 %d\n", update_every);
1834 for (w = target_root; w ; w = w->next) {
1835 if(w->target || (!w->processes && !w->exposed)) continue;
1837 fprintf(stdout, "DIMENSION %s '' incremental 1 %d noreset\n", w->name, 1024);
1840 fprintf(stdout, "CHART apps.lwrites '' 'Apps I/O Logical Writes' 'kilobytes/s' disk apps.lwrites stacked 20042 %d\n", update_every);
1841 for (w = target_root; w ; w = w->next) {
1842 if(w->target || (!w->processes && !w->exposed)) continue;
1844 fprintf(stdout, "DIMENSION %s '' incremental 1 %d noreset\n", w->name, 1024);
1847 fprintf(stdout, "CHART apps.preads '' 'Apps Disk Reads' 'kilobytes/s' disk apps.preads stacked 20002 %d\n", update_every);
1848 for (w = target_root; w ; w = w->next) {
1849 if(w->target || (!w->processes && !w->exposed)) continue;
1851 fprintf(stdout, "DIMENSION %s '' incremental 1 %d noreset\n", w->name, 1024);
1854 fprintf(stdout, "CHART apps.pwrites '' 'Apps Disk Writes' 'kilobytes/s' disk apps.pwrites stacked 20002 %d\n", update_every);
1855 for (w = target_root; w ; w = w->next) {
1856 if(w->target || (!w->processes && !w->exposed)) continue;
1858 fprintf(stdout, "DIMENSION %s '' incremental 1 %d noreset\n", w->name, 1024);
1861 fprintf(stdout, "CHART apps.files '' 'Apps Open Files' 'open files' disk apps.files stacked 20050 %d\n", update_every);
1862 for (w = target_root; w ; w = w->next) {
1863 if(w->target || (!w->processes && !w->exposed)) continue;
1865 fprintf(stdout, "DIMENSION %s '' absolute 1 1 noreset\n", w->name);
1868 fprintf(stdout, "CHART apps.sockets '' 'Apps Open Sockets' 'open sockets' net apps.sockets stacked 20051 %d\n", update_every);
1869 for (w = target_root; w ; w = w->next) {
1870 if(w->target || (!w->processes && !w->exposed)) continue;
1872 fprintf(stdout, "DIMENSION %s '' absolute 1 1 noreset\n", w->name);
1875 fprintf(stdout, "CHART apps.pipes '' 'Apps Pipes' 'open pipes' processes apps.pipes stacked 20053 %d\n", update_every);
1876 for (w = target_root; w ; w = w->next) {
1877 if(w->target || (!w->processes && !w->exposed)) continue;
1879 fprintf(stdout, "DIMENSION %s '' absolute 1 1 noreset\n", w->name);
1882 fprintf(stdout, "CHART netdata.apps_cpu '' 'Apps Plugin CPU' 'milliseconds/s' apps.plugin netdata.apps_cpu stacked 140000 %d\n", update_every);
1883 fprintf(stdout, "DIMENSION user '' incremental 1 %d\n", 1000);
1884 fprintf(stdout, "DIMENSION system '' incremental 1 %d\n", 1000);
1886 fprintf(stdout, "CHART netdata.apps_files '' 'Apps Plugin Files' 'files/s' apps.plugin netdata.apps_files line 140001 %d\n", update_every);
1887 fprintf(stdout, "DIMENSION files '' incremental 1 1\n");
1888 fprintf(stdout, "DIMENSION pids '' absolute 1 1\n");
1889 fprintf(stdout, "DIMENSION fds '' absolute 1 1\n");
1890 fprintf(stdout, "DIMENSION targets '' absolute 1 1\n");
1896 // ----------------------------------------------------------------------------
1897 // parse command line arguments
1899 void parse_args(int argc, char **argv)
1904 for(i = 1; i < argc; i++) {
1906 int n = atoi(argv[i]);
1913 if(strcmp("debug", argv[i]) == 0) {
1915 debug_flags = 0xffffffff;
1924 error("Cannot understand option %s", argv[i]);
1928 if(freq > 0) update_every = freq;
1929 if(!name) name = "groups";
1931 if(read_process_groups(name)) {
1932 error("Cannot read process groups %s", name);
1937 int main(int argc, char **argv)
1939 // debug_flags = D_PROCFILE;
1941 // set the name for logging
1942 program_name = "apps.plugin";
1944 host_prefix = getenv("NETDATA_HOST_PREFIX");
1945 if(host_prefix == NULL) {
1946 info("NETDATA_HOST_PREFIX is not passed from netdata");
1949 else info("Found NETDATA_HOST_PREFIX='%s'", host_prefix);
1951 config_dir = getenv("NETDATA_CONFIG_DIR");
1952 if(config_dir == NULL) {
1953 info("NETDATA_CONFIG_DIR is not passed from netdata");
1954 config_dir = CONFIG_DIR;
1956 else info("Found NETDATA_CONFIG_DIR='%s'", config_dir);
1958 info("starting...");
1960 procfile_adaptive_initial_allocation = 1;
1962 time_t started_t = time(NULL);
1964 Hertz = get_hertz();
1965 pid_max = get_pid_max();
1966 processors = get_processors();
1968 parse_args(argc, argv);
1970 all_pids = calloc(sizeof(struct pid_stat *), (size_t) pid_max);
1972 error("Cannot allocate %lu bytes of memory.", sizeof(struct pid_stat *) * pid_max);
1973 printf("DISABLE\n");
1977 unsigned long long counter = 1;
1978 unsigned long long usec = 0, susec = 0;
1979 struct timeval last, now;
1980 gettimeofday(&last, NULL);
1982 for(;1; counter++) {
1983 if(!update_from_proc()) {
1984 error("Cannot allocate %lu bytes of memory.", sizeof(struct pid_stat *) * pid_max);
1985 printf("DISABLE\n");
1989 update_statistics();
1990 show_charts(); // this is smart enough to show only newly added apps, when needed
1993 if(debug) fprintf(stderr, "apps.plugin: done Loop No %llu\n", counter);
1996 gettimeofday(&now, NULL);
1997 usec = usecdiff(&now, &last) - susec;
1998 if(debug) fprintf(stderr, "apps.plugin: last loop took %llu usec (worked for %llu, sleeped for %llu).\n", usec + susec, usec, susec);
2000 // if the last loop took less than half the time
2001 // wait the rest of the time
2002 if(usec < (update_every * 1000000ULL / 2)) susec = (update_every * 1000000ULL) - usec;
2003 else susec = update_every * 1000000ULL / 2;
2005 usleep((useconds_t) susec);
2006 bcopy(&now, &last, sizeof(struct timeval));
2008 // restart once per day (14400 seconds)
2009 current_t = time(NULL);
2010 if(current_t - started_t > 14400) exit(0);