xref: /freebsd/usr.bin/mkuzip/mkuzip.c (revision cd0d51baaa4509a1db83251a601d34404d20c990)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright (c) 2004-2016 Maxim Sobolev <sobomax@FreeBSD.org>
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  */
28 
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
31 
32 #include <sys/types.h>
33 #include <sys/endian.h>
34 #include <sys/param.h>
35 #include <sys/sysctl.h>
36 #include <sys/stat.h>
37 #include <sys/uio.h>
38 #include <netinet/in.h>
39 #include <assert.h>
40 #include <ctype.h>
41 #include <err.h>
42 #include <fcntl.h>
43 #include <pthread.h>
44 #include <signal.h>
45 #include <stdint.h>
46 #include <stdio.h>
47 #include <stdlib.h>
48 #include <string.h>
49 #include <unistd.h>
50 
51 #include "mkuzip.h"
52 #include "mkuz_cloop.h"
53 #include "mkuz_blockcache.h"
54 #include "mkuz_lzma.h"
55 #include "mkuz_zlib.h"
56 #include "mkuz_zstd.h"
57 #include "mkuz_blk.h"
58 #include "mkuz_cfg.h"
59 #include "mkuz_conveyor.h"
60 #include "mkuz_format.h"
61 #include "mkuz_fqueue.h"
62 #include "mkuz_time.h"
63 #include "mkuz_insize.h"
64 
65 #define DEFAULT_CLSTSIZE	16384
66 
67 enum UZ_ALGORITHM {
68 	UZ_ZLIB = 0,
69 	UZ_LZMA,
70 	UZ_ZSTD,
71 	UZ_INVALID
72 };
73 
74 static const struct mkuz_format uzip_fmts[] = {
75 	[UZ_ZLIB] = {
76 		.option = "zlib",
77 		.magic = CLOOP_MAGIC_ZLIB,
78 		.default_sufx = DEFAULT_SUFX_ZLIB,
79 		.f_compress_bound = mkuz_zlib_cbound,
80 		.f_init = mkuz_zlib_init,
81 		.f_compress = mkuz_zlib_compress,
82 	},
83 	[UZ_LZMA] = {
84 		.option = "lzma",
85 		.magic = CLOOP_MAGIC_LZMA,
86 		.default_sufx = DEFAULT_SUFX_LZMA,
87 		.f_compress_bound = mkuz_lzma_cbound,
88 		.f_init = mkuz_lzma_init,
89 		.f_compress = mkuz_lzma_compress,
90 	},
91 	[UZ_ZSTD] = {
92 		.option = "zstd",
93 		.magic = CLOOP_MAGIC_ZSTD,
94 		.default_sufx = DEFAULT_SUFX_ZSTD,
95 		.f_compress_bound = mkuz_zstd_cbound,
96 		.f_init = mkuz_zstd_init,
97 		.f_compress = mkuz_zstd_compress,
98 	},
99 };
100 
101 static struct mkuz_blk *readblock(int, u_int32_t);
102 static void usage(void);
103 static void cleanup(void);
104 
105 static char *cleanfile = NULL;
106 
107 static int
108 cmp_blkno(const struct mkuz_blk *bp, void *p)
109 {
110 	uint32_t *ap;
111 
112 	ap = (uint32_t *)p;
113 
114 	return (bp->info.blkno == *ap);
115 }
116 
117 int main(int argc, char **argv)
118 {
119 	struct mkuz_cfg cfs;
120 	char *oname;
121 	uint64_t *toc;
122 	int i, io, opt, tmp;
123 	struct {
124 		int en;
125 		FILE *f;
126 	} summary;
127 	struct iovec iov[2];
128 	uint64_t offset, last_offset;
129 	struct cloop_header hdr;
130 	struct mkuz_conveyor *cvp;
131         void *c_ctx;
132 	struct mkuz_blk_info *chit;
133 	size_t ncpusz, ncpu, magiclen;
134 	double st, et;
135 	enum UZ_ALGORITHM comp_alg;
136 	int comp_level;
137 
138 	st = getdtime();
139 
140 	ncpusz = sizeof(size_t);
141 	if (sysctlbyname("hw.ncpu", &ncpu, &ncpusz, NULL, 0) < 0) {
142 		ncpu = 1;
143 	} else if (ncpu > MAX_WORKERS_AUTO) {
144 		ncpu = MAX_WORKERS_AUTO;
145 	}
146 
147 	memset(&hdr, 0, sizeof(hdr));
148 	cfs.blksz = DEFAULT_CLSTSIZE;
149 	oname = NULL;
150 	cfs.verbose = 0;
151 	cfs.no_zcomp = 0;
152 	cfs.en_dedup = 0;
153 	summary.en = 0;
154 	summary.f = stderr;
155 	comp_alg = UZ_ZLIB;
156 	comp_level = USE_DEFAULT_LEVEL;
157 	cfs.nworkers = ncpu;
158 	struct mkuz_blk *iblk, *oblk;
159 
160 	while((opt = getopt(argc, argv, "A:C:o:s:vZdLSj:")) != -1) {
161 		switch(opt) {
162 		case 'A':
163 			for (tmp = UZ_ZLIB; tmp < UZ_INVALID; tmp++) {
164 				if (strcmp(uzip_fmts[tmp].option, optarg) == 0)
165 					break;
166 			}
167 			if (tmp == UZ_INVALID)
168 				errx(1, "invalid algorithm specified: %s",
169 				    optarg);
170 				/* Not reached */
171 			comp_alg = tmp;
172 			break;
173 		case 'C':
174 			comp_level = atoi(optarg);
175 			break;
176 		case 'o':
177 			oname = optarg;
178 			break;
179 
180 		case 's':
181 			tmp = atoi(optarg);
182 			if (tmp <= 0) {
183 				errx(1, "invalid cluster size specified: %s",
184 				    optarg);
185 				/* Not reached */
186 			}
187 			cfs.blksz = tmp;
188 			break;
189 
190 		case 'v':
191 			cfs.verbose = 1;
192 			break;
193 
194 		case 'Z':
195 			cfs.no_zcomp = 1;
196 			break;
197 
198 		case 'd':
199 			cfs.en_dedup = 1;
200 			break;
201 
202 		case 'L':
203 			comp_alg = UZ_LZMA;
204 			break;
205 
206 		case 'S':
207 			summary.en = 1;
208 			summary.f = stdout;
209 			break;
210 
211 		case 'j':
212 			tmp = atoi(optarg);
213 			if (tmp <= 0) {
214 				errx(1, "invalid number of compression threads"
215                                     " specified: %s", optarg);
216 				/* Not reached */
217 			}
218 			cfs.nworkers = tmp;
219 			break;
220 
221 		default:
222 			usage();
223 			/* Not reached */
224 		}
225 	}
226 	argc -= optind;
227 	argv += optind;
228 
229 	if (argc != 1) {
230 		usage();
231 		/* Not reached */
232 	}
233 
234 	cfs.handler = &uzip_fmts[comp_alg];
235 
236 	magiclen = strlcpy(hdr.magic, cfs.handler->magic, sizeof(hdr.magic));
237 	assert(magiclen < sizeof(hdr.magic));
238 
239 	if (cfs.en_dedup != 0) {
240 		/*
241 		 * Dedupe requires a version 3 format.  Don't downgrade newer
242 		 * formats.
243 		 */
244 		if (hdr.magic[CLOOP_OFS_VERSN] == CLOOP_MAJVER_2)
245 			hdr.magic[CLOOP_OFS_VERSN] = CLOOP_MAJVER_3;
246 		hdr.magic[CLOOP_OFS_COMPR] =
247 		    tolower(hdr.magic[CLOOP_OFS_COMPR]);
248 	}
249 
250 	if (cfs.blksz % DEV_BSIZE != 0)
251 		errx(1, "cluster size should be multiple of %d", DEV_BSIZE);
252 
253 	cfs.cbound_blksz = cfs.handler->f_compress_bound(cfs.blksz);
254 	if (cfs.cbound_blksz > MAXPHYS)
255 		errx(1, "maximal compressed cluster size %zu greater than MAXPHYS %zu",
256 		    cfs.cbound_blksz, (size_t)MAXPHYS);
257 
258 	c_ctx = cfs.handler->f_init(&comp_level);
259 	cfs.comp_level = comp_level;
260 
261 	cfs.iname = argv[0];
262 	if (oname == NULL) {
263 		asprintf(&oname, "%s%s", cfs.iname, cfs.handler->default_sufx);
264 		if (oname == NULL) {
265 			err(1, "can't allocate memory");
266 			/* Not reached */
267 		}
268 	}
269 
270 	signal(SIGHUP, exit);
271 	signal(SIGINT, exit);
272 	signal(SIGTERM, exit);
273 	signal(SIGXCPU, exit);
274 	signal(SIGXFSZ, exit);
275 	atexit(cleanup);
276 
277 	cfs.fdr = open(cfs.iname, O_RDONLY);
278 	if (cfs.fdr < 0) {
279 		err(1, "open(%s)", cfs.iname);
280 		/* Not reached */
281 	}
282 	cfs.isize = mkuz_get_insize(&cfs);
283 	if (cfs.isize < 0) {
284 		errx(1, "can't determine input image size");
285 		/* Not reached */
286 	}
287 	hdr.nblocks = cfs.isize / cfs.blksz;
288 	if ((cfs.isize % cfs.blksz) != 0) {
289 		if (cfs.verbose != 0)
290 			fprintf(stderr, "file size is not multiple "
291 			"of %d, padding data\n", cfs.blksz);
292 		hdr.nblocks++;
293 	}
294 	toc = mkuz_safe_malloc((hdr.nblocks + 1) * sizeof(*toc));
295 
296 	/*
297 	 * Initialize last+1 entry with non-heap trash.  If final padding is
298 	 * added later, it may or may not be overwritten with an offset
299 	 * representing the length of the final compressed block.  If not,
300 	 * initialize to a defined value.
301 	 */
302 	toc[hdr.nblocks] = 0;
303 
304 	cfs.fdw = open(oname, (cfs.en_dedup ? O_RDWR : O_WRONLY) | O_TRUNC | O_CREAT,
305 		   S_IRWXU | S_IRGRP | S_IXGRP | S_IROTH | S_IXOTH);
306 	if (cfs.fdw < 0) {
307 		err(1, "open(%s)", oname);
308 		/* Not reached */
309 	}
310 	cleanfile = oname;
311 
312 	/* Prepare header that we will write later when we have index ready. */
313 	iov[0].iov_base = (char *)&hdr;
314 	iov[0].iov_len = sizeof(hdr);
315 	iov[1].iov_base = (char *)toc;
316 	iov[1].iov_len = (hdr.nblocks + 1) * sizeof(*toc);
317 	offset = iov[0].iov_len + iov[1].iov_len;
318 
319 	/* Reserve space for header */
320 	lseek(cfs.fdw, offset, SEEK_SET);
321 
322 	if (cfs.verbose != 0) {
323 		fprintf(stderr, "data size %ju bytes, number of clusters "
324 		    "%u, index length %zu bytes\n", cfs.isize,
325 		    hdr.nblocks, iov[1].iov_len);
326 	}
327 
328 	cvp = mkuz_conveyor_ctor(&cfs);
329 
330 	last_offset = 0;
331         iblk = oblk = NULL;
332 	for(i = io = 0; iblk != MKUZ_BLK_EOF; i++) {
333 		iblk = readblock(cfs.fdr, cfs.blksz);
334 		mkuz_fqueue_enq(cvp->wrk_queue, iblk);
335 		if (iblk != MKUZ_BLK_EOF &&
336 		    (i < (cfs.nworkers * ITEMS_PER_WORKER))) {
337 			continue;
338 		}
339 drain:
340 		oblk = mkuz_fqueue_deq_when(cvp->results, cmp_blkno, &io);
341 		assert(oblk->info.blkno == (unsigned)io);
342 		oblk->info.offset = offset;
343 		chit = NULL;
344 		if (cfs.en_dedup != 0 && oblk->info.len > 0) {
345 			chit = mkuz_blkcache_regblock(cfs.fdw, oblk);
346 			/*
347 			 * There should be at least one non-empty block
348 			 * between us and the backref'ed offset, otherwise
349 			 * we won't be able to parse that sequence correctly
350 			 * as it would be indistinguishible from another
351 			 * empty block.
352 			 */
353 			if (chit != NULL && chit->offset == last_offset) {
354 				chit = NULL;
355 			}
356 		}
357 		if (chit != NULL) {
358 			toc[io] = htobe64(chit->offset);
359 			oblk->info.len = 0;
360 		} else {
361 			if (oblk->info.len > 0 && write(cfs.fdw, oblk->data,
362 			    oblk->info.len) < 0) {
363 				err(1, "write(%s)", oname);
364 				/* Not reached */
365 			}
366 			toc[io] = htobe64(offset);
367 			last_offset = offset;
368 			offset += oblk->info.len;
369 		}
370 		if (cfs.verbose != 0) {
371 			fprintf(stderr, "cluster #%d, in %u bytes, "
372 			    "out len=%lu offset=%lu", io, cfs.blksz,
373 			    (u_long)oblk->info.len, (u_long)be64toh(toc[io]));
374 			if (chit != NULL) {
375 				fprintf(stderr, " (backref'ed to #%d)",
376 				    chit->blkno);
377 			}
378 			fprintf(stderr, "\n");
379 		}
380 		free(oblk);
381 		io += 1;
382 		if (iblk == MKUZ_BLK_EOF) {
383 			if (io < i)
384 				goto drain;
385 			/* Last block, see if we need to add some padding */
386 			if ((offset % DEV_BSIZE) == 0)
387 				continue;
388 			oblk = mkuz_blk_ctor(DEV_BSIZE - (offset % DEV_BSIZE));
389 			oblk->info.blkno = io;
390 			oblk->info.len = oblk->alen;
391 			if (cfs.verbose != 0) {
392 				fprintf(stderr, "padding data with %lu bytes "
393 				    "so that file size is multiple of %d\n",
394 				    (u_long)oblk->alen, DEV_BSIZE);
395 			}
396 			mkuz_fqueue_enq(cvp->results, oblk);
397 			goto drain;
398 		}
399 	}
400 
401 	close(cfs.fdr);
402 
403 	if (cfs.verbose != 0 || summary.en != 0) {
404 		et = getdtime();
405 		fprintf(summary.f, "compressed data to %ju bytes, saved %lld "
406 		    "bytes, %.2f%% decrease, %.2f bytes/sec.\n", offset,
407 		    (long long)(cfs.isize - offset),
408 		    100.0 * (long long)(cfs.isize - offset) /
409 		    (float)cfs.isize, (float)cfs.isize / (et - st));
410 	}
411 
412 	/* Convert to big endian */
413 	hdr.blksz = htonl(cfs.blksz);
414 	hdr.nblocks = htonl(hdr.nblocks);
415 	/* Write headers into pre-allocated space */
416 	lseek(cfs.fdw, 0, SEEK_SET);
417 	if (writev(cfs.fdw, iov, 2) < 0) {
418 		err(1, "writev(%s)", oname);
419 		/* Not reached */
420 	}
421 	cleanfile = NULL;
422 	close(cfs.fdw);
423 
424 	exit(0);
425 }
426 
427 static struct mkuz_blk *
428 readblock(int fd, u_int32_t clstsize)
429 {
430 	int numread;
431 	struct mkuz_blk *rval;
432 	static int blockcnt;
433 	off_t cpos;
434 
435 	rval = mkuz_blk_ctor(clstsize);
436 
437 	rval->info.blkno = blockcnt;
438 	blockcnt += 1;
439 	cpos = lseek(fd, 0, SEEK_CUR);
440 	if (cpos < 0) {
441 		err(1, "readblock: lseek() failed");
442 		/* Not reached */
443 	}
444 	rval->info.offset = cpos;
445 
446 	numread = read(fd, rval->data, clstsize);
447 	if (numread < 0) {
448 		err(1, "readblock: read() failed");
449 		/* Not reached */
450 	}
451 	if (numread == 0) {
452 		free(rval);
453 		return MKUZ_BLK_EOF;
454 	}
455 	rval->info.len = numread;
456 	return rval;
457 }
458 
459 static void
460 usage(void)
461 {
462 
463 	fprintf(stderr, "usage: mkuzip [-vZdLS] [-o outfile] [-s cluster_size] "
464 	    "[-j ncompr] infile\n");
465 	exit(1);
466 }
467 
468 void *
469 mkuz_safe_malloc(size_t size)
470 {
471 	void *retval;
472 
473 	retval = malloc(size);
474 	if (retval == NULL) {
475 		err(1, "can't allocate memory");
476 		/* Not reached */
477 	}
478 	return retval;
479 }
480 
481 void *
482 mkuz_safe_zmalloc(size_t size)
483 {
484 	void *retval;
485 
486 	retval = mkuz_safe_malloc(size);
487 	bzero(retval, size);
488 	return retval;
489 }
490 
491 static void
492 cleanup(void)
493 {
494 
495 	if (cleanfile != NULL)
496 		unlink(cleanfile);
497 }
498 
499 int
500 mkuz_memvcmp(const void *memory, unsigned char val, size_t size)
501 {
502     const u_char *mm;
503 
504     mm = (const u_char *)memory;
505     return (*mm == val) && memcmp(mm, mm + 1, size - 1) == 0;
506 }
507