]> sigrok.org Git - libsigrok.git/blob - src/input/vcd.c
input: vcd: refactor parsing a bit
[libsigrok.git] / src / input / vcd.c
1 /*
2  * This file is part of the libsigrok project.
3  *
4  * Copyright (C) 2012 Petteri Aimonen <jpa@sr.mail.kapsi.fi>
5  * Copyright (C) 2014 Bert Vermeulen <bert@biot.com>
6  *
7  * This program is free software: you can redistribute it and/or modify
8  * it under the terms of the GNU General Public License as published by
9  * the Free Software Foundation, either version 3 of the License, or
10  * (at your option) any later version.
11  *
12  * This program is distributed in the hope that it will be useful,
13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
15  * GNU General Public License for more details.
16  *
17  * You should have received a copy of the GNU General Public License
18  * along with this program.  If not, see <http://www.gnu.org/licenses/>.
19  */
20
21 /* The VCD input module has the following options:
22  *
23  * numchannels: Maximum number of channels to use. The channels are
24  *              detected in the same order as they are listed
25  *              in the $var sections of the VCD file.
26  *
27  * skip:        Allows skipping until given timestamp in the file.
28  *              This can speed up analyzing of long captures.
29  *            
30  *              Value < 0: Skip until first timestamp listed in
31  *              the file. (default)
32  *
33  *              Value = 0: Do not skip, instead generate samples
34  *              beginning from timestamp 0.
35  *
36  *              Value > 0: Start at the given timestamp.
37  *
38  * downsample:  Divide the samplerate by the given factor.
39  *              This can speed up analyzing of long captures.
40  *
41  * compress:    Compress idle periods longer than this value.
42  *              This can speed up analyzing of long captures.
43  *              Default 0 = don't compress.
44  *
45  * Based on Verilog standard IEEE Std 1364-2001 Version C
46  *
47  * Supported features:
48  * - $var with 'wire' and 'reg' types of scalar variables
49  * - $timescale definition for samplerate
50  * - multiple character variable identifiers
51  *
52  * Most important unsupported features:
53  * - vector variables (bit vectors etc.)
54  * - analog, integer and real number variables
55  * - $dumpvars initial value declaration
56  * - $scope namespaces
57  * - more than 64 channels
58  */
59
60 #include <config.h>
61 #include <stdlib.h>
62 #include <glib.h>
63 #include <stdio.h>
64 #include <string.h>
65 #include <libsigrok/libsigrok.h>
66 #include "libsigrok-internal.h"
67
68 #define LOG_PREFIX "input/vcd"
69
70 #define DEFAULT_NUM_CHANNELS 8
71 #define CHUNKSIZE (1024 * 1024)
72
73 struct context {
74         gboolean started;
75         gboolean got_header;
76         uint64_t samplerate;
77         unsigned int maxchannels;
78         unsigned int channelcount;
79         int downsample;
80         unsigned compress;
81         int64_t skip;
82         gboolean skip_until_end;
83         GSList *channels;
84         size_t bytes_per_sample;
85         size_t samples_in_buffer;
86         uint8_t *buffer;
87         uint8_t *current_levels;
88 };
89
90 struct vcd_channel {
91         gchar *name;
92         gchar *identifier;
93 };
94
95 /*
96  * Reads a single VCD section from input file and parses it to name/contents.
97  * e.g. $timescale 1ps $end  => "timescale" "1ps"
98  */
99 static gboolean parse_section(GString *buf, gchar **name, gchar **contents)
100 {
101         GString *sname, *scontent;
102         gboolean status;
103         unsigned int pos;
104
105         *name = *contents = NULL;
106         status = FALSE;
107         pos = 0;
108
109         /* Skip any initial white-space. */
110         while (pos < buf->len && g_ascii_isspace(buf->str[pos]))
111                 pos++;
112
113         /* Section tag should start with $. */
114         if (buf->str[pos++] != '$')
115                 return FALSE;
116
117         sname = g_string_sized_new(32);
118         scontent = g_string_sized_new(128);
119
120         /* Read the section tag. */
121         while (pos < buf->len && !g_ascii_isspace(buf->str[pos]))
122                 g_string_append_c(sname, buf->str[pos++]);
123
124         /* Skip whitespace before content. */
125         while (pos < buf->len && g_ascii_isspace(buf->str[pos]))
126                 pos++;
127
128         /* Read the content. */
129         while (pos < buf->len - 4 && strncmp(buf->str + pos, "$end", 4))
130                 g_string_append_c(scontent, buf->str[pos++]);
131
132         if (sname->len && pos < buf->len - 4 && !strncmp(buf->str + pos, "$end", 4)) {
133                 status = TRUE;
134                 pos += 4;
135                 while (pos < buf->len && g_ascii_isspace(buf->str[pos]))
136                         pos++;
137                 g_string_erase(buf, 0, pos);
138         }
139
140         *name = g_string_free(sname, !status);
141         *contents = g_string_free(scontent, !status);
142         if (*contents)
143                 g_strchomp(*contents);
144
145         return status;
146 }
147
148 static void free_channel(void *data)
149 {
150         struct vcd_channel *vcd_ch = data;
151         g_free(vcd_ch->name);
152         g_free(vcd_ch->identifier);
153         g_free(vcd_ch);
154 }
155
156 /* Remove empty parts from an array returned by g_strsplit. */
157 static void remove_empty_parts(gchar **parts)
158 {
159         gchar **src = parts;
160         gchar **dest = parts;
161         while (*src != NULL) {
162                 if (**src != '\0')
163                         *dest++ = *src;
164                 src++;
165         }
166
167         *dest = NULL;
168 }
169
170 /*
171  * Parse VCD header to get values for context structure.
172  * The context structure should be zeroed before calling this.
173  */
174 static gboolean parse_header(const struct sr_input *in, GString *buf)
175 {
176         struct vcd_channel *vcd_ch;
177         uint64_t p, q;
178         struct context *inc;
179         gboolean status;
180         gchar *name, *contents, **parts;
181
182         inc = in->priv;
183         name = contents = NULL;
184         status = FALSE;
185         while (parse_section(buf, &name, &contents)) {
186                 sr_dbg("Section '%s', contents '%s'.", name, contents);
187
188                 if (g_strcmp0(name, "enddefinitions") == 0) {
189                         status = TRUE;
190                         break;
191                 } else if (g_strcmp0(name, "timescale") == 0) {
192                         /*
193                          * The standard allows for values 1, 10 or 100
194                          * and units s, ms, us, ns, ps and fs.
195                          */
196                         if (sr_parse_period(contents, &p, &q) == SR_OK) {
197                                 inc->samplerate = q / p;
198                                 if (q % p != 0) {
199                                         /* Does not happen unless time value is non-standard */
200                                         sr_warn("Inexact rounding of samplerate, %" PRIu64 " / %" PRIu64 " to %" PRIu64 " Hz.",
201                                                 q, p, inc->samplerate);
202                                 }
203
204                                 sr_dbg("Samplerate: %" PRIu64, inc->samplerate);
205                         } else {
206                                 sr_err("Parsing timescale failed.");
207                         }
208                 } else if (g_strcmp0(name, "var") == 0) {
209                         /* Format: $var type size identifier reference [opt. index] $end */
210                         unsigned int length;
211
212                         parts = g_strsplit_set(contents, " \r\n\t", 0);
213                         remove_empty_parts(parts);
214                         length = g_strv_length(parts);
215
216                         if (length != 4 && length != 5)
217                                 sr_warn("$var section should have 4 or 5 items");
218                         else if (g_strcmp0(parts[0], "reg") != 0 && g_strcmp0(parts[0], "wire") != 0)
219                                 sr_info("Unsupported signal type: '%s'", parts[0]);
220                         else if (strtol(parts[1], NULL, 10) != 1)
221                                 sr_info("Unsupported signal size: '%s'", parts[1]);
222                         else if (inc->channelcount >= inc->maxchannels)
223                                 sr_warn("Skipping '%s%s' because only %d channels requested.",
224                                         parts[3], parts[4] ? : "", inc->maxchannels);
225                         else {
226                                 vcd_ch = g_malloc(sizeof(struct vcd_channel));
227                                 vcd_ch->identifier = g_strdup(parts[2]);
228                                 if (length == 4)
229                                         vcd_ch->name = g_strdup(parts[3]);
230                                 else
231                                         vcd_ch->name = g_strconcat(parts[3], parts[4], NULL);
232
233                                 sr_info("Channel %d is '%s' identified by '%s'.",
234                                                 inc->channelcount, vcd_ch->name, vcd_ch->identifier);
235
236                                 inc->channels = g_slist_append(inc->channels, vcd_ch);
237                                 inc->channelcount++;
238                         }
239
240                         g_strfreev(parts);
241                 }
242
243                 g_free(name);
244                 name = NULL;
245                 g_free(contents);
246                 contents = NULL;
247         }
248         g_free(name);
249         g_free(contents);
250
251         /*
252          * Compute how many bytes each sample will have and initialize the
253          * current levels. The current levels will be updated whenever VCD
254          * has changes.
255          */
256         inc->bytes_per_sample = (inc->channelcount + 7) / 8;
257         inc->current_levels = g_malloc0(inc->bytes_per_sample);
258
259         inc->got_header = status;
260
261         return status;
262 }
263
264 static int format_match(GHashTable *metadata)
265 {
266         GString *buf, *tmpbuf;
267         gboolean status;
268         gchar *name, *contents;
269
270         buf = g_hash_table_lookup(metadata, GINT_TO_POINTER(SR_INPUT_META_HEADER));
271         tmpbuf = g_string_new_len(buf->str, buf->len);
272
273         /*
274          * If we can parse the first section correctly,
275          * then it is assumed to be a VCD file.
276          */
277         status = parse_section(tmpbuf, &name, &contents);
278         g_string_free(tmpbuf, TRUE);
279         g_free(name);
280         g_free(contents);
281
282         return status ? SR_OK : SR_ERR;
283 }
284
285 /* Send all accumulated bytes from inc->buffer. */
286 static void send_buffer(const struct sr_input *in)
287 {
288         struct context *inc;
289         struct sr_datafeed_packet packet;
290         struct sr_datafeed_logic logic;
291
292         inc = in->priv;
293
294         if (inc->samples_in_buffer == 0)
295                 return;
296
297         packet.type = SR_DF_LOGIC;
298         packet.payload = &logic;
299         logic.unitsize = inc->bytes_per_sample;
300         logic.data = inc->buffer;
301         logic.length = inc->bytes_per_sample * inc->samples_in_buffer;
302         sr_session_send(in->sdi, &packet);
303         inc->samples_in_buffer = 0;
304 }
305
306 /*
307  * Add N copies of the current sample to buffer.
308  * When the buffer fills up, automatically send it.
309  */
310 static void add_samples(const struct sr_input *in, size_t count)
311 {
312         struct context *inc;
313         size_t samples_per_chunk;
314         size_t space_left, i;
315         uint8_t *p;
316
317         inc = in->priv;
318         samples_per_chunk = CHUNKSIZE / inc->bytes_per_sample;
319
320         while (count) {
321                 space_left = samples_per_chunk - inc->samples_in_buffer;
322
323                 if (space_left > count)
324                         space_left = count;
325
326                 p = inc->buffer + inc->samples_in_buffer * inc->bytes_per_sample;
327                 for (i = 0; i < space_left; i++) {
328                         memcpy(p, inc->current_levels, inc->bytes_per_sample);
329                         p += inc->bytes_per_sample;
330                         inc->samples_in_buffer++;
331                         count--;
332                 }
333
334                 if (inc->samples_in_buffer == samples_per_chunk)
335                         send_buffer(in);
336         }
337 }
338
339 /* Set the channel level depending on the identifier and parsed value. */
340 static void process_bit(struct context *inc, char *identifier, unsigned int bit)
341 {
342         GSList *l;
343         struct vcd_channel *vcd_ch;
344         unsigned int j;
345
346         for (j = 0, l = inc->channels; j < inc->channelcount && l; j++, l = l->next) {
347                 vcd_ch = l->data;
348                 if (g_strcmp0(identifier, vcd_ch->identifier) == 0) {
349                         /* Found our channel. */
350                         size_t byte_idx = (j / 8);
351                         size_t bit_idx = j - 8 * byte_idx;
352                         if (bit)
353                                 inc->current_levels[byte_idx] |= (uint8_t)1 << bit_idx;
354                         else
355                                 inc->current_levels[byte_idx] &= ~((uint8_t)1 << bit_idx);
356                         break;
357                 }
358         }
359         if (j == inc->channelcount)
360                 sr_dbg("Did not find channel for identifier '%s'.", identifier);
361 }
362
363 /* Parse a set of lines from the data section. */
364 static void parse_contents(const struct sr_input *in, char *data)
365 {
366         struct context *inc;
367         uint64_t timestamp, prev_timestamp;
368         unsigned int bit, i;
369         char **tokens;
370
371         inc = in->priv;
372         prev_timestamp = 0;
373
374         /* Read one space-delimited token at a time. */
375         tokens = g_strsplit_set(data, " \t\r\n", 0);
376         remove_empty_parts(tokens);
377         for (i = 0; tokens[i]; i++) {
378                 if (inc->skip_until_end) {
379                         if (!strcmp(tokens[i], "$end")) {
380                                 /* Done with unhandled/unknown section. */
381                                 inc->skip_until_end = FALSE;
382                                 break;
383                         }
384                 }
385                 if (tokens[i][0] == '#' && g_ascii_isdigit(tokens[i][1])) {
386                         /* Numeric value beginning with # is a new timestamp value */
387                         timestamp = strtoull(tokens[i] + 1, NULL, 10);
388
389                         if (inc->downsample > 1)
390                                 timestamp /= inc->downsample;
391
392                         /*
393                          * Skip < 0 => skip until first timestamp.
394                          * Skip = 0 => don't skip
395                          * Skip > 0 => skip until timestamp >= skip.
396                          */
397                         if (inc->skip < 0) {
398                                 inc->skip = timestamp;
399                                 prev_timestamp = timestamp;
400                         } else if (inc->skip > 0 && timestamp < (uint64_t)inc->skip) {
401                                 prev_timestamp = inc->skip;
402                         } else if (timestamp == prev_timestamp) {
403                                 /* Ignore repeated timestamps (e.g. sigrok outputs these) */
404                         } else {
405                                 if (inc->compress != 0 && timestamp - prev_timestamp > inc->compress) {
406                                         /* Compress long idle periods */
407                                         prev_timestamp = timestamp - inc->compress;
408                                 }
409
410                                 sr_dbg("New timestamp: %" PRIu64, timestamp);
411
412                                 /* Generate samples from prev_timestamp up to timestamp - 1. */
413                                 add_samples(in, timestamp - prev_timestamp);
414                                 prev_timestamp = timestamp;
415                         }
416                 } else if (tokens[i][0] == '$' && tokens[i][1] != '\0') {
417                         /*
418                          * This is probably a $dumpvars, $comment or similar.
419                          * $dump* contain useful data.
420                          */
421                         if (g_strcmp0(tokens[i], "$dumpvars") == 0
422                                         || g_strcmp0(tokens[i], "$dumpon") == 0
423                                         || g_strcmp0(tokens[i], "$dumpoff") == 0
424                                         || g_strcmp0(tokens[i], "$end") == 0) {
425                                 /* Ignore, parse contents as normally. */
426                         } else {
427                                 /* Ignore this and future lines until $end. */
428                                 inc->skip_until_end = TRUE;
429                                 break;
430                         }
431                 } else if (strchr("bBrR", tokens[i][0]) != NULL) {
432                         sr_dbg("Vector values not supported yet");
433                         if (!tokens[++i])
434                                 /* No tokens left, bail out */
435                                 break;
436                         else
437                                 /* Process next token */
438                                 continue;
439                 } else if (strchr("01xXzZ", tokens[i][0]) != NULL) {
440                         char *identifier;
441
442                         /* A new 1-bit sample value */
443                         bit = (tokens[i][0] == '1');
444
445                         /*
446                          * The identifier is either the next character, or, if
447                          * there was whitespace after the bit, the next token.
448                          */
449                         if (tokens[i][1] == '\0') {
450                                 if (!tokens[++i]) {
451                                         sr_dbg("Identifier missing!");
452                                         break;
453                                 }
454                                 identifier = tokens[i];
455                         } else {
456                                 identifier = tokens[i] + 1;
457                         }
458                         process_bit(inc, identifier, bit);
459                 } else {
460                         sr_warn("Skipping unknown token '%s'.", tokens[i]);
461                 }
462         }
463         g_strfreev(tokens);
464 }
465
466 static int init(struct sr_input *in, GHashTable *options)
467 {
468         int num_channels, i;
469         char name[16];
470         struct context *inc;
471
472         num_channels = g_variant_get_int32(g_hash_table_lookup(options, "numchannels"));
473         if (num_channels < 1) {
474                 sr_err("Invalid value for numchannels: must be at least 1.");
475                 return SR_ERR_ARG;
476         }
477         inc = in->priv = g_malloc0(sizeof(struct context));
478         inc->maxchannels = num_channels;
479
480         inc->downsample = g_variant_get_int32(g_hash_table_lookup(options, "downsample"));
481         if (inc->downsample < 1)
482                 inc->downsample = 1;
483
484         inc->compress = g_variant_get_int32(g_hash_table_lookup(options, "compress"));
485         inc->skip = g_variant_get_int32(g_hash_table_lookup(options, "skip"));
486         inc->skip /= inc->downsample;
487
488         in->sdi = g_malloc0(sizeof(struct sr_dev_inst));
489         in->priv = inc;
490
491         inc->buffer = g_malloc(CHUNKSIZE);
492
493         for (i = 0; i < num_channels; i++) {
494                 snprintf(name, 16, "%d", i);
495                 sr_channel_new(in->sdi, i, SR_CHANNEL_LOGIC, TRUE, name);
496         }
497
498         return SR_OK;
499 }
500
501 static gboolean have_header(GString *buf)
502 {
503         unsigned int pos;
504         char *p;
505
506         if (!(p = g_strstr_len(buf->str, buf->len, "$enddefinitions")))
507                 return FALSE;
508         pos = p - buf->str + 15;
509         while (pos < buf->len - 4 && g_ascii_isspace(buf->str[pos]))
510                 pos++;
511         if (!strncmp(buf->str + pos, "$end", 4))
512                 return TRUE;
513
514         return FALSE;
515 }
516
517 static int process_buffer(struct sr_input *in)
518 {
519         struct sr_datafeed_packet packet;
520         struct sr_datafeed_meta meta;
521         struct sr_config *src;
522         struct context *inc;
523         uint64_t samplerate;
524         char *p;
525
526         inc = in->priv;
527         if (!inc->started) {
528                 std_session_send_df_header(in->sdi, LOG_PREFIX);
529
530                 packet.type = SR_DF_META;
531                 packet.payload = &meta;
532                 samplerate = inc->samplerate / inc->downsample;
533                 src = sr_config_new(SR_CONF_SAMPLERATE, g_variant_new_uint64(samplerate));
534                 meta.config = g_slist_append(NULL, src);
535                 sr_session_send(in->sdi, &packet);
536                 g_slist_free(meta.config);
537                 sr_config_free(src);
538
539                 inc->started = TRUE;
540         }
541
542         while ((p = g_strrstr_len(in->buf->str, in->buf->len, "\n"))) {
543                 *p = '\0';
544                 g_strstrip(in->buf->str);
545                 if (in->buf->str[0] != '\0')
546                         parse_contents(in, in->buf->str);
547                 g_string_erase(in->buf, 0, p - in->buf->str + 1);
548         }
549
550         return SR_OK;
551 }
552
553 static int receive(struct sr_input *in, GString *buf)
554 {
555         struct context *inc;
556         int ret;
557
558         g_string_append_len(in->buf, buf->str, buf->len);
559
560         inc = in->priv;
561         if (!inc->got_header) {
562                 if (!have_header(in->buf))
563                         return SR_OK;
564                 if (!parse_header(in, in->buf))
565                         /* There was a header in there, but it was malformed. */
566                         return SR_ERR;
567
568                 in->sdi_ready = TRUE;
569                 /* sdi is ready, notify frontend. */
570                 return SR_OK;
571         }
572
573         ret = process_buffer(in);
574
575         return ret;
576 }
577
578 static int end(struct sr_input *in)
579 {
580         struct sr_datafeed_packet packet;
581         struct context *inc;
582         int ret;
583
584         inc = in->priv;
585
586         if (in->sdi_ready)
587                 ret = process_buffer(in);
588         else
589                 ret = SR_OK;
590
591         /* Send any samples that haven't been sent yet. */
592         send_buffer(in);
593
594         if (inc->started) {
595                 packet.type = SR_DF_END;
596                 sr_session_send(in->sdi, &packet);
597         }
598
599         return ret;
600 }
601
602 static void cleanup(struct sr_input *in)
603 {
604         struct context *inc;
605
606         inc = in->priv;
607         g_slist_free_full(inc->channels, free_channel);
608         g_free(inc->buffer);
609         inc->buffer = NULL;
610         g_free(inc->current_levels);
611         inc->current_levels = NULL;
612 }
613
614 static struct sr_option options[] = {
615         { "numchannels", "Number of channels", "Number of channels", NULL, NULL },
616         { "skip", "Skip", "Skip until timestamp", NULL, NULL },
617         { "downsample", "Downsample", "Divide samplerate by factor", NULL, NULL },
618         { "compress", "Compress", "Compress idle periods longer than this value", NULL, NULL },
619         ALL_ZERO
620 };
621
622 static const struct sr_option *get_options(void)
623 {
624         if (!options[0].def) {
625                 options[0].def = g_variant_ref_sink(g_variant_new_int32(DEFAULT_NUM_CHANNELS));
626                 options[1].def = g_variant_ref_sink(g_variant_new_int32(-1));
627                 options[2].def = g_variant_ref_sink(g_variant_new_int32(1));
628                 options[3].def = g_variant_ref_sink(g_variant_new_int32(0));
629         }
630
631         return options;
632 }
633
634 SR_PRIV struct sr_input_module input_vcd = {
635         .id = "vcd",
636         .name = "VCD",
637         .desc = "Value Change Dump",
638         .exts = (const char*[]){"vcd", NULL},
639         .metadata = { SR_INPUT_META_HEADER | SR_INPUT_META_REQUIRED },
640         .options = get_options,
641         .format_match = format_match,
642         .init = init,
643         .receive = receive,
644         .end = end,
645         .cleanup = cleanup,
646 };