1 /* file.c - describe file type
2 *
3 * Copyright 2016 The Android Open Source Project
4 *
5 * See http://pubs.opengroup.org/onlinepubs/9699919799/utilities/file.html
6
7 USE_FILE(NEWTOY(file, "<1hL[!hL]", TOYFLAG_USR|TOYFLAG_BIN))
8
9 config FILE
10 bool "file"
11 default y
12 help
13 usage: file [-hL] [file...]
14
15 Examine the given files and describe their content types.
16
17 -h don't follow symlinks (default)
18 -L follow symlinks
19 */
20
21 #define FOR_file
22 #include "toys.h"
23
GLOBALS(int max_name_len;)24 GLOBALS(
25 int max_name_len;
26 )
27
28 // We don't trust elf.h to be there, and two codepaths for 32/64 is awkward
29 // anyway, so calculate struct offsets manually. (It's a fixed ABI.)
30 static void do_elf_file(int fd, struct stat *sb)
31 {
32 int endian = toybuf[5], bits = toybuf[4], i, j;
33 int64_t (*elf_int)(void *ptr, unsigned size);
34 // Values from include/linux/elf-em.h (plus arch/*/include/asm/elf.h)
35 // Names are linux/arch/ directory (sometimes before 32/64 bit merges)
36 struct {int val; char *name;} type[] = {{0x9026, "alpha"}, {93, "arc"},
37 {195, "arcv2"}, {40, "arm"}, {183, "arm64"}, {0x18ad, "avr32"},
38 {247, "bpf"}, {106, "blackfin"}, {140, "c6x"}, {23, "cell"}, {76, "cris"},
39 {0x5441, "frv"}, {46, "h8300"}, {164, "hexagon"}, {50, "ia64"},
40 {88, "m32r"}, {0x9041, "m32r"}, {4, "m68k"}, {174, "metag"},
41 {189, "microblaze"}, {0xbaab, "microblaze-old"}, {8, "mips"},
42 {10, "mips-old"}, {89, "mn10300"}, {0xbeef, "mn10300-old"}, {113, "nios2"},
43 {92, "openrisc"}, {0x8472, "openrisc-old"}, {15, "parisc"}, {20, "ppc"},
44 {21, "ppc64"}, {22, "s390"}, {0xa390, "s390-old"}, {135, "score"},
45 {42, "sh"}, {2, "sparc"}, {18, "sparc8+"}, {43, "sparc9"}, {188, "tile"},
46 {191, "tilegx"}, {3, "386"}, {6, "486"}, {62, "x86-64"}, {94, "xtensa"},
47 {0xabc7, "xtensa-old"}
48 };
49 int dynamic = 0;
50 int stripped = 1;
51 char *map;
52 off_t phoff, shoff;
53 int phentsize, phnum, shsize, shnum;
54
55 printf("ELF ");
56 elf_int = (endian==2) ? peek_be : peek_le;
57
58 // executable type
59 i = elf_int(toybuf+16, 2);
60 if (i == 1) printf("relocatable");
61 else if (i == 2) printf("executable");
62 else if (i == 3) printf("shared object");
63 else if (i == 4) printf("core dump");
64 else printf("(bad type %d)", i);
65 if (elf_int(toybuf+36+12*(bits==2), 4) & 0x8000) printf(" (fdpic)");
66 printf(", ");
67
68 // "64-bit"
69 if (bits == 1) printf("32-bit ");
70 else if (bits == 2) printf("64-bit ");
71 else {
72 printf("(bad class %d) ", bits);
73 bits = 0;
74 }
75
76 // "LSB"
77 if (endian == 1) printf("LSB ");
78 else if (endian == 2) printf("MSB ");
79 else {
80 printf("(bad endian %d) \n", endian);
81 endian = 0;
82 }
83
84 // e_machine, ala "x86", from big table above
85 j = elf_int(toybuf+18, 2);
86 for (i = 0; i<ARRAY_LEN(type); i++) if (j==type[i].val) break;
87 if (i<ARRAY_LEN(type)) printf("%s", type[i].name);
88 else printf("(unknown arch %d)", j);
89
90 bits--;
91 // If what we've seen so far doesn't seem consistent, bail.
92 if (!((bits&1)==bits && endian)) {
93 printf(", corrupt?\n");
94 return;
95 }
96
97 // Stash what we need from the header; it's okay to reuse toybuf after this.
98 phentsize = elf_int(toybuf+42+12*bits, 2);
99 phnum = elf_int(toybuf+44+12*bits, 2);
100 phoff = elf_int(toybuf+28+4*bits, 4+4*bits);
101 shsize = elf_int(toybuf+46+12*bits, 2);
102 shnum = elf_int(toybuf+48+12*bits, 2);
103 shoff = elf_int(toybuf+32+8*bits, 4+4*bits);
104
105 // With binutils, phentsize seems to only be non-zero if phnum is non-zero.
106 // Such ELF files are rare, but do exist. (Android's crtbegin files, say.)
107 if (phnum && (phentsize != 32+24*bits)) {
108 printf(", corrupt phentsize %d?\n", phentsize);
109 return;
110 }
111
112 map = xmmap(0, sb->st_size, PROT_READ, MAP_SHARED, fd, 0);
113
114 // We need to read the phdrs for dynamic vs static.
115 // (Note: fields got reordered for 64 bit)
116 for (i = 0; i<phnum; i++) {
117 char *phdr = map+phoff+i*phentsize;
118 int p_type = elf_int(phdr, 4);
119 long long p_offset, p_filesz;
120
121 if (p_type==2 /*PT_DYNAMIC*/) dynamic = 1;
122 if (p_type!=3 /*PT_INTERP*/ && p_type!=4 /*PT_NOTE*/) continue;
123
124 j = bits+1;
125 p_offset = elf_int(phdr+4*j, 4*j);
126 p_filesz = elf_int(phdr+16*j, 4*j);
127
128 if (p_type==3 /*PT_INTERP*/)
129 printf(", dynamic (%.*s)", (int)p_filesz, map+p_offset);
130 }
131 if (!dynamic) printf(", static");
132
133 // We need to read the shdrs for stripped/unstripped and any notes.
134 // Notes are in program headers *and* section headers, but some files don't
135 // contain program headers, so we prefer to check here.
136 // (Note: fields got reordered for 64 bit)
137 for (i = 0; i<shnum; i++) {
138 char *shdr = map+shoff+i*shsize;
139 int sh_type = elf_int(shdr+4, 4);
140 long sh_offset = elf_int(shdr+8+8*(bits+1), 4*(bits+1));
141 int sh_size = elf_int(shdr+8+12*(bits+1), 4);
142
143 if (sh_type == 2 /*SHT_SYMTAB*/) {
144 stripped = 0;
145 break;
146 } else if (sh_type == 7 /*SHT_NOTE*/) {
147 char *note = map+sh_offset;
148
149 // An ELF note is a sequence of entries, each consisting of an
150 // ndhr followed by n_namesz+n_descsz bytes of data (each of those
151 // rounded up to the next 4 bytes, without this being reflected in
152 // the header byte counts themselves).
153 while (sh_size >= 3*4) { // Don't try to read a truncated entry.
154 int n_namesz = elf_int(note, 4);
155 int n_descsz = elf_int(note+4, 4);
156 int n_type = elf_int(note+8, 4);
157 int notesz = 3*4 + ((n_namesz+3)&~3) + ((n_descsz+3)&~3);
158
159 if (n_namesz==4 && !memcmp(note+12, "GNU", 4)) {
160 if (n_type==3 /*NT_GNU_BUILD_ID*/) {
161 printf(", BuildID=");
162 for (j = 0; j < n_descsz; ++j) printf("%02x", note[16 + j]);
163 }
164 } else if (n_namesz==8 && !memcmp(note+12, "Android", 8)) {
165 if (n_type==1 /*.android.note.ident*/) {
166 printf(", for Android %d", (int)elf_int(note+20, 4));
167 if (n_descsz > 24)
168 printf(", built by NDK %.64s (%.64s)", note+24, note+24+64);
169 }
170 }
171
172 note += notesz;
173 sh_size -= notesz;
174 }
175 }
176 }
177 printf(", %sstripped", stripped ? "" : "not ");
178 xputc('\n');
179
180 munmap(map, sb->st_size);
181 }
182
do_regular_file(int fd,char * name,struct stat * sb)183 static void do_regular_file(int fd, char *name, struct stat *sb)
184 {
185 char *s;
186 int len = read(fd, s = toybuf, sizeof(toybuf)-256);
187 int magic;
188
189 if (len<0) perror_msg("%s", name);
190
191 if (len>40 && strstart(&s, "\177ELF")) do_elf_file(fd, sb);
192 else if (len>=8 && strstart(&s, "!<arch>\n")) xprintf("ar archive\n");
193 else if (len>28 && strstart(&s, "\x89PNG\x0d\x0a\x1a\x0a")) {
194 // PNG is big-endian: https://www.w3.org/TR/PNG/#7Integers-and-byte-order
195 int chunk_length = peek_be(s, 4);
196
197 xprintf("PNG image data");
198
199 // The IHDR chunk comes first: https://www.w3.org/TR/PNG/#11IHDR
200 s += 4;
201 if (chunk_length == 13 && strstart(&s, "IHDR")) {
202 // https://www.w3.org/TR/PNG/#6Colour-values
203 char *c = 0, *colors[] = {"grayscale", 0, "color RGB", "indexed color",
204 "grayscale with alpha", 0, "color RGBA"};
205
206 if (s[9]<ARRAY_LEN(colors)) c = colors[s[9]];
207 if (!c) c = "unknown";
208
209 xprintf(", %d x %d, %d-bit/%s, %sinterlaced", (int)peek_be(s, 4),
210 (int)peek_be(s+4, 4), s[8], c, s[12] ? "" : "non-");
211 }
212
213 xputc('\n');
214
215 // https://www.w3.org/Graphics/GIF/spec-gif89a.txt
216 } else if (len>16 && (strstart(&s, "GIF87a") || strstart(&s, "GIF89a")))
217 xprintf("GIF image data, %d x %d\n",
218 (int)peek_le(s, 2), (int)peek_le(s+8, 2));
219
220 // TODO: parsing JPEG for width/height is harder than GIF or PNG.
221 else if (len>32 && !memcmp(toybuf, "\xff\xd8", 2)) xputs("JPEG image data");
222
223 // https://docs.oracle.com/javase/specs/jvms/se7/html/jvms-4.html
224 else if (len>8 && strstart(&s, "\xca\xfe\xba\xbe"))
225 xprintf("Java class file, version %d.%d\n",
226 (int)peek_be(s+2, 2), (int)peek_be(s, 2));
227
228 // https://people.freebsd.org/~kientzle/libarchive/man/cpio.5.txt
229 // the lengths for cpio are size of header + 9 bytes, since any valid
230 // cpio archive ends with a record for "TARGET!!!"
231 else if (len>85 && strstart(&s, "07070")) {
232 char *cpioformat = "unknown type";
233
234 if (toybuf[5] == '7') cpioformat = "pre-SVR4 or odc";
235 else if (toybuf[5] == '1') cpioformat = "SVR4 with no CRC";
236 else if (toybuf[5] == '2') cpioformat = "SVR4 with CRC";
237 xprintf("ASCII cpio archive (%s)\n", cpioformat);
238 } else if (len>33 && (magic=peek(&s,2), magic==0143561 || magic==070707)) {
239 if (magic == 0143561) printf("byte-swapped ");
240 xprintf("cpio archive\n");
241 // tar archive (ustar/pax or gnu)
242 } else if (len>500 && !strncmp(s+257, "ustar", 5))
243 xprintf("POSIX tar archive%s\n", strncmp(s+262," ",2)?"":" (GNU)");
244 // zip/jar/apk archive, ODF/OOXML document, or such
245 else if (len>5 && strstart(&s, "PK\03\04")) {
246 int ver = toybuf[4];
247
248 xprintf("Zip archive data");
249 if (ver) xprintf(", requires at least v%d.%d to extract", ver/10, ver%10);
250 xputc('\n');
251 } else if (len>4 && strstart(&s, "BZh") && isdigit(*s))
252 xprintf("bzip2 compressed data, block size = %c00k\n", *s);
253 else if (len>10 && strstart(&s, "\x1f\x8b")) xputs("gzip compressed data");
254 else if (len>32 && !memcmp(s+1, "\xfa\xed\xfe", 3)) {
255 int bit = s[0]=='\xce'?32:64;
256 char *what;
257
258 xprintf("Mach-O %d-bit ", bit);
259
260 if (s[4] == 7) what = (bit==32)?"x86":"x86-";
261 else if (s[4] == 12) what = "arm";
262 else if (s[4] == 18) what = "ppc";
263 else what = NULL;
264 if (what) xprintf("%s%s ", what, (bit==32)?"":"64");
265 else xprintf("(bad arch %d) ", s[4]);
266
267 if (s[12] == 1) what = "object";
268 else if (s[12] == 2) what = "executable";
269 else if (s[12] == 6) what = "shared library";
270 else what = NULL;
271 if (what) xprintf("%s\n", what);
272 else xprintf("(bad type %d)\n", s[9]);
273 } else if (len>36 && !memcmp(s, "OggS\x00\x02", 6)) {
274 xprintf("Ogg data");
275 // https://wiki.xiph.org/MIMETypesCodecs
276 if (!memcmp(s+28, "CELT ", 8)) xprintf(", celt audio");
277 if (!memcmp(s+28, "CMML ", 8)) xprintf(", cmml text");
278 if (!memcmp(s+28, "BBCD\0", 5)) xprintf(", dirac video");
279 if (!memcmp(s+28, "\177FLAC", 5)) xprintf(", flac audio");
280 if (!memcmp(s+28, "\x8bJNG\r\n\x1a\n", 8)) xprintf(", jng video");
281 if (!memcmp(s+28, "\x80kate\0\0\0", 8)) xprintf(", kate text");
282 if (!memcmp(s+28, "OggMIDI\0", 8)) xprintf(", midi text");
283 if (!memcmp(s+28, "\x8aMNG\r\n\x1a\n", 8)) xprintf(", mng video");
284 if (!memcmp(s+28, "OpusHead", 8)) xprintf(", opus audio");
285 if (!memcmp(s+28, "PCM ", 8)) xprintf(", pcm audio");
286 if (!memcmp(s+28, "\x89PNG\r\n\x1a\n", 8)) xprintf(", png video");
287 if (!memcmp(s+28, "Speex ", 8)) xprintf(", speex audio");
288 if (!memcmp(s+28, "\x80theora", 7)) xprintf(", theora video");
289 if (!memcmp(s+28, "\x01vorbis", 7)) xprintf(", vorbis audio");
290 if (!memcmp(s+28, "YUV4MPEG", 8)) xprintf(", yuv4mpeg video");
291 xputc('\n');
292 } else if (len>12 && !memcmp(s, "\x00\x01\x00\x00", 4)) {
293 xputs("TrueType font");
294 } else if (len>12 && !memcmp(s, "ttcf\x00", 5)) {
295 xprintf("TrueType font collection, version %d, %d fonts\n",
296 (int)peek_be(s+4, 2), (int)peek_be(s+8, 4));
297 } else if (len>4 && !memcmp(s, "BC\xc0\xde", 4)) {
298 xputs("LLVM IR bitcode");
299 } else if (strstart(&s, "-----BEGIN CERTIFICATE-----")) {
300 xputs("PEM certificate");
301
302 // https://msdn.microsoft.com/en-us/library/windows/desktop/ms680547(v=vs.85).aspx
303 } else if (len>0x70 && !memcmp(s, "MZ", 2) &&
304 (magic=peek_le(s+0x3c,4))<len-4 && !memcmp(s+magic, "\x50\x45\0\0", 4)) {
305 xprintf("MS PE32%s executable %s", (peek_le(s+magic+24, 2)==0x20b)?"+":"",
306 (peek_le(s+magic+22, 2)&0x2000)?"(DLL) ":"");
307 if (peek_le(s+magic+20, 2)>70) {
308 char *types[] = {0, "native", "GUI", "console", "OS/2", "driver", "CE",
309 "EFI", "EFI boot", "EFI runtime", "EFI ROM", "XBOX", 0, "boot"};
310 int type = peek_le(s+magic+92, 2);
311 char *name = (type>0 && type<ARRAY_LEN(types))?types[type]:0;
312
313 xprintf("(%s) ", name?name:"unknown");
314 }
315 xprintf("%s\n", (peek_le(s+magic+4, 2)==0x14c)?"x86":"x86-64");
316 } else {
317 char *what = 0;
318 int i, bytes;
319
320 // If shell script, report which interpreter
321 if (len>3 && strstart(&s, "#!")) {
322 // Whitespace is allowed between the #! and the interpreter
323 while (isspace(*s)) s++;
324 if (strstart(&s, "/usr/bin/env")) while (isspace(*s)) s++;
325 for (what = s; (s-toybuf)<len && !isspace(*s); s++);
326 strcpy(s, " script");
327
328 // Distinguish ASCII text, UTF-8 text, or data
329 } else for (i = 0; i<len; ++i) {
330 if (!(isprint(toybuf[i]) || isspace(toybuf[i]))) {
331 wchar_t wc;
332 if ((bytes = utf8towc(&wc, s+i, len-i))>0 && wcwidth(wc)>=0) {
333 i += bytes-1;
334 if (!what) what = "UTF-8 text";
335 } else {
336 what = "data";
337 break;
338 }
339 }
340 }
341 xputs(what ? what : "ASCII text");
342 }
343 }
344
file_main(void)345 void file_main(void)
346 {
347 char **arg;
348
349 for (arg = toys.optargs; *arg; ++arg) {
350 int name_len = strlen(*arg);
351
352 if (name_len > TT.max_name_len) TT.max_name_len = name_len;
353 }
354
355 // Can't use loopfiles here because it doesn't call function when can't open
356 for (arg = toys.optargs; *arg; arg++) {
357 char *name = *arg, *what = "cannot open";
358 struct stat sb;
359 int fd = !strcmp(name, "-");
360
361 xprintf("%s: %*s", name, (int)(TT.max_name_len - strlen(name)), "");
362
363 if (fd || !((toys.optflags & FLAG_L) ? stat : lstat)(name, &sb)) {
364 if (fd || S_ISREG(sb.st_mode)) {
365 if (!sb.st_size) what = "empty";
366 else if ((fd = openro(name, O_RDONLY)) != -1) {
367 do_regular_file(fd, name, &sb);
368 if (fd) close(fd);
369 continue;
370 }
371 } else if (S_ISFIFO(sb.st_mode)) what = "fifo";
372 else if (S_ISBLK(sb.st_mode)) what = "block special";
373 else if (S_ISCHR(sb.st_mode)) what = "character special";
374 else if (S_ISDIR(sb.st_mode)) what = "directory";
375 else if (S_ISSOCK(sb.st_mode)) what = "socket";
376 else if (S_ISLNK(sb.st_mode)) what = "symbolic link";
377 else what = "unknown";
378 }
379
380 xputs(what);
381 }
382 }
383