1 /*****************************************************************************
2  * RRDtool 1.2.30  Copyright by Tobi Oetiker, 1997-2009
3  *****************************************************************************
4  * rrd_fetch.c  read date from an rrd to use for further processing
5  *****************************************************************************
6  * $Id: rrd_fetch.c 1735 2009-01-19 14:29:11Z oetiker $
7  * $Log$
8  * Revision 1.8  2004/05/18 18:53:03  oetiker
9  * big spell checking patch -- slif@bellsouth.net
10  *
11  * Revision 1.7  2003/11/11 19:46:21  oetiker
12  * replaced time_value with rrd_time_value as MacOS X introduced a struct of that name in their standard headers
13  *
14  * Revision 1.6  2003/01/16 23:27:54  oetiker
15  * fix border condition in rra selection of rrd_fetch
16  * -- Stanislav Sinyagin <ssinyagin@yahoo.com>
17  *
18  * Revision 1.5  2002/06/23 22:29:40  alex
19  * Added "step=1800" and such to "DEF"
20  * Cleaned some of the signed vs. unsigned problems
21  *
22  * Revision 1.4  2002/02/01 20:34:49  oetiker
23  * fixed version number and date/time
24  *
25  * Revision 1.3  2001/12/24 06:51:49  alex
26  * A patch of size 44Kbytes... in short:
27  *
28  * Found and repaired the off-by-one error in rrd_fetch_fn().
29  * As a result I had to remove the hacks in rrd_fetch_fn(),
30  * rrd_tool.c, vdef_calc(), data_calc(), data_proc() and
31  * reduce_data().  There may be other places which I didn't
32  * find so be careful.
33  *
34  * Enhanced debugging in rrd_fetch_fn(), it shows the RRA selection
35  * process.
36  *
37  * Added the ability to print VDEF timestamps.  At the moment it
38  * is a hack, I needed it now to fix the off-by-one error.
39  * If the format string is "%c" (and nothing else!), the time
40  * will be printed by both ctime() and as a long int.
41  *
42  * Moved some code around (slightly altering it) from rrd_graph()
43  *   initializing     now in rrd_graph_init()
44  *   options parsing  now in rrd_graph_options()
45  *   script parsing   now in rrd_graph_script()
46  *
47  * Revision 1.2  2001/12/17 12:48:43  oetiker
48  * fix overflow error ...
49  *
50  * Revision 1.1.1.1  2001/02/25 22:25:05  oetiker
51  * checkin
52  *
53  *****************************************************************************/
54 
55 #include "rrd_tool.h"
56 
57 #include "rrd_is_thread_safe.h"
58 /*#define DEBUG*/
59 
60 int
rrd_fetch(int argc,char ** argv,time_t * start,time_t * end,unsigned long * step,unsigned long * ds_cnt,char *** ds_namv,rrd_value_t ** data)61 rrd_fetch(int argc,
62 	  char **argv,
63 	  time_t         *start,
64 	  time_t         *end,       /* which time frame do you want ?
65 				      * will be changed to represent reality */
66 	  unsigned long  *step,      /* which stepsize do you want?
67 				      * will be changed to represent reality */
68 	  unsigned long  *ds_cnt,    /* number of data sources in file */
69 	  char           ***ds_namv,   /* names of data sources */
70 	  rrd_value_t    **data)     /* two dimensional array containing the data */
71 {
72 
73 
74     long     step_tmp =1;
75     time_t   start_tmp=0, end_tmp=0;
76     const char *cf;
77 
78     struct rrd_time_value start_tv, end_tv;
79     char     *parsetime_error = NULL;
80     optind = 0; opterr = 0;  /* initialize getopt */
81 
82     /* init start and end time */
83     parsetime("end-24h", &start_tv);
84     parsetime("now", &end_tv);
85 
86     while (1){
87 	static struct option long_options[] =
88 	{
89 	    {"resolution",      required_argument, 0, 'r'},
90 	    {"start",      required_argument, 0, 's'},
91 	    {"end",      required_argument, 0, 'e'},
92 	    {0,0,0,0}
93 	};
94 	int option_index = 0;
95 	int opt;
96 	opt = getopt_long(argc, argv, "r:s:e:",
97 			  long_options, &option_index);
98 
99 	if (opt == EOF)
100 	    break;
101 
102 	switch(opt) {
103 	case 's':
104             if ((parsetime_error = parsetime(optarg, &start_tv))) {
105                 rrd_set_error( "start time: %s", parsetime_error );
106                 return -1;
107 	    }
108 	    break;
109 	case 'e':
110             if ((parsetime_error = parsetime(optarg, &end_tv))) {
111                 rrd_set_error( "end time: %s", parsetime_error );
112                 return -1;
113 	    }
114 	    break;
115 	case 'r':
116 	    step_tmp = atol(optarg);
117 	    break;
118 	case '?':
119 	    rrd_set_error("unknown option '-%c'",optopt);
120 	    return(-1);
121 	}
122     }
123 
124 
125     if (proc_start_end(&start_tv,&end_tv,&start_tmp,&end_tmp) == -1){
126 	return -1;
127     }
128 
129 
130     if (start_tmp < 3600*24*365*10){
131 	rrd_set_error("the first entry to fetch should be after 1980");
132 	return(-1);
133     }
134 
135     if (end_tmp < start_tmp) {
136 	rrd_set_error("start (%ld) should be less than end (%ld)", start_tmp, end_tmp);
137 	return(-1);
138     }
139 
140     *start = start_tmp;
141     *end = end_tmp;
142 
143     if (step_tmp < 1) {
144 	rrd_set_error("step must be >= 1 second");
145 	return -1;
146     }
147     *step = step_tmp;
148 
149     if (optind + 1 >= argc){
150 	rrd_set_error("not enough arguments");
151 	return -1;
152     }
153 
154     cf = argv[optind+1];
155 
156     if (rrd_fetch_r(argv[optind],cf,start,end,step,ds_cnt,ds_namv,data) == -1)
157 	return(-1);
158     return (0);
159 }
160 
161 int
rrd_fetch_r(const char * filename,const char * cf,time_t * start,time_t * end,unsigned long * step,unsigned long * ds_cnt,char *** ds_namv,rrd_value_t ** data)162 rrd_fetch_r(
163     const char           *filename,  /* name of the rrd */
164     const char           *cf,        /* which consolidation function ?*/
165     time_t         *start,
166     time_t         *end,       /* which time frame do you want ?
167                                 * will be changed to represent reality */
168     unsigned long  *step,      /* which stepsize do you want?
169                                 * will be changed to represent reality */
170     unsigned long  *ds_cnt,    /* number of data sources in file */
171     char           ***ds_namv, /* names of data_sources */
172     rrd_value_t    **data)     /* two dimensional array containing the data */
173 {
174     enum     cf_en cf_idx;
175 
176     if ((int)(cf_idx=cf_conv(cf)) == -1 ){
177         return -1;
178     }
179 
180     return (rrd_fetch_fn(filename,cf_idx,start,end,step,ds_cnt,ds_namv,data));
181 }
182 
183 int
rrd_fetch_fn(const char * filename,enum cf_en cf_idx,time_t * start,time_t * end,unsigned long * step,unsigned long * ds_cnt,char *** ds_namv,rrd_value_t ** data)184 rrd_fetch_fn(
185     const char     *filename,  /* name of the rrd */
186     enum cf_en     cf_idx,         /* which consolidation function ?*/
187     time_t         *start,
188     time_t         *end,       /* which time frame do you want ?
189 			        * will be changed to represent reality */
190     unsigned long  *step,      /* which stepsize do you want?
191 				* will be changed to represent reality */
192     unsigned long  *ds_cnt,    /* number of data sources in file */
193     char           ***ds_namv,   /* names of data_sources */
194     rrd_value_t    **data)     /* two dimensional array containing the data */
195 {
196     long           i,ii;
197     FILE           *in_file;
198     time_t         cal_start,cal_end, rra_start_time,rra_end_time;
199     long  best_full_rra=0, best_part_rra=0, chosen_rra=0, rra_pointer=0;
200     long  best_full_step_diff=0, best_part_step_diff=0, tmp_step_diff=0, tmp_match=0, best_match=0;
201     long  full_match, rra_base;
202     long           start_offset, end_offset;
203     int            first_full = 1;
204     int            first_part = 1;
205     rrd_t     rrd;
206     rrd_value_t    *data_ptr;
207     unsigned long  rows;
208 
209 #ifdef DEBUG
210 fprintf(stderr,"Entered rrd_fetch_fn() searching for the best match\n");
211 fprintf(stderr,"Looking for: start %10lu end %10lu step %5lu\n",
212 						*start,*end,*step);
213 #endif
214 
215     if(rrd_open(filename,&in_file,&rrd, RRD_READONLY)==-1)
216 	return(-1);
217 
218     /* when was the really last update of this file ? */
219 
220     if (((*ds_namv) = (char **) malloc(rrd.stat_head->ds_cnt * sizeof(char*)))==NULL){
221 	rrd_set_error("malloc fetch ds_namv array");
222 	rrd_free(&rrd);
223 	fclose(in_file);
224 	return(-1);
225     }
226 
227     for(i=0;(unsigned long)i<rrd.stat_head->ds_cnt;i++){
228 	if ((((*ds_namv)[i]) = malloc(sizeof(char) * DS_NAM_SIZE))==NULL){
229 	    rrd_set_error("malloc fetch ds_namv entry");
230 	    rrd_free(&rrd);
231 	    free(*ds_namv);
232 	    fclose(in_file);
233 	    return(-1);
234 	}
235 	strncpy((*ds_namv)[i],rrd.ds_def[i].ds_nam,DS_NAM_SIZE-1);
236 	(*ds_namv)[i][DS_NAM_SIZE-1]='\0';
237 
238     }
239 
240     /* find the rra which best matches the requirements */
241     for(i=0;(unsigned)i<rrd.stat_head->rra_cnt;i++){
242 	if(cf_conv(rrd.rra_def[i].cf_nam) == cf_idx){
243 
244 	    cal_end = (rrd.live_head->last_up - (rrd.live_head->last_up
245 			  % (rrd.rra_def[i].pdp_cnt
246 			     * rrd.stat_head->pdp_step)));
247 	    cal_start = (cal_end
248 			 - (rrd.rra_def[i].pdp_cnt
249 			    * rrd.rra_def[i].row_cnt
250 			    * rrd.stat_head->pdp_step));
251 
252 	    full_match = *end -*start;
253 #ifdef DEBUG
254 fprintf(stderr,"Considering: start %10lu end %10lu step %5lu ",
255 							cal_start,cal_end,
256 			rrd.stat_head->pdp_step * rrd.rra_def[i].pdp_cnt);
257 #endif
258  	    /* we need step difference in either full or partial case */
259  	    tmp_step_diff = labs(*step - (rrd.stat_head->pdp_step
260 					   * rrd.rra_def[i].pdp_cnt));
261 	    /* best full match */
262 	    if(cal_start <= *start){
263 		if (first_full || (tmp_step_diff < best_full_step_diff)){
264 		    first_full=0;
265 		    best_full_step_diff = tmp_step_diff;
266 		    best_full_rra=i;
267 #ifdef DEBUG
268 fprintf(stderr,"best full match so far\n");
269 #endif
270 		} else {
271 #ifdef DEBUG
272 fprintf(stderr,"full match, not best\n");
273 #endif
274 		}
275 
276 	    } else {
277 		/* best partial match */
278 		tmp_match = full_match;
279 		if (cal_start>*start)
280 		    tmp_match -= (cal_start-*start);
281 		if (first_part ||
282                     (best_match < tmp_match) ||
283                     (best_match == tmp_match &&
284                      tmp_step_diff < best_part_step_diff)){
285 #ifdef DEBUG
286 fprintf(stderr,"best partial so far\n");
287 #endif
288 		    first_part=0;
289 		    best_match = tmp_match;
290 		    best_part_step_diff = tmp_step_diff;
291 		    best_part_rra =i;
292 		} else {
293 #ifdef DEBUG
294 fprintf(stderr,"partial match, not best\n");
295 #endif
296 		}
297 	    }
298 	}
299     }
300 
301     /* lets see how the matching went. */
302     if (first_full==0)
303 	chosen_rra = best_full_rra;
304     else if (first_part==0)
305 	chosen_rra = best_part_rra;
306     else {
307 	rrd_set_error("the RRD does not contain an RRA matching the chosen CF");
308 	rrd_free(&rrd);
309 	fclose(in_file);
310 	return(-1);
311     }
312 
313     /* set the wish parameters to their real values */
314     *step = rrd.stat_head->pdp_step * rrd.rra_def[chosen_rra].pdp_cnt;
315     *start -= (*start % *step);
316     *end += (*step - *end % *step);
317     rows = (*end - *start) / *step + 1;
318 
319 #ifdef DEBUG
320     fprintf(stderr,"We found:    start %10lu end %10lu step %5lu rows  %lu\n",
321 						*start,*end,*step,rows);
322 #endif
323 
324 /* Start and end are now multiples of the step size.  The amount of
325 ** steps we want is (end-start)/step and *not* an extra one.
326 ** Reasoning:  if step is s and we want to graph from t to t+s,
327 ** we need exactly ((t+s)-t)/s rows.  The row to collect from the
328 ** database is the one with time stamp (t+s) which means t to t+s.
329 */
330     *ds_cnt =   rrd.stat_head->ds_cnt;
331     if (((*data) = malloc(*ds_cnt * rows * sizeof(rrd_value_t)))==NULL){
332 	rrd_set_error("malloc fetch data area");
333 	for (i=0;(unsigned long)i<*ds_cnt;i++)
334 	      free((*ds_namv)[i]);
335 	free(*ds_namv);
336 	rrd_free(&rrd);
337 	fclose(in_file);
338 	return(-1);
339     }
340 
341     data_ptr=(*data);
342 
343     /* find base address of rra */
344     rra_base=ftell(in_file);
345     for(i=0;i<chosen_rra;i++)
346 	rra_base += ( *ds_cnt
347 		      * rrd.rra_def[i].row_cnt
348 		      * sizeof(rrd_value_t));
349 
350     /* find start and end offset */
351     rra_end_time = (rrd.live_head->last_up
352 		    - (rrd.live_head->last_up % *step));
353     rra_start_time = (rra_end_time
354 		 - ( *step * (rrd.rra_def[chosen_rra].row_cnt-1)));
355     /* here's an error by one if we don't be careful */
356     start_offset =(long)(*start + *step - rra_start_time) / (long)*step;
357     end_offset = (long)(rra_end_time - *end ) / (long)*step;
358 #ifdef DEBUG
359     fprintf(stderr,"rra_start %lu, rra_end %lu, start_off %li, end_off %li\n",
360 	    rra_start_time,rra_end_time,start_offset,end_offset);
361 #endif
362 
363     /* fill the gap at the start if needs be */
364     if (*start <= rra_end_time && *end >= rra_start_time - *step){
365 
366         if (start_offset <= 0)
367             rra_pointer = rrd.rra_ptr[chosen_rra].cur_row+1;
368         else
369             rra_pointer = rrd.rra_ptr[chosen_rra].cur_row+1+start_offset;
370 
371         rra_pointer = rra_pointer % (signed) rrd.rra_def[chosen_rra].row_cnt;
372 
373         if(fseek(in_file,(rra_base
374 		   + (rra_pointer
375 		      * *ds_cnt
376 		      * sizeof(rrd_value_t))),SEEK_SET) != 0){
377             rrd_set_error("seek error in RRA");
378             for (i=0;(unsigned)i<*ds_cnt;i++)
379                 free((*ds_namv)[i]);
380             free(*ds_namv);
381             rrd_free(&rrd);
382             free(*data);
383             *data = NULL;
384             fclose(in_file);
385             return(-1);
386 
387         }
388 #ifdef DEBUG
389         fprintf(stderr,"First Seek: rra_base %lu rra_pointer %lu\n",
390                 rra_base, rra_pointer);
391 #endif
392     }
393 
394     /* step trough the array */
395 
396     for (i=start_offset;
397 	 i< (signed)rrd.rra_def[chosen_rra].row_cnt - end_offset;
398 	 i++){
399 	/* no valid data yet */
400 	if (i<0) {
401 #ifdef DEBUG
402 	    fprintf(stderr,"pre fetch %li -- ",i);
403 #endif
404 	    for(ii=0;(unsigned)ii<*ds_cnt;ii++){
405 		*(data_ptr++) = DNAN;
406 #ifdef DEBUG
407 		fprintf(stderr,"%10.2f ",*(data_ptr-1));
408 #endif
409 	    }
410 	}
411 	/* past the valid data area */
412 	else if (i >= (signed)rrd.rra_def[chosen_rra].row_cnt) {
413 #ifdef DEBUG
414 	    fprintf(stderr,"post fetch %li -- ",i);
415 #endif
416 	    for(ii=0;(unsigned)ii<*ds_cnt;ii++){
417 		*(data_ptr++) = DNAN;
418 #ifdef DEBUG
419 		fprintf(stderr,"%10.2f ",*(data_ptr-1));
420 #endif
421 	    }
422 	} else {
423 	    /* OK we are inside the valid area but the pointer has to
424 	     * be wrapped*/
425 	    if (rra_pointer >= (signed)rrd.rra_def[chosen_rra].row_cnt) {
426 		rra_pointer -= rrd.rra_def[chosen_rra].row_cnt;
427 		if(fseek(in_file,(rra_base+rra_pointer
428 			       * *ds_cnt
429 			       * sizeof(rrd_value_t)),SEEK_SET) != 0){
430 		    rrd_set_error("wrap seek in RRA did fail");
431 		    for (ii=0;(unsigned)ii<*ds_cnt;ii++)
432 			free((*ds_namv)[ii]);
433 		    free(*ds_namv);
434 		    rrd_free(&rrd);
435 		    free(*data);
436 		    *data = NULL;
437 		    fclose(in_file);
438 		    return(-1);
439 		}
440 #ifdef DEBUG
441 		fprintf(stderr,"wrap seek ...\n");
442 #endif
443 	    }
444 
445 	    if(fread(data_ptr,
446 		     sizeof(rrd_value_t),
447 		     *ds_cnt,in_file) != rrd.stat_head->ds_cnt){
448 		rrd_set_error("fetching cdp from rra");
449 		for (ii=0;(unsigned)ii<*ds_cnt;ii++)
450 		    free((*ds_namv)[ii]);
451 		free(*ds_namv);
452 		rrd_free(&rrd);
453 		free(*data);
454 		*data = NULL;
455 		fclose(in_file);
456 		return(-1);
457 	    }
458 #ifdef DEBUG
459 	    fprintf(stderr,"post fetch %li -- ",i);
460 	    for(ii=0;ii<*ds_cnt;ii++)
461 		fprintf(stderr,"%10.2f ",*(data_ptr+ii));
462 #endif
463 	    data_ptr += *ds_cnt;
464 	    rra_pointer ++;
465 	}
466 #ifdef DEBUG
467 	    fprintf(stderr,"\n");
468 #endif
469 
470     }
471     rrd_free(&rrd);
472     fclose(in_file);
473     return(0);
474 }
475