ia64/xen-unstable

view xen/tools/symbols.c @ 15927:b7eb2bb9b625

IRQ injection changes for HVM PCI passthru.
Signed-off-by: Allen Kay <allen.m.kay@intel.com>
Signed-off-by: Guy Zana <guy@neocleus.com>
author kfraser@localhost.localdomain
date Tue Sep 18 16:09:19 2007 +0100 (2007-09-18)
parents 33242afccb32
children 8d5474a5c66b
line source
1 /* Generate assembler source containing symbol information
2 *
3 * Copyright 2002 by Kai Germaschewski
4 *
5 * This software may be used and distributed according to the terms
6 * of the GNU General Public License, incorporated herein by reference.
7 *
8 * Usage: nm -n vmlinux | scripts/symbols [--all-symbols] > symbols.S
9 *
10 * ChangeLog:
11 *
12 * (25/Aug/2004) Paulo Marques <pmarques@grupopie.com>
13 * Changed the compression method from stem compression to "table lookup"
14 * compression
15 *
16 * Table compression uses all the unused char codes on the symbols and
17 * maps these to the most used substrings (tokens). For instance, it might
18 * map char code 0xF7 to represent "write_" and then in every symbol where
19 * "write_" appears it can be replaced by 0xF7, saving 5 bytes.
20 * The used codes themselves are also placed in the table so that the
21 * decompresion can work without "special cases".
22 * Applied to kernel symbols, this usually produces a compression ratio
23 * of about 50%.
24 *
25 */
27 #define _GNU_SOURCE
29 #include <stdio.h>
30 #include <stdlib.h>
31 #include <string.h>
32 #include <stdint.h>
33 #include <ctype.h>
35 #define KSYM_NAME_LEN 127
38 struct sym_entry {
39 unsigned long long addr;
40 unsigned int len;
41 unsigned char *sym;
42 };
45 static struct sym_entry *table;
46 static unsigned int table_size, table_cnt;
47 static unsigned long long _stext, _etext, _sinittext, _einittext, _sextratext, _eextratext;
48 static int all_symbols = 0;
49 static char symbol_prefix_char = '\0';
51 int token_profit[0x10000];
53 /* the table that holds the result of the compression */
54 unsigned char best_table[256][2];
55 unsigned char best_table_len[256];
58 static void usage(void)
59 {
60 fprintf(stderr, "Usage: symbols [--all-symbols] [--symbol-prefix=<prefix char>] < in.map > out.S\n");
61 exit(1);
62 }
64 /*
65 * This ignores the intensely annoying "mapping symbols" found
66 * in ARM ELF files: $a, $t and $d.
67 */
68 static inline int is_arm_mapping_symbol(const char *str)
69 {
70 return str[0] == '$' && strchr("atd", str[1])
71 && (str[2] == '\0' || str[2] == '.');
72 }
74 static int read_symbol(FILE *in, struct sym_entry *s)
75 {
76 char str[500];
77 char *sym, stype;
78 int rc;
80 rc = fscanf(in, "%llx %c %499s\n", &s->addr, &stype, str);
81 if (rc != 3) {
82 if (rc != EOF) {
83 /* skip line */
84 fgets(str, 500, in);
85 }
86 return -1;
87 }
89 sym = str;
90 /* skip prefix char */
91 if (symbol_prefix_char && str[0] == symbol_prefix_char)
92 sym++;
94 /* Ignore most absolute/undefined (?) symbols. */
95 if (strcmp(sym, "_stext") == 0)
96 _stext = s->addr;
97 else if (strcmp(sym, "_etext") == 0)
98 _etext = s->addr;
99 else if (strcmp(sym, "_sinittext") == 0)
100 _sinittext = s->addr;
101 else if (strcmp(sym, "_einittext") == 0)
102 _einittext = s->addr;
103 else if (strcmp(sym, "_sextratext") == 0)
104 _sextratext = s->addr;
105 else if (strcmp(sym, "_eextratext") == 0)
106 _eextratext = s->addr;
107 else if (toupper((uint8_t)stype) == 'A')
108 {
109 /* Keep these useful absolute symbols */
110 if (strcmp(sym, "__kernel_syscall_via_break") &&
111 strcmp(sym, "__kernel_syscall_via_epc") &&
112 strcmp(sym, "__kernel_sigtramp") &&
113 strcmp(sym, "__gp"))
114 return -1;
116 }
117 else if (toupper((uint8_t)stype) == 'U' ||
118 is_arm_mapping_symbol(sym))
119 return -1;
120 /* exclude also MIPS ELF local symbols ($L123 instead of .L123) */
121 else if (str[0] == '$')
122 return -1;
124 /* include the type field in the symbol name, so that it gets
125 * compressed together */
126 s->len = strlen(str) + 1;
127 s->sym = malloc(s->len + 1);
128 strcpy((char *)s->sym + 1, str);
129 s->sym[0] = stype;
131 return 0;
132 }
134 static int symbol_valid(struct sym_entry *s)
135 {
136 /* Symbols which vary between passes. Passes 1 and 2 must have
137 * identical symbol lists. The symbols_* symbols below are only added
138 * after pass 1, they would be included in pass 2 when --all-symbols is
139 * specified so exclude them to get a stable symbol list.
140 */
141 static char *special_symbols[] = {
142 "symbols_addresses",
143 "symbols_num_syms",
144 "symbols_names",
145 "symbols_markers",
146 "symbols_token_table",
147 "symbols_token_index",
149 /* Exclude linker generated symbols which vary between passes */
150 "_SDA_BASE_", /* ppc */
151 "_SDA2_BASE_", /* ppc */
152 NULL };
153 int i;
154 int offset = 1;
156 /* skip prefix char */
157 if (symbol_prefix_char && *(s->sym + 1) == symbol_prefix_char)
158 offset++;
160 /* if --all-symbols is not specified, then symbols outside the text
161 * and inittext sections are discarded */
162 if (!all_symbols) {
163 if ((s->addr < _stext || s->addr > _etext)
164 && (s->addr < _sinittext || s->addr > _einittext)
165 && (s->addr < _sextratext || s->addr > _eextratext))
166 return 0;
167 /* Corner case. Discard any symbols with the same value as
168 * _etext _einittext or _eextratext; they can move between pass
169 * 1 and 2 when the symbols data are added. If these symbols
170 * move then they may get dropped in pass 2, which breaks the
171 * symbols rules.
172 */
173 if ((s->addr == _etext && strcmp((char*)s->sym + offset, "_etext")) ||
174 (s->addr == _einittext && strcmp((char*)s->sym + offset, "_einittext")) ||
175 (s->addr == _eextratext && strcmp((char*)s->sym + offset, "_eextratext")))
176 return 0;
177 }
179 /* Exclude symbols which vary between passes. */
180 if (strstr((char *)s->sym + offset, "_compiled."))
181 return 0;
183 for (i = 0; special_symbols[i]; i++)
184 if( strcmp((char *)s->sym + offset, special_symbols[i]) == 0 )
185 return 0;
187 return 1;
188 }
190 static void read_map(FILE *in)
191 {
192 while (!feof(in)) {
193 if (table_cnt >= table_size) {
194 table_size += 10000;
195 table = realloc(table, sizeof(*table) * table_size);
196 if (!table) {
197 fprintf(stderr, "out of memory\n");
198 exit (1);
199 }
200 }
201 if (read_symbol(in, &table[table_cnt]) == 0)
202 table_cnt++;
203 }
204 }
206 static void output_label(char *label)
207 {
208 if (symbol_prefix_char)
209 printf(".globl %c%s\n", symbol_prefix_char, label);
210 else
211 printf(".globl %s\n", label);
212 printf("\tALGN\n");
213 if (symbol_prefix_char)
214 printf("%c%s:\n", symbol_prefix_char, label);
215 else
216 printf("%s:\n", label);
217 }
219 /* uncompress a compressed symbol. When this function is called, the best table
220 * might still be compressed itself, so the function needs to be recursive */
221 static int expand_symbol(unsigned char *data, int len, char *result)
222 {
223 int c, rlen, total=0;
225 while (len) {
226 c = *data;
227 /* if the table holds a single char that is the same as the one
228 * we are looking for, then end the search */
229 if (best_table[c][0]==c && best_table_len[c]==1) {
230 *result++ = c;
231 total++;
232 } else {
233 /* if not, recurse and expand */
234 rlen = expand_symbol(best_table[c], best_table_len[c], result);
235 total += rlen;
236 result += rlen;
237 }
238 data++;
239 len--;
240 }
241 *result=0;
243 return total;
244 }
246 static void write_src(void)
247 {
248 unsigned int i, k, off;
249 unsigned int best_idx[256];
250 unsigned int *markers;
251 char buf[KSYM_NAME_LEN+1];
253 printf("#include <asm/types.h>\n");
254 printf("#if BITS_PER_LONG == 64\n");
255 printf("#define PTR .quad\n");
256 printf("#define ALGN .align 8\n");
257 printf("#else\n");
258 printf("#define PTR .long\n");
259 printf("#define ALGN .align 4\n");
260 printf("#endif\n");
262 printf(".data\n");
264 output_label("symbols_addresses");
265 for (i = 0; i < table_cnt; i++) {
266 printf("\tPTR\t%#llx\n", table[i].addr);
267 }
268 printf("\n");
270 output_label("symbols_num_syms");
271 printf("\tPTR\t%d\n", table_cnt);
272 printf("\n");
274 /* table of offset markers, that give the offset in the compressed stream
275 * every 256 symbols */
276 markers = (unsigned int *) malloc(sizeof(unsigned int) * ((table_cnt + 255) / 256));
278 output_label("symbols_names");
279 off = 0;
280 for (i = 0; i < table_cnt; i++) {
281 if ((i & 0xFF) == 0)
282 markers[i >> 8] = off;
284 printf("\t.byte 0x%02x", table[i].len);
285 for (k = 0; k < table[i].len; k++)
286 printf(", 0x%02x", table[i].sym[k]);
287 printf("\n");
289 off += table[i].len + 1;
290 }
291 printf("\n");
293 output_label("symbols_markers");
294 for (i = 0; i < ((table_cnt + 255) >> 8); i++)
295 printf("\tPTR\t%d\n", markers[i]);
296 printf("\n");
298 free(markers);
300 output_label("symbols_token_table");
301 off = 0;
302 for (i = 0; i < 256; i++) {
303 best_idx[i] = off;
304 expand_symbol(best_table[i], best_table_len[i], buf);
305 printf("\t.asciz\t\"%s\"\n", buf);
306 off += strlen(buf) + 1;
307 }
308 printf("\n");
310 output_label("symbols_token_index");
311 for (i = 0; i < 256; i++)
312 printf("\t.short\t%d\n", best_idx[i]);
313 printf("\n");
314 }
317 /* table lookup compression functions */
319 /* count all the possible tokens in a symbol */
320 static void learn_symbol(unsigned char *symbol, int len)
321 {
322 int i;
324 for (i = 0; i < len - 1; i++)
325 token_profit[ symbol[i] + (symbol[i + 1] << 8) ]++;
326 }
328 /* decrease the count for all the possible tokens in a symbol */
329 static void forget_symbol(unsigned char *symbol, int len)
330 {
331 int i;
333 for (i = 0; i < len - 1; i++)
334 token_profit[ symbol[i] + (symbol[i + 1] << 8) ]--;
335 }
337 /* remove all the invalid symbols from the table and do the initial token count */
338 static void build_initial_tok_table(void)
339 {
340 unsigned int i, pos;
342 pos = 0;
343 for (i = 0; i < table_cnt; i++) {
344 if ( symbol_valid(&table[i]) ) {
345 if (pos != i)
346 table[pos] = table[i];
347 learn_symbol(table[pos].sym, table[pos].len);
348 pos++;
349 }
350 }
351 table_cnt = pos;
352 }
354 static void *memmem_pvt(void *h, size_t hlen, void *n, size_t nlen)
355 {
356 char *p;
357 for (p = h; (p - (char *)h) <= (long)(hlen - nlen); p++)
358 if (!memcmp(p, n, nlen)) return p;
359 return NULL;
360 }
362 /* replace a given token in all the valid symbols. Use the sampled symbols
363 * to update the counts */
364 static void compress_symbols(unsigned char *str, int idx)
365 {
366 unsigned int i, len, size;
367 unsigned char *p1, *p2;
369 for (i = 0; i < table_cnt; i++) {
371 len = table[i].len;
372 p1 = table[i].sym;
374 /* find the token on the symbol */
375 p2 = memmem_pvt(p1, len, str, 2);
376 if (!p2) continue;
378 /* decrease the counts for this symbol's tokens */
379 forget_symbol(table[i].sym, len);
381 size = len;
383 do {
384 *p2 = idx;
385 p2++;
386 size -= (p2 - p1);
387 memmove(p2, p2 + 1, size);
388 p1 = p2;
389 len--;
391 if (size < 2) break;
393 /* find the token on the symbol */
394 p2 = memmem_pvt(p1, size, str, 2);
396 } while (p2);
398 table[i].len = len;
400 /* increase the counts for this symbol's new tokens */
401 learn_symbol(table[i].sym, len);
402 }
403 }
405 /* search the token with the maximum profit */
406 static int find_best_token(void)
407 {
408 int i, best, bestprofit;
410 bestprofit=-10000;
411 best = 0;
413 for (i = 0; i < 0x10000; i++) {
414 if (token_profit[i] > bestprofit) {
415 best = i;
416 bestprofit = token_profit[i];
417 }
418 }
419 return best;
420 }
422 /* this is the core of the algorithm: calculate the "best" table */
423 static void optimize_result(void)
424 {
425 int i, best;
427 /* using the '\0' symbol last allows compress_symbols to use standard
428 * fast string functions */
429 for (i = 255; i >= 0; i--) {
431 /* if this table slot is empty (it is not used by an actual
432 * original char code */
433 if (!best_table_len[i]) {
435 /* find the token with the breates profit value */
436 best = find_best_token();
438 /* place it in the "best" table */
439 best_table_len[i] = 2;
440 best_table[i][0] = best & 0xFF;
441 best_table[i][1] = (best >> 8) & 0xFF;
443 /* replace this token in all the valid symbols */
444 compress_symbols(best_table[i], i);
445 }
446 }
447 }
449 /* start by placing the symbols that are actually used on the table */
450 static void insert_real_symbols_in_table(void)
451 {
452 unsigned int i, j, c;
454 memset(best_table, 0, sizeof(best_table));
455 memset(best_table_len, 0, sizeof(best_table_len));
457 for (i = 0; i < table_cnt; i++) {
458 for (j = 0; j < table[i].len; j++) {
459 c = table[i].sym[j];
460 best_table[c][0]=c;
461 best_table_len[c]=1;
462 }
463 }
464 }
466 static void optimize_token_table(void)
467 {
468 build_initial_tok_table();
470 insert_real_symbols_in_table();
472 /* When valid symbol is not registered, exit to error */
473 if (!table_cnt) {
474 fprintf(stderr, "No valid symbol.\n");
475 exit(1);
476 }
478 optimize_result();
479 }
482 int main(int argc, char **argv)
483 {
484 if (argc >= 2) {
485 int i;
486 for (i = 1; i < argc; i++) {
487 if(strcmp(argv[i], "--all-symbols") == 0)
488 all_symbols = 1;
489 else if (strncmp(argv[i], "--symbol-prefix=", 16) == 0) {
490 char *p = &argv[i][16];
491 /* skip quote */
492 if ((*p == '"' && *(p+2) == '"') || (*p == '\'' && *(p+2) == '\''))
493 p++;
494 symbol_prefix_char = *p;
495 } else
496 usage();
497 }
498 } else if (argc != 1)
499 usage();
501 read_map(stdin);
502 optimize_token_table();
503 write_src();
505 return 0;
506 }